)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":16708,"name":"Kendall Nelson","display_name":"Kendall (diablo_rojo)","email":"kennelson11@gmail.com","username":"kjnelson"},"change_message_id":"18a9055f023a6389fbc275536d9c5f1247882725","unresolved":false,"context_lines":[{"line_number":10,"context_line":"disaster recovery scenario when an entire host can be failed over to"},{"line_number":11,"context_line":"a secondary site."},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"At the Mitaka mid-cycle meetup, Cinder team decided that after Cheesecake"},{"line_number":14,"context_line":"is implemented, Tiramisu will be the next step. Tiramisu is the code name"},{"line_number":15,"context_line":"for a new set of replication APIs with more granularity. It gives tenant"},{"line_number":16,"context_line":"more control on what should be replicated together, i.e., a volume or a"},{"line_number":17,"context_line":"group of volumes."}],"source_content_type":"text/x-gerrit-commit-message","patch_set":10,"id":"fa0719c6_49741167","line":14,"range":{"start_line":13,"start_character":64,"end_line":14,"end_character":73},"updated":"2016-03-23 15:25:40.000000000","message":"Don\u0027t think its that big a deal, but these lines are over 72 characters.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"}],"specs/mitaka/replication-cg.rst":[{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"91deec01d05280913cdbd375f383bc94229cc39a","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Add the following CG APIs that support replication."},{"line_number":60,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"1a26ad4f_e208bc93","line":57,"updated":"2015-10-30 01:02:47.000000000","message":"I feel that the best way to do this when you try and replicate a single volume that is in a CG is to just say \u0027No\u0027. There is a risk that the volume selected (maybe by mistake) is a member of a VERY large CG that could cause all sorts of capacity issues on the destination array and that array could be overwhelmed when you force the CG to be replicated. \n\nWhen CG replication is ready should it do a capacity check on the destination array to ensure that capacity is available for the replicated CG?","commit_id":"b8b1e904c192539c20c08daef5a73da7e6bb805e"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"21cee7384c7076f7b2a7307901e8de9b43054246","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Add the following CG APIs that support replication."},{"line_number":60,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"1a26ad4f_5539faf0","line":57,"in_reply_to":"1a26ad4f_e208bc93","updated":"2015-11-02 05:08:25.000000000","message":"In Pure\u0027s replication implementation, the remote volume is created when the primary volume is created.  Other drivers are doing the same.  So you should do capacity check at volume creation time.\n\nI don\u0027t think enable_replication should do capacity check.  The remote volume is created at volume creation time, not at the enable_replication time.\n\nI don\u0027t see why replicating a large CG is different when you trigger it by a CG name vs one of the volumes.","commit_id":"b8b1e904c192539c20c08daef5a73da7e6bb805e"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"6b6310ba961001446114bb0223a9be8e094318df","unresolved":false,"context_lines":[{"line_number":41,"context_line":""},{"line_number":42,"context_line":"Phase 1: Integrate CG and replication at the volume manager level (without"},{"line_number":43,"context_line":"changing API). Change the following manager functions to handle CG. Also"},{"line_number":44,"context_line":"equivalent driver methods will be provided to handle CG."},{"line_number":45,"context_line":""},{"line_number":46,"context_line":"* enable_replication"},{"line_number":47,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_26719f44","line":44,"updated":"2015-11-02 17:53:05.000000000","message":"I think we can do this implicitly without a ton of change.  For example allow the creation of a type that specifies CG and Replication and leave the \u0027work\" up to the driver.\n\nI\u0027m leery of going down the path of cramming a bunch of rigid functionality in the base code for something that is not what I consider a first class feature and more importantly varies greatly in semantics between one vendor and another.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8bf5819281d00a60ef4da40b843c04e03ad29197","unresolved":false,"context_lines":[{"line_number":41,"context_line":""},{"line_number":42,"context_line":"Phase 1: Integrate CG and replication at the volume manager level (without"},{"line_number":43,"context_line":"changing API). Change the following manager functions to handle CG. Also"},{"line_number":44,"context_line":"equivalent driver methods will be provided to handle CG."},{"line_number":45,"context_line":""},{"line_number":46,"context_line":"* enable_replication"},{"line_number":47,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_d3dc1e75","line":44,"in_reply_to":"1a26ad4f_26719f44","updated":"2015-11-03 01:42:03.000000000","message":"See later comments.  We probably don\u0027t need phase 1 as there\u0027s no interest in it.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"b0cb6b4291d1560da0e15b7612eb47ae29224d4c","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Change capability reporting and scheduler to support different"},{"line_number":60,"context_line":"level of grouping."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_6fb9bf78","line":57,"updated":"2015-11-02 09:43:59.000000000","message":"I think my only concern here is an operator requesting a replication of a volume not knowing it is in a large CG. I think, at this phase, we should should just say \u0027No\u0027 when a single volume is set to enable_replication if it is already part of a CG.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"b8a94b4e4cb13fa2ffb8d96697cba4220e0dbacb","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Change capability reporting and scheduler to support different"},{"line_number":60,"context_line":"level of grouping."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_8db3fee8","line":57,"in_reply_to":"1a26ad4f_1245cbb1","updated":"2015-11-02 13:00:42.000000000","message":"I agree on the capacity issue, but my main point is if selecting a volume for replication that is in a CG you should not allow that to force the whole CG to be replicated. You should error with an \u0027in CG\u0027 message. Explicit CG replication should be the rule, not implicit","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8bf5819281d00a60ef4da40b843c04e03ad29197","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Change capability reporting and scheduler to support different"},{"line_number":60,"context_line":"level of grouping."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_88a51548","line":57,"in_reply_to":"1a26ad4f_1556032a","updated":"2015-11-03 01:42:03.000000000","message":"Yes, the local and remote volumes are created at volume creation time.  I think the concern is that you want to perform an action on one volume but you get it on a group of volumes.\n\nThis is just one way of moving forward with a small step without introducing new API\u0027s.  If no one likes it, we\u0027ll just wait until everyone is ready for the big change then.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"db4fbcb537976b5d46630028dcd71e3ef169cc9b","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Change capability reporting and scheduler to support different"},{"line_number":60,"context_line":"level of grouping."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_1a8e3079","line":57,"in_reply_to":"1a26ad4f_46413ba8","updated":"2015-11-02 19:54:02.000000000","message":"I tend to lean towards what Simon is suggesting, at least for phase 1 of this, unless we skip directly to phase 2 \u0026 3.\n\nThe more I think about it, the more I don\u0027t like the implicit behavior of specifying a volume but sometimes getting a whole group replicated or failed over.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"05840cccfb2b6b8c805abd35b299795c8d0a1cd4","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Change capability reporting and scheduler to support different"},{"line_number":60,"context_line":"level of grouping."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_1245cbb1","line":57,"in_reply_to":"1a26ad4f_6fb9bf78","updated":"2015-11-02 12:40:43.000000000","message":"Going back to your original comment in patch set 1, your concern is regarding the capacity.  Capacity check should happen at volume creation time, I don\u0027t see why enabling replication should check capacity on the remote site, regardless of whether it is a large CG or small CG.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":12112,"name":"Tom Swanson","email":"tom_swanson@dell.com","username":"tomswanson"},"change_message_id":"6feae1647b459d63bea951d5945002b86b12106c","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Change capability reporting and scheduler to support different"},{"line_number":60,"context_line":"level of grouping."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_1556032a","line":57,"in_reply_to":"1a26ad4f_6fb9bf78","updated":"2015-11-02 21:01:36.000000000","message":"I\u0027m confused.  Enable/disable is only resume/pause, right?  Replications are created at the time off volume creation, right?","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"6b6310ba961001446114bb0223a9be8e094318df","unresolved":false,"context_lines":[{"line_number":54,"context_line":"the CG. A driver method, that handles replication_enable for a CG will be"},{"line_number":55,"context_line":"introduced.  The manager will need to handle volume status update for all"},{"line_number":56,"context_line":"volumes. Create volume will need to handle both replication_enabled and"},{"line_number":57,"context_line":"consistencygroup_support capabilities."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Phase 2: Change capability reporting and scheduler to support different"},{"line_number":60,"context_line":"level of grouping."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_46413ba8","line":57,"in_reply_to":"1a26ad4f_8db3fee8","updated":"2015-11-02 17:53:05.000000000","message":"Would it be possible to push this into the creation spec?  So for example if you want replicated CG\u0027s that becomes a specifier in the type when you create the volume.  This pushes the burden on to the driver and also allows all the various permutations and combinations of things up to admin when he/she creates types and the details are up to the driver to figure out what it can support.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"6b6310ba961001446114bb0223a9be8e094318df","unresolved":false,"context_lines":[{"line_number":66,"context_line":"pool: Volumes in the group are from the same pool."},{"line_number":67,"context_line":"volgrp: Volumes in the group are in the same specific group setting defined"},{"line_number":68,"context_line":"by the storage backend."},{"line_number":69,"context_line":"None: CG is not supported."},{"line_number":70,"context_line":""},{"line_number":71,"context_line":"If a driver reports consistencygroup_support \u003d True, it will be treated"},{"line_number":72,"context_line":"like consistencygroup_support \u003d backend for backward compatibility."}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_e61887d5","line":69,"updated":"2015-11-02 17:53:05.000000000","message":"Yeah, this makes a lot of sense and I *think* solves the concerns raised by a number of vendors that can\u0027t implement the way we have it today.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"6b6310ba961001446114bb0223a9be8e094318df","unresolved":false,"context_lines":[{"line_number":71,"context_line":"If a driver reports consistencygroup_support \u003d True, it will be treated"},{"line_number":72,"context_line":"like consistencygroup_support \u003d backend for backward compatibility."},{"line_number":73,"context_line":"If a driver reports consistencygroup_support \u003d False, it will be treated"},{"line_number":74,"context_line":"like consistencygroup_support \u003d None for backward compatibility."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"Change the scheduler to add a ConsistencyGroupFilter to make sure the correct"},{"line_number":77,"context_line":"backend is picked for creating a consistency group. The extra specs for the"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_86ce732b","line":74,"updated":"2015-11-02 17:53:05.000000000","message":"I\u0027d prefer we didn\u0027t do a bunch of backward compatibility gymnastics if we could.  This is all internal code for the most part and until rolling upgrades is a reality probably no real benefit that I can see.  Yes, we\u0027ll have to go and update any current reporting, but that seems better than having multiple ways to do the same thing.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8bf5819281d00a60ef4da40b843c04e03ad29197","unresolved":false,"context_lines":[{"line_number":71,"context_line":"If a driver reports consistencygroup_support \u003d True, it will be treated"},{"line_number":72,"context_line":"like consistencygroup_support \u003d backend for backward compatibility."},{"line_number":73,"context_line":"If a driver reports consistencygroup_support \u003d False, it will be treated"},{"line_number":74,"context_line":"like consistencygroup_support \u003d None for backward compatibility."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"Change the scheduler to add a ConsistencyGroupFilter to make sure the correct"},{"line_number":77,"context_line":"backend is picked for creating a consistency group. The extra specs for the"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_882af58e","line":74,"in_reply_to":"1a26ad4f_86ce732b","updated":"2015-11-03 01:42:03.000000000","message":"Sure, we can certainly make all the changes if we can agree this is the right way to go.  No need to keep the old code there.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"db4fbcb537976b5d46630028dcd71e3ef169cc9b","unresolved":false,"context_lines":[{"line_number":78,"context_line":"volume type need to contain the following:"},{"line_number":79,"context_line":"consistencygroup_support: backend/pool/volgrp/None"},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"Phase 3: Add the following CG APIs that support replication."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":"* enable_replication"},{"line_number":84,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_dad19858","line":81,"updated":"2015-11-02 19:54:02.000000000","message":"Maybe instead of adding new cg specific api\u0027s we modify the existing replication ones to support targeting a volume, group, pool, whatever.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"4049140124f6eca7e57e720432bebd6a3f530fd2","unresolved":false,"context_lines":[{"line_number":78,"context_line":"volume type need to contain the following:"},{"line_number":79,"context_line":"consistencygroup_support: backend/pool/volgrp/None"},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"Phase 3: Add the following CG APIs that support replication."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":"* enable_replication"},{"line_number":84,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_e2e5c5a3","line":81,"in_reply_to":"1a26ad4f_a8049149","updated":"2015-11-03 04:10:10.000000000","message":"Yea I think that would potentially simplify the usage of this, debatable whether or not the internal volume api/manager code gets much better, but at least it wouldn\u0027t require a new set of api\u0027s for each type of thing we want to replicate.\n\nI think checking the type and/or expecting a hint as to what type it is as part of the api would be the way to go if we do consolidate them.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8bf5819281d00a60ef4da40b843c04e03ad29197","unresolved":false,"context_lines":[{"line_number":78,"context_line":"volume type need to contain the following:"},{"line_number":79,"context_line":"consistencygroup_support: backend/pool/volgrp/None"},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"Phase 3: Add the following CG APIs that support replication."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":"* enable_replication"},{"line_number":84,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_a8049149","line":81,"in_reply_to":"1a26ad4f_dad19858","updated":"2015-11-03 01:42:03.000000000","message":"This is an interesting idea.  What if enable_replication can take a volume, or a CG, or a pool as an input?  We check the type and decide what to do?","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"7f79150316a6a0ca0eed0b94a2b01ab19037fc00","unresolved":false,"context_lines":[{"line_number":78,"context_line":"volume type need to contain the following:"},{"line_number":79,"context_line":"consistencygroup_support: backend/pool/volgrp/None"},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"Phase 3: Add the following CG APIs that support replication."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":"* enable_replication"},{"line_number":84,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa80f949_f60bdf7b","line":81,"in_reply_to":"1a26ad4f_e2e5c5a3","updated":"2015-11-03 20:30:38.000000000","message":"I like the way this is going!\n\nThere are however backends where a set or group of cinder volumes is replicated and enable, disable, failover, and for that matter list-targets act on the group, where the group is just a logical thing and does not correspond to a CG or to a storage pool.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"6b6310ba961001446114bb0223a9be8e094318df","unresolved":false,"context_lines":[{"line_number":88,"context_line":"Alternatives"},{"line_number":89,"context_line":"------------"},{"line_number":90,"context_line":""},{"line_number":91,"context_line":"Without the proposed changes, we cannot support CG for replication."},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Data model impact"},{"line_number":94,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_06e343a5","line":91,"updated":"2015-11-02 17:53:05.000000000","message":"That may not be such a bad thing :)  I\u0027m not convinced that clever driver code couldn\u0027t accomplish this though even as it is today.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":12112,"name":"Tom Swanson","email":"tom_swanson@dell.com","username":"tomswanson"},"change_message_id":"6feae1647b459d63bea951d5945002b86b12106c","unresolved":false,"context_lines":[{"line_number":88,"context_line":"Alternatives"},{"line_number":89,"context_line":"------------"},{"line_number":90,"context_line":""},{"line_number":91,"context_line":"Without the proposed changes, we cannot support CG for replication."},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Data model impact"},{"line_number":94,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_b56f0ffb","line":91,"in_reply_to":"1a26ad4f_06e343a5","updated":"2015-11-02 21:01:36.000000000","message":"This all depends a lot on how you replicate things.  We have loose connections.  So if you want a replicated CG group you need to replicate each individual volume in the CG.  As replication happens at create if everything is on the same backend with replication enabled then there will be replication on the CG.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"db4fbcb537976b5d46630028dcd71e3ef169cc9b","unresolved":false,"context_lines":[{"line_number":88,"context_line":"Alternatives"},{"line_number":89,"context_line":"------------"},{"line_number":90,"context_line":""},{"line_number":91,"context_line":"Without the proposed changes, we cannot support CG for replication."},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Data model impact"},{"line_number":94,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_3f7c3eb4","line":91,"in_reply_to":"1a26ad4f_06e343a5","updated":"2015-11-02 19:54:02.000000000","message":"While i\u0027m not entirely convinced we need to support replication for CG\u0027s I do think leaving it entirely up to the driver might create a worse user experience. IMO at a minimum we should define the expected behavior so that at a high level an admin doesn\u0027t need to know how each driver was coded to handle the cases where CG volumes are replicated. As-is each driver can choose to do it however they want, but some things might have unexpected behaviors for the user. Even worse they will probably be different from vendor to vendor...\n\nEven if we don\u0027t bake in the behavior into the apis/manager/etc we should define what it should be doing so we can be at least a little bit consistent.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8bf5819281d00a60ef4da40b843c04e03ad29197","unresolved":false,"context_lines":[{"line_number":88,"context_line":"Alternatives"},{"line_number":89,"context_line":"------------"},{"line_number":90,"context_line":""},{"line_number":91,"context_line":"Without the proposed changes, we cannot support CG for replication."},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Data model impact"},{"line_number":94,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_f3a27afb","line":91,"in_reply_to":"1a26ad4f_b56f0ffb","updated":"2015-11-03 01:42:03.000000000","message":"Agree that it\u0027s good to define common behavior for drivers that can support this, even though driver can make it happen using \"types\" already.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"6b6310ba961001446114bb0223a9be8e094318df","unresolved":false,"context_lines":[{"line_number":105,"context_line":"* replication_enable_cg"},{"line_number":106,"context_line":"* replication_disable_cg"},{"line_number":107,"context_line":"* replication_failover_cg"},{"line_number":108,"context_line":"* list_replication_targets_cg"},{"line_number":109,"context_line":""},{"line_number":110,"context_line":"Phase 2:"},{"line_number":111,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_a6a12f48","line":108,"updated":"2015-11-02 17:53:05.000000000","message":"I think we can skip 1 - 3, and for 4 use the standard list_replication_targets but add a member variable that indicates if it\u0027s a cg or not.  I\u0027m not a fan of adding cg specific API methods if we can avoid it.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8bf5819281d00a60ef4da40b843c04e03ad29197","unresolved":false,"context_lines":[{"line_number":105,"context_line":"* replication_enable_cg"},{"line_number":106,"context_line":"* replication_disable_cg"},{"line_number":107,"context_line":"* replication_failover_cg"},{"line_number":108,"context_line":"* list_replication_targets_cg"},{"line_number":109,"context_line":""},{"line_number":110,"context_line":"Phase 2:"},{"line_number":111,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_3328f285","line":108,"in_reply_to":"1a26ad4f_a6a12f48","updated":"2015-11-03 01:42:03.000000000","message":"If we are stilling keeping \"phase 1\", I\u0027ll keep one driver interface but call two helper functions, one for volume and one for CG, as you suggested.  However, not sure if there\u0027s still any interest in pursuing phase 1.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"6b6310ba961001446114bb0223a9be8e094318df","unresolved":false,"context_lines":[{"line_number":113,"context_line":""},{"line_number":114,"context_line":"Phase 3:"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"New Consistency Group API in phase 3:"},{"line_number":117,"context_line":"* enable_replication"},{"line_number":118,"context_line":"* disable_replication"},{"line_number":119,"context_line":"* failover_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_a66feffa","line":116,"updated":"2015-11-02 17:53:05.000000000","message":"Same as above, if I understand this would be like:\ncg:enable_replication ....\n\nWe\u0027re creating an awful lot of API calls for two features that aren\u0027t even considered \u0027common\u0027.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"046b6d8702054455fbdb5504b9a90ac9d895a423","unresolved":false,"context_lines":[{"line_number":113,"context_line":""},{"line_number":114,"context_line":"Phase 3:"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"New Consistency Group API in phase 3:"},{"line_number":117,"context_line":"* enable_replication"},{"line_number":118,"context_line":"* disable_replication"},{"line_number":119,"context_line":"* failover_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_8eed1db9","line":116,"in_reply_to":"1a26ad4f_135a962a","updated":"2015-11-03 03:04:28.000000000","message":"Since all the replication API\u0027s are admin API\u0027s, the most straightforward way I can think of is this:\n\n\n\u0027/v2/\u003ctenant_id\u003e/volumes/\u003cvolume uuid\u003e/action\u0027\nmethod \u003d \u0027POST\u0027\nbody \u003d \n            {\u0027os-enable_replication\u0027: \n                {\u0027replication_group_type\u0027: \u0027cg\u0027 or \u0027pool\u0027,\n                 \u0027resource\u0027: \u003ccg_id\u003e or \u003cpool name\u003e}}\n\nIf \u0027replication_group_type\u0027 is not specified, just replicate volume; If \u0027replication_group_type\u0027  is \u0027cg\u0027, replication the CG (cg_id); If \u0027replication_group_type\u0027 is a \u0027pool\u0027, replication the pool (\u003cpool_name\u003e).","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"edb760df5eadfa893819b355d676ae3540382c08","unresolved":false,"context_lines":[{"line_number":113,"context_line":""},{"line_number":114,"context_line":"Phase 3:"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"New Consistency Group API in phase 3:"},{"line_number":117,"context_line":"* enable_replication"},{"line_number":118,"context_line":"* disable_replication"},{"line_number":119,"context_line":"* failover_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_657f7fab","line":116,"in_reply_to":"1a26ad4f_22c97d26","updated":"2015-11-03 05:04:57.000000000","message":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027 has problem too as we have to figure out whether the resource is a volume, pool, or CG.  A pool doesn\u0027t have a UUID.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"4049140124f6eca7e57e720432bebd6a3f530fd2","unresolved":false,"context_lines":[{"line_number":113,"context_line":""},{"line_number":114,"context_line":"Phase 3:"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"New Consistency Group API in phase 3:"},{"line_number":117,"context_line":"* enable_replication"},{"line_number":118,"context_line":"* disable_replication"},{"line_number":119,"context_line":"* failover_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_22c97d26","line":116,"in_reply_to":"1a26ad4f_8eed1db9","updated":"2015-11-03 04:10:10.000000000","message":"Seems like that would work, although the syntax does seem a bit awkward by having to hit an endpoint with /volumes/\u003cvolume uuid\u003e/action to enable replication on a cg or pool.\n\nIn a perfect world it would maybe be something more like \u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027 but that would mean a breaking change, or duplicating the api\u0027s :(\n\nMaybe its not too late to change the initial api\u0027s? Or at least change them as soon as we have microversions (and leave the old versions of them behind to just call the new impl specifying volume as the resource type)?","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8bf5819281d00a60ef4da40b843c04e03ad29197","unresolved":false,"context_lines":[{"line_number":113,"context_line":""},{"line_number":114,"context_line":"Phase 3:"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"New Consistency Group API in phase 3:"},{"line_number":117,"context_line":"* enable_replication"},{"line_number":118,"context_line":"* disable_replication"},{"line_number":119,"context_line":"* failover_replication"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a26ad4f_135a962a","line":116,"in_reply_to":"1a26ad4f_a66feffa","updated":"2015-11-03 01:42:03.000000000","message":"Exploring Patrick\u0027s suggestion of building this on top of your replication V2 API, may be we don\u0027t need to add new API\u0027s.  Let\u0027s see how that goes.","commit_id":"e9b47d565386e8c61401c9ab0640629ce28b1a08"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":25,"context_line":"Some backends can replicate a group of volumes, but not a single volume."},{"line_number":26,"context_line":"Some backends can replicate either a group of volumes or a single volume."},{"line_number":27,"context_line":"There are some differences in terminologies used by different backends,"},{"line_number":28,"context_line":"however, the goal of replicating a group of volumes is common, that is to"},{"line_number":29,"context_line":"achieve data consistency across all volumes replicated together."},{"line_number":30,"context_line":""},{"line_number":31,"context_line":"In Cinder, we also have consistencygroups and cgsnapshots. Cgsnapshots"}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_72b5962f","line":28,"updated":"2015-11-09 18:25:09.000000000","message":"\"to achieve data consistency across all volumes replicated together\" is I think either a vague notion or not necessarily the objective of all replication groups.\n\nIn some cases a replication group is a mechanism for applying a common policy, schedule, and destination (e.g., async, kick off every 5 minutes) but does not imply any tight relation among the volumes replicated other than that.  In particular, it doesn\u0027t imply that on the destination writes *across* the replicated volumes are necessarily in the proper order even when the replication status for each volume in the group is consistent (i.e., not in the middle of being replicated) on the destination.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"526d631b594ab5d51e2554b8deacb49e9f21ad35","unresolved":false,"context_lines":[{"line_number":25,"context_line":"Some backends can replicate a group of volumes, but not a single volume."},{"line_number":26,"context_line":"Some backends can replicate either a group of volumes or a single volume."},{"line_number":27,"context_line":"There are some differences in terminologies used by different backends,"},{"line_number":28,"context_line":"however, the goal of replicating a group of volumes is common, that is to"},{"line_number":29,"context_line":"achieve data consistency across all volumes replicated together."},{"line_number":30,"context_line":""},{"line_number":31,"context_line":"In Cinder, we also have consistencygroups and cgsnapshots. Cgsnapshots"}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_924595b4","line":28,"in_reply_to":"da85f559_72b5962f","updated":"2015-11-10 03:21:39.000000000","message":"So it is a policy based grouping.  Not sure what is the best way to handle this one.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":45,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":46,"context_line":"important to replicate all the volumes used by the application"},{"line_number":47,"context_line":"together to ensure data consistency."},{"line_number":48,"context_line":""},{"line_number":49,"context_line":"Proposed change"}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_d241e20c","line":46,"updated":"2015-11-09 18:25:09.000000000","message":"This is certainly a legitimate use case, and one that requires strong consistency (e.g. proper write ordering) *across* all the volumes in a replication group.\n\nI think there are also looser use cases, wherein one wants a set of volumes replicated with a common policy, schedule, destination and the like but it is strong consistency across the replicated volumes at any given point in time is not guaranteed.  Here the volume with DB and volume with DB logs might both be replicated at time t (and still not \u0027inconsistent\u0027 in the sense that they are in the process of being copied over, but not done) but the two might not be perfectly in sync.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"2b68dfec888a148ab358098eab904b8c328545a2","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa80f949_c5a1b547","line":83,"updated":"2015-11-06 18:02:09.000000000","message":"While i\u0027m not really opposed to this... the proposed workflow here seems like kind of a weird thing we are putting on a user of this.\n\nAs-is replication is an admin action and a normal user doesn\u0027t know much or anything about it, they just create a volume with a replicated volume type and thats it, right?\n\nNow though a user needs to know that some particular replicated type of volume needs to be in a CG first, and volumes created from that type are only replicated if they are added to the CG? Or maybe not...\n\nSeems like it might be confusing because if the volume type gets the volume scheduled into a pool that would be replicated, and there is a CG with replication enabled for that pool, you could have volumes that are *not* part of the CG but *are* being replicated because they happened to be the in same pool?\n\nPart of this which I think is going to complicate things is having the ability to add/remove volumes from a cg. In the case where you want to stop replicating a volume that is part of a pool based replication setup you would have to not only remove it from the cg but also migrate it to another pool/backend.\n\nUnfortunately I don\u0027t have a great suggestion for how we handle this... but IMO we do need to consider what workflow we would want a user to have and adjust the code to suite. I get the feeling right now we are trying to kind of massage pools and replication into the existing cg workflow which is complicating things. The idea that as a user you just create a volume of a certain type and things are replicated is, IMO, a good bar to set for this. If its not possible thats ok, but we should at least consider it... If we are going to push replication complexities onto a normal user maybe we should be looking more at what that new project (smaug?) does and how we can better integrate with that type of setup.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_7d47a74b","line":83,"in_reply_to":"da85f559_200374ba","updated":"2015-11-09 18:25:09.000000000","message":"I agree that pool based replication operations (enable, disable, failover, list-targets) would be admin-only, not tenant/user controlled.\n\nFWIW, that fits with replication v2, whereas CG-based replication will innovate relative to v2 by enabling tenant/user control of those operations.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"0e0391d20b4a9eedfab0d84d1eee7ca94f491600","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_a6345c4a","line":83,"in_reply_to":"da85f559_200374ba","updated":"2015-11-07 03:12:21.000000000","message":"I initially only plan to address the problem of replicating a CG because that was the only case I knew.  After reading Tom\u0027s comments on replicating a pool, I thought that can be solved here too because a few folks from NetApp have implemented CG for snapshot in Manila (based on the Cinder implementation with some changes) which allows shares in the same pool to be snapshotted together.  I know it is not replication, but it does group all the shares in the same pool together in a CG.\n\nNot that I want to boil the ocean, but if we don\u0027t think about how to solve both replicating CG and replicating a pool, we could come up with a design that needs to be changed soon.  This design can be for CG replication only, but some other ideas such as the Volume Group will change everything.  Also Tom thought about Replication Group.  If a CG Replication is not a Replication Group, what is it?  Maybe at the end, we decide these two cases should be handled differently, but I\u0027d like at least to explore the possibility of using common APIs.\n\nI don\u0027t know if pool replication can be done without any change to the common code.  This is because a driver cannot update volume statuses directly.  Somehow the manager needs to know that a pool of volumes are being replicated and the statuses of all volumes in the pool need to be updated.\n\nAll the concerns you raised are valid.  I don\u0027t have a good answer for them either.  That\u0027s why we are still discussing:).","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"3770697cfd438afba254c6b7996c20a790d57645","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_768a261d","line":83,"in_reply_to":"da85f559_31edbaef","updated":"2015-11-10 20:37:52.000000000","message":"OK, got it.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"79933ae5195e9bd366132dec1823a5e3e77cedf5","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_200374ba","line":83,"in_reply_to":"da85f559_372c20ab","updated":"2015-11-07 00:13:35.000000000","message":"Haha, I\u0027ll try to keep these to short stories ;)\n\nI get what you are saying about us having the problems with volumes moving between pools and adding/removing from groups regardless of grouping construct, totally agree with your assessment.\n\nMaybe what we need to clarify is whether or not this spec is proposing a way to allow for replicating CG\u0027s *only*, or if it is to support replication for backends who can only do pool or whole backend replication and want a way to expose that. If its only CG\u0027s than no problem, I\u0027ll stop rambling and +1 this.\n\nIf it isn\u0027t, and we are trying to allow for pools to be replicated, then I have some issue with the way we are exposing this. The main point I\u0027m hoping to make is that we shouldn\u0027t force a user of the cloud (not the operator/deployer/admin) to know whether or not they need to use a CG with some replicated volume types to actually get a volume replicated.\n\nSomeone correct me if I\u0027m wrong, but my assumption is that the whole point of this being brought up is that the pool based replicators want to have the ability to manage that feature of the array. Thats great, we should do that. It should however be something an admin deals with. Does that make sense?","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_12320a11","line":83,"in_reply_to":"da85f559_372c20ab","updated":"2015-11-09 18:25:09.000000000","message":"Yes, I agree that there is a tension between the requirements of pool-based replication, where volumes cannot be added and removed from the pool/pool replication group except by creating and destroying the volumes, and CG/logical replication groups, where they can be added and removed while continuing to exist.  Correspondingly, there\u0027s a tension between admin control of the pool/replication group and tenant/user control of CG/logical replication groups.\n\nSo, to be explicit, I think you are correct in saying that when I push the idea of a more general Volume Group (more general than CG group) that may or may not be valid, but that it does not by itself solve the question of how to handle pool replication (or more generally, replication based on \"physical\" groups - pools, sets of pools, entire backends).","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"526d631b594ab5d51e2554b8deacb49e9f21ad35","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_31edbaef","line":83,"in_reply_to":"da85f559_72a9167b","updated":"2015-11-10 03:21:39.000000000","message":"Tenant/user will create CG and create volumes in the CG by providing volume types.  With replication V2, a tenant/user will still need to create volume with a specific volume type to make it replicable.  enable/disable/failover replication are still admin functions.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_fdb717f3","line":83,"in_reply_to":"da85f559_a6345c4a","updated":"2015-11-09 18:25:09.000000000","message":"Yeah, on our DataOntap platforms we can replicate and snapshot the entire pool because we have a filesystem 1-1 with the pool and can snapshot the filesystem as a whole and replicate the snapshots.  Other filesystem based backends (e.g. ZFS) can likely do something similar.  We don\u0027t expose *those* capabilities in Cinder today because they are at the pool (or set of pools) level and would need to be under cloud admin control, whereas Cinder CGs as implemented are tenant/user controlled constructs.\n\nI recognize, but do not have any immediate answer for, the problem about how to update the volume statuses for all the volumes in a pool.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_72a9167b","line":83,"in_reply_to":"fa80f949_c5a1b547","updated":"2015-11-09 18:25:09.000000000","message":"Nice job showing exactly what has been giving me headaches when I think about this stuff!","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"3a88690cf99483c3e0abc269cbfdb5f6de9dab4c","unresolved":false,"context_lines":[{"line_number":80,"context_line":"      consistencygroup_replication: \u0027pool\u0027"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"- User creates a CG with the potential to support replication by"},{"line_number":83,"context_line":"  specifying volume_type_A."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"cinder consisgroup-create --name \u0027replication_group\u0027 volume_type_A"},{"line_number":86,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_372c20ab","line":83,"in_reply_to":"fa80f949_c5a1b547","updated":"2015-11-06 22:45:34.000000000","message":"Wow, I know Patrick can write a novel:).\n\nPlease note that when a user creates a volume that supports replication, he/she has to provide a volume type.  Creating a CG is the same and it needs user\u0027s input for a volume type (or more than 1 volume types).\n\nWhen you specify \u0027pool\u0027, that means you want all volumes in that pool to be replicated together.  If that is not what you want, then you can specify a different grouping \u0027volgrp\u0027.  It is up to the driver to decide what kind of special grouping makes sense.\n\nI think with pool based replication, this problem exists any way.  Correct me if I\u0027m wrong, Tom.  There may be no easy way to remove a volume from the pool.  This problem will be there even if you go with a new Volume Group construct.  With any grouping, you have to add individual volume to the grouping.  With the pool type of grouping, you just have to make sure that you add all volumes in that pool to the grouping.  There may be no way to remove a volume from that grouping (the pool) unless if you delete it.  I want to stress again that this problem exists whether you use CG as the grouping construct or create a new Volume Group construct.\n\nCG to me means replication first, and then snapshot, because I worked on a DR product that only uses CG replication for failover/failback without using CG snapshot.  So I never feel it complicates things when trying to add replication to the existing CG flow.  I\u0027m certainly open to design this differently if folks prefer a different solution.\n\nEven if we introduce Volume Group, we\u0027ll have to solve many of the same problems that we try to solve here.  We\u0027ll have to create a Volume Group first and then create volume and add to the group.  Add to / remove from the group will be needed as well.  The challenge with not being able to remove a volume from a pool will still be there.  To create a Volume Group, we need to add logic to the scheduler to pick a backend for the Volume Group that can support more than 1 volume types (or 1 group type).\n\nSure, I\u0027ll take a look of Smaug.  It has always been a debate on how much Cinder should do and how much a service at a higher layer should do.  I think that because the storage array level APIs to create a replication group of volumes, enable/disable replication, failover, etc are different at volume level and group level, we still need to enable it in Cinder for another project to use.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"79933ae5195e9bd366132dec1823a5e3e77cedf5","unresolved":false,"context_lines":[{"line_number":98,"context_line":""},{"line_number":99,"context_line":"- Enable replication, assuming the CLI is as follows:"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_a025245d","line":101,"updated":"2015-11-07 00:13:35.000000000","message":"The replication api\u0027s are, afaik, intended to be admin only. A user who created a CG may not even have access to these.\n\nThat might be something we need to address as far as replication usability in general..","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"3770697cfd438afba254c6b7996c20a790d57645","unresolved":false,"context_lines":[{"line_number":98,"context_line":""},{"line_number":99,"context_line":"- Enable replication, assuming the CLI is as follows:"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_36801ef7","line":101,"in_reply_to":"da85f559_95b57f32","updated":"2015-11-10 20:37:52.000000000","message":"Thanks for the clarification.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"0e0391d20b4a9eedfab0d84d1eee7ca94f491600","unresolved":false,"context_lines":[{"line_number":98,"context_line":""},{"line_number":99,"context_line":"- Enable replication, assuming the CLI is as follows:"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_a66c5c09","line":101,"in_reply_to":"da85f559_a025245d","updated":"2015-11-07 03:12:21.000000000","message":"Yes, we need to think about this.  I think a user should create CG and add volumes to the CG because only the user knows what application should be running on the volumes in the CG and needs to be protected together.  An admin doesn\u0027t know what the volumes are used for.  An admin can create a volume type for replication CG and tell the user what type to use.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":98,"context_line":""},{"line_number":99,"context_line":"- Enable replication, assuming the CLI is as follows:"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_fd38773f","line":101,"in_reply_to":"da85f559_a66c5c09","updated":"2015-11-09 18:25:09.000000000","message":"So even if we step aside a moment from all the issues about generalizing CGs to Volume Groups, and about whether pool replication can also be handled, this is an important point that needs to be called out and made more explicit.  This original spec introduces client/user control of replication whereas replication v2 is admin only.\n\nThat may or may not be a good thing at this stage of the Cinder replication saga, it just needs to be highlighted a bit more in this spec IMO.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"526d631b594ab5d51e2554b8deacb49e9f21ad35","unresolved":false,"context_lines":[{"line_number":98,"context_line":""},{"line_number":99,"context_line":"- Enable replication, assuming the CLI is as follows:"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_95b57f32","line":101,"in_reply_to":"da85f559_fd38773f","updated":"2015-11-10 03:21:39.000000000","message":"The cg-replication-enable API will still require admin context.  I didn\u0027t intend to change replication API from admin only to tenant API.  I should have added a note for that.  I was only saying that adding volume to CG can be a tenant task because a tenant knows what volumes to add to a CG.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"79933ae5195e9bd366132dec1823a5e3e77cedf5","unresolved":false,"context_lines":[{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"Alternatives"},{"line_number":106,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_801d8011","line":103,"updated":"2015-11-07 00:13:35.000000000","message":"For pool based replication it would replicate all volumes in the pool (including the ones in the replication group) which may contain volumes for multiple tenants... right? This is another place where I\u0027m concerned about us mixing user facing with admin/deployer facing things.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"Alternatives"},{"line_number":106,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_9d55ab51","line":103,"in_reply_to":"da85f559_1a2b3c84","updated":"2015-11-09 18:25:09.000000000","message":"I don\u0027t think we are planning to dedicate pools to a single project or to a nested project set currently.  Something to think on though ...","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"Alternatives"},{"line_number":106,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_9d27eb11","line":103,"in_reply_to":"da85f559_801d8011","updated":"2015-11-09 18:25:09.000000000","message":"Agreed.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"0e0391d20b4a9eedfab0d84d1eee7ca94f491600","unresolved":false,"context_lines":[{"line_number":100,"context_line":""},{"line_number":101,"context_line":"cinder cg-replication-enable \u003cCG uuid\u003e"},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"This will enable replication for all volumes in the replication group."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"Alternatives"},{"line_number":106,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_1a2b3c84","line":103,"in_reply_to":"da85f559_801d8011","updated":"2015-11-07 03:12:21.000000000","message":"That is a valid point, but seems beyond our control.  In Cinder, I don\u0027t see how we can support a pool based replication that spans multiple tenants, unless if the tenants are all under the same parent tenant in a nested project?  Admin has to set it up and make sure it works.  We can probably say we support it in a nested project structure (one parent with multiple child tenants) or a single tenant.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"e7c252e0e2e114f447afde8a36ec212ad2dc3aae","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_b7547554","line":157,"updated":"2015-11-09 03:56:52.000000000","message":"Some more thoughts about this option:\n\n\u0027/v2/\u003ctenant_id\u003e/replication_resources/\u003cresource uuid\u003e/action\u0027\n\n1) Have a common replication API for volume, CG, or pool.\n\n2) Add a new table for pool in the db.\n\n3) Add a new column pool_id in the volume table.\n\n4) Add new APIs to import/export a pool.\n\n5) When a volume is created, if it is part of a pool, driver can return {\u0027pool_name\u0027: \u0027xxx\u0027} and manager needs to figure out which pool it is in the db and assign pool_id to the volume in the volumes table.  (maybe there\u0027s a better way to associate a volume to a pool_id)\n\nPros: One set of APIs for volume, CG, and pools.\n\nCons: The existing Replication API will be changed; Need to add new API to manage pools.\n\nI don\u0027t know which backend has support for a different grouping other than CG or pool.  If it is replicating a group of volumes without ensuring data consistency, I don\u0027t see why it cannot be achieved by replicating each volume individually.  If it is replicating a group of volumes to ensure data consistency and a volume can be add/remove to/from a group, then we can use CG for it.  If it is replicating a group of volumes to ensure data consistency and a volume can not be add/remove to/from a group, then it seems more like a pool.  I\u0027d like to know more about this different grouping.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a9161a8da0df7ba2a494c23cb4b98db217cf94b0","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_7076aea5","line":157,"in_reply_to":"da85f559_15ebcfb1","updated":"2015-11-10 19:45:56.000000000","message":"Thinking more about this option.  The replication resource can be a volume, CG, pool, or this other replication grouping:\n\n/v2/\u003ctenant_id\u003e/replication_resources/\u003cresource uuid\u003e/action\n\nWhat if we have a register replication resource action to handle pools and other replication grouping?  Admin calls this API and provides a resource identifier (a pool name or a name of this other grouping) and UUID\u0027s of all volumes in this pool or grouping.  We\u0027ll need a replication_resources table in the db.  After registering, an entry will be added to the table and each volume will have a foreign key referencing the id of this replication_resource.\n\nFor Replication CG, tenant will create CG and add volumes to the CG, using the volume types provided by the admin.  The replication CG will be associated with a replication resource db entry as well.\n\nA volume that is replicable will also be associated with a replication resource db entry.\n\nOnly admin can run the replication actions such as enable-replication/disable-replication/failover-replication.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"3785222cf86b62a04b143dd40de70cabf731a0d4","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_78238356","line":157,"in_reply_to":"da85f559_163f1a25","updated":"2015-11-12 16:36:47.000000000","message":"Can we do:\n\n   \u0027/v2/\u003ctenant_id\u003e/replication/action\u0027\n\nand in the body something like:\n\n   {\u0027os_enable_replication\u0027:\n       {\u0027replication_granularity\u0027: \u0027volume\u0027|\u0027CG\u0027|\u0027pool\u0027|\u0027backend\u0027,\n         \u0027resource\u0027: \u003cuuid\u003e|\u003chost_string\u003e\n       }\n   }\n\nPros: \n    * small change to API\n    * no need to \"register\" pool|backend to get a uuid for them\n    * admin and tenant just need to create and use volume types, then admin can use the (generalized) replication v2 methods\n\nCons:\n    * more logic in the replication controller\n    * replication v2 API would change somewhat","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"4f93a1c6afd3293ba1ef1281f0a5fa8a000e21bc","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_681d9e65","line":157,"in_reply_to":"da85f559_234d7e72","updated":"2015-11-14 12:50:38.000000000","message":"Maybe we can have pretty much the same url as what you originally proposed:\n\n\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource-id\u003e/action\u0027\n\nbut resource-id can either be a uuid or a host.\n\nThe controller will then look in the body for \u0027replication_granularity\u0027 and pass its value (say, \u0027volume\u0027|\u0027CG\u0027|\u0027host\u0027) as another argument when invoking volume_api.enable_replication.\n\nvolume_api.enable_replication would itself call _enable_replication_on volume, _enable_replication_on_CG, _enable_replication_on_host as appropriate.\n\n_enable_replication_on_volume will just be the code from today\u0027s replication v2 enable_replication method.\n\nSo there would be no impact on the current effort in drivers to implement replication v2 itself.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a224932547db577e65edd032fa7715bb206523b7","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_7a723fe5","line":157,"in_reply_to":"da85f559_681d9e65","updated":"2015-11-16 16:32:24.000000000","message":"If \u003chost_string\u003e is acceptable to be part of a URL, then it simplifies a lot.  I like it.\n\nSure, we can have helper functions to handle volume, CG, or backend for the volume_api.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"1cce2dcea69a78c767e26d3d59f2c1f249968562","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_fbd03f21","line":157,"in_reply_to":"da85f559_7076aea5","updated":"2015-11-10 20:09:01.000000000","message":"That sounds pretty reasonable. To make sure I\u0027m understanding, we would automatically associate cg\u0027s and single volumes with a replication resource on creation? There would then be api\u0027s to add/remove/create replication resources where someone could set up a pool or whatever they need in addition to those two?","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"f9657a3aac74f37f41fd9165ed600223f5b38a52","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_c26e2296","line":157,"in_reply_to":"da85f559_78238356","updated":"2015-11-13 03:21:29.000000000","message":"Interesting idea.  What is \u003chost_string\u003e?  Is it some kind of identifier for backend to recognize a pool or do you mean a host in Cinder\u0027s term like host@backend#pool?\n\nI don\u0027t know if this is an acceptable format for an URL.  Usually we should have a resource id in the URL itself:\n\u0027/v2/\u003ctenant_id\u003e/replication/action\u0027\n\nIn this case, how do you update statuses of all the volumes in the pool?  Does the driver know all the volume UUID\u0027s and will return statuses for all UUID\u0027s back to the manager without querying the db?","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"ecaca7fc50eba244dc0d82e7fe54f36db601947d","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_234d7e72","line":157,"in_reply_to":"da85f559_8c9276f5","updated":"2015-11-14 06:09:02.000000000","message":"Thought about this again. The URL I suggested below won\u0027t work because we can\u0027t call it again on the same resource.\n\n/v2/\u003ctenant_id\u003e/replications/enable_replication","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_f87165a9","line":157,"in_reply_to":"da85f559_b7547554","updated":"2015-11-09 18:25:09.000000000","message":"\"I don\u0027t know which backend has support for a different grouping other than CG or pool.\"\n\nI need to talk to our E-Series folks to understand that backend better, but there CGs and Replication Groups are distinct from storage pools (they both slice and span storage pools).  Moreover, CGs and Replication Groups are themselves distinct from one another and consume different resources (the first, snapshots and snapshot groups, the second async mirror groups and mirror pairs).","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"d50880c4b7a067cb45dbc3a34710eaf8b95dc2e2","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_cde1659c","line":157,"in_reply_to":"da85f559_c26e2296","updated":"2015-11-13 11:17:09.000000000","message":"\"What is \u003chost string\u003e? ... do you mean a host in Cinder\u0027s term like host@backend#pool?\"\n\nYes, that\u0027s what I meant.\n\n\"I don\u0027t know if this is an acceptable format for an URL. Usually we should have a resource id in the URL itself: \u0027/v2/\u003ctenant_id\u003e/replication/action\"\n\nI wasn\u0027t sure what the acceptable formats for a URL were either and after looking around a bit consulted with ameade on this subject, since he knows the domain well (and thingee asked him to be API liason/checker for Cinder).  The above is what we came up with when considering this problem, so I think it or something equivalent will be acceptable.  The critical point, as you say, is whether we have to have the resource id in the URL itself.  Being able to put a uuid or host string in the bodyinstead frees us from having to create a uuid for pools and backends just to get this job done.\n\nOn the question of how to update statuses, I\u0027m thinking that the manager could pass in a list of the volume uuids the way you are now doing for CGs (https://review.openstack.org/#/c/229580/) and the driver can return a model update with\nstatuses for all these.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"3c51de27fda04445a11272aab494e205a978f486","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_8c9276f5","line":157,"in_reply_to":"da85f559_cde1659c","updated":"2015-11-13 16:11:49.000000000","message":"Sure, so the URL format is probably ok.  Still looks weird as I haven\u0027t seen an example like this.  With this, at least we don\u0027t need to add new APIs for registering a pool so that\u0027s good.\n\nI wonder if it makes sense to create a new replication resource with this enable API.  After the resource is created, we can do disable/failover on this resource.  Something like this:\n\n/v2/\u003ctenant_id\u003e/replications/enable_replication\nmethod: post\nbody:\n   {\u0027replication\u0027:\n       {\u0027replication_granularity\u0027: \u0027volume\u0027|\u0027CG\u0027|\u0027pool\u0027|\u0027backend\u0027,\n         \u0027resource\u0027: \u003cuuid\u003e|\u003chost_string\u003e\n       }\n   }\n\n/v2/\u003ctenant_id\u003e/replications/\u003cresource uuid\u003e/action\nmethod: post\nbody:\n   {\u0027os-disable_replication\u0027: {}\n   }\n\nThis does need to create a new replications table though.\n\nIn the CG case, the manager passes a list of all volumes in that CG.  So in the case that a host string \"host@backend#pool\" is passed, the manager can get a list of all the volumes from that host.  That probably works.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"3770697cfd438afba254c6b7996c20a790d57645","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_163f1a25","line":157,"in_reply_to":"da85f559_d618b277","updated":"2015-11-10 20:37:52.000000000","message":"This sounds promising.  I am thinking about whether we can somehow automate invocation of this API as volumes are created and added  with VolumeType specifying Replication \u003d True and replication_type \u003d \u0027pool\u0027 or \u0027backend\u0027.  We certainly don\u0027t want to require a manual administrator command as each new volume is created.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"526d631b594ab5d51e2554b8deacb49e9f21ad35","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_15ebcfb1","line":157,"in_reply_to":"da85f559_f87165a9","updated":"2015-11-10 03:21:39.000000000","message":"There are two tables in db today, one is consistencygroups and the other is cgsnapshots.  cgsnapshots hold the snapshot groups.  consistencygroups hold groups of volumes that can be used for snapshot group as well as replication group.  The cgsnapshots table has a relationship with the consistencygroups table.  We can add another table cgreplications if needed and that will hold mirror groups.  This is for Replication Group that is consistent.  If the Replication Group is not consistent, then that is different.\n\nIt will be great if you can provide a list of all the groupings with names that you are trying to support. Based on my understanding of your comments, there are two types groupings for replication that you are trying to support:\n\n1. Replicating all volumes in a pool.\n\n2. Replicating a group of volumes based on policy but they are not consistent.\n\n3. CG for replication (Sounds like you don\u0027t support this one today.  Is there plan to support it in the future)?\n\nDid I miss anything?\n\nSo far we have 3 kinds of groupings.  I think we need to understand all the use cases and then try to abstract them.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"70324ba65fb24854ca63314c20bb2125b4258e67","unresolved":false,"context_lines":[{"line_number":154,"context_line":"3. Keep the capability reporting change described in the proposed change"},{"line_number":155,"context_line":"section, and change the existing APIs."},{"line_number":156,"context_line":""},{"line_number":157,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replication/\u003cresource uuid\u003e/action\u0027"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Replication resource could be a volume or a CG."},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_d618b277","line":157,"in_reply_to":"da85f559_fbd03f21","updated":"2015-11-10 20:25:55.000000000","message":"Great:).  Yes for CG and single volume.\n\nFor pool and other types of grouping, the admin will set up pool and grouping on the array and set up volume types.  When creating a volume given correct volume type, the volume will be placed in the pool or grouping.  So this part is taken care of by volume type and extra specs.\n\nThen we will have an API for admin to add (or create) a replication_resource (which is a pool or grouping) with all the volumes (that are already created and placed in the pool or grouping in the array).  When admin calls this API, Cinder will know all these volumes are part of this replication_resource (pool or grouping) and make the change in the db.  We are not providing APIs to add a volume to the pool or grouping for the reasons we discussed previously.\n\nWe will need an API to delete this replication_resource, which will just remove this entry and the relationship with the volumes.  Deleting volume is still a separate operation as before.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"2b68dfec888a148ab358098eab904b8c328545a2","unresolved":false,"context_lines":[{"line_number":164,"context_line":"4. Introduce a new VolumeGroup construct with a group type (Proposed by"},{"line_number":165,"context_line":"Winston). A group type can be anything defined by the users, just like volume"},{"line_number":166,"context_line":"type but it is for a group. The VolumeGroup can be a replication group for a"},{"line_number":167,"context_line":"pool, CG, or anything the user wants to define."},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Pros: Very flexible."},{"line_number":170,"context_line":"Cons: Will introduce a lot of new code and the definition is too general."}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa80f949_c80b0c23","line":167,"updated":"2015-11-06 18:02:09.000000000","message":"I kind of like this idea, but agree it would add a ton of extra code and complexity (especially since we can\u0027t just yank out the existing cg api\u0027s).\n\nIt would help abstract a lot of these things though...","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"79933ae5195e9bd366132dec1823a5e3e77cedf5","unresolved":false,"context_lines":[{"line_number":164,"context_line":"4. Introduce a new VolumeGroup construct with a group type (Proposed by"},{"line_number":165,"context_line":"Winston). A group type can be anything defined by the users, just like volume"},{"line_number":166,"context_line":"type but it is for a group. The VolumeGroup can be a replication group for a"},{"line_number":167,"context_line":"pool, CG, or anything the user wants to define."},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Pros: Very flexible."},{"line_number":170,"context_line":"Cons: Will introduce a lot of new code and the definition is too general."}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_e0a00ca8","line":167,"in_reply_to":"da85f559_7ac7ed45","updated":"2015-11-07 00:13:35.000000000","message":"100% agree it doesn\u0027t solve all of our problems... just abstracts it maybe in a more clean way (or maybe would just complicate things even more XD)","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"0e0391d20b4a9eedfab0d84d1eee7ca94f491600","unresolved":false,"context_lines":[{"line_number":164,"context_line":"4. Introduce a new VolumeGroup construct with a group type (Proposed by"},{"line_number":165,"context_line":"Winston). A group type can be anything defined by the users, just like volume"},{"line_number":166,"context_line":"type but it is for a group. The VolumeGroup can be a replication group for a"},{"line_number":167,"context_line":"pool, CG, or anything the user wants to define."},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Pros: Very flexible."},{"line_number":170,"context_line":"Cons: Will introduce a lot of new code and the definition is too general."}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_85927256","line":167,"in_reply_to":"da85f559_7ac7ed45","updated":"2015-11-07 03:12:21.000000000","message":"By the way, this Volume Group with a group type reminded me of the Type Group that John proposed back in Juno when I first started working on CG.  I was supposed to make the CG dependent on the Type Group.  A Type Group is a group that supports a group of volume types.  A Type Group can have different capabilities depending on how admin sets it up.  It can be associated with a CG or some other groupings.  I just went back to the review history and refreshed my memory.  Type Group was in the spec up to patch set 6 when Duncan asked a question on how to handle backend supporting multiple Type Groups.  After that it got dropped due to the complexity.\n\nhttps://review.openstack.org/#/c/96665/6/specs/juno/consistency-groups.rst #line 69\n\nIn that review, Ben also raised concerns on replication CG.  After reading his comments again and thinking about Tom\u0027s concerns, I think one problem is exactly what Patrick raised earlier that you cannot remove a volume from a pool without deleting it or migrating it to another pool.  For a grouping not in a pool, we can remove a volume from it without deleting the volume.  Sounds like we either need to have a different set of APIs just for a pool or we can have a common set of APIs but pool can support create/delete volume but not add/remove volume (add/remove will be optional)?  Even if we use a Volume Group, we still have this challenge.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"39215e3c2b551377e1c60ae3152bc71b6cc462e5","unresolved":false,"context_lines":[{"line_number":164,"context_line":"4. Introduce a new VolumeGroup construct with a group type (Proposed by"},{"line_number":165,"context_line":"Winston). A group type can be anything defined by the users, just like volume"},{"line_number":166,"context_line":"type but it is for a group. The VolumeGroup can be a replication group for a"},{"line_number":167,"context_line":"pool, CG, or anything the user wants to define."},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Pros: Very flexible."},{"line_number":170,"context_line":"Cons: Will introduce a lot of new code and the definition is too general."}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_bdac2f36","line":167,"in_reply_to":"da85f559_85927256","updated":"2015-11-09 18:25:09.000000000","message":"Again, agree about this tension between physical groupings (pools) and logical groupings (CGs, Replication Groups) and appreciate the historical reference.  Looks like you folks have been working these issues for some time!","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"526d631b594ab5d51e2554b8deacb49e9f21ad35","unresolved":false,"context_lines":[{"line_number":164,"context_line":"4. Introduce a new VolumeGroup construct with a group type (Proposed by"},{"line_number":165,"context_line":"Winston). A group type can be anything defined by the users, just like volume"},{"line_number":166,"context_line":"type but it is for a group. The VolumeGroup can be a replication group for a"},{"line_number":167,"context_line":"pool, CG, or anything the user wants to define."},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Pros: Very flexible."},{"line_number":170,"context_line":"Cons: Will introduce a lot of new code and the definition is too general."}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_91a58605","line":167,"in_reply_to":"da85f559_bdac2f36","updated":"2015-11-10 03:21:39.000000000","message":"Yes, it did have a long history!  I thought CG for snapshot was phase 1, and CG for replication will come some day when we are ready.  I didn\u0027t know back then you guys have some different concepts about replication:).","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"3a88690cf99483c3e0abc269cbfdb5f6de9dab4c","unresolved":false,"context_lines":[{"line_number":164,"context_line":"4. Introduce a new VolumeGroup construct with a group type (Proposed by"},{"line_number":165,"context_line":"Winston). A group type can be anything defined by the users, just like volume"},{"line_number":166,"context_line":"type but it is for a group. The VolumeGroup can be a replication group for a"},{"line_number":167,"context_line":"pool, CG, or anything the user wants to define."},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Pros: Very flexible."},{"line_number":170,"context_line":"Cons: Will introduce a lot of new code and the definition is too general."}],"source_content_type":"text/x-rst","patch_set":5,"id":"da85f559_7ac7ed45","line":167,"in_reply_to":"fa80f949_c80b0c23","updated":"2015-11-06 22:45:34.000000000","message":"I can look at how to get this to work, but I do think we will still need to solve many of the same problems that we run into if we use consistencygroups regardless of whether we need to yank out the existing CG API\u0027s or not.","commit_id":"eee81564859a37c79e48f7f6527db3476c377db0"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"719e4c7260e08370b853c60a41f42e75543d51ed","unresolved":false,"context_lines":[{"line_number":43,"context_line":"Use case 2:"},{"line_number":44,"context_line":"There is also another use case where the user wants to replicate"},{"line_number":45,"context_line":"a group of volumes together based on a common policy but the data"},{"line_number":46,"context_line":"on the volumes is not guaranteed to be consistent."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"Proposed change"},{"line_number":49,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":6,"id":"ba8a016a_ea7e72ad","line":46,"updated":"2015-11-17 10:04:41.000000000","message":"It is consistent on *each* volume but write-order consistency is not guaranteed across the set of volumes in the group.  We might want to make this clear.","commit_id":"7499c7f67d54bbb75f07c7c138cb08fec2347b41"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"e7c89afc2e708a8674c17c955c4a00f14a18bf42","unresolved":false,"context_lines":[{"line_number":43,"context_line":"Use case 2:"},{"line_number":44,"context_line":"There is also another use case where the user wants to replicate"},{"line_number":45,"context_line":"a group of volumes together based on a common policy but the data"},{"line_number":46,"context_line":"on the volumes is not guaranteed to be consistent."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"Proposed change"},{"line_number":49,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":6,"id":"ba8a016a_8bbb02e5","line":46,"in_reply_to":"ba8a016a_ea7e72ad","updated":"2015-11-17 22:10:43.000000000","message":"Sure, will add this.","commit_id":"7499c7f67d54bbb75f07c7c138cb08fec2347b41"},{"author":{"_account_id":9003,"name":"Tom Barron","email":"tpb@dyncloud.net","username":"tbarron"},"change_message_id":"719e4c7260e08370b853c60a41f42e75543d51ed","unresolved":false,"context_lines":[{"line_number":67,"context_line":"* A host string such as host@backend or host@backend@pool"},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"Note: Need to verify whether it is acceptable for a host string"},{"line_number":70,"context_line":"to appear as part of the URL."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Method: POST"},{"line_number":73,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"ba8a016a_2aaeba2e","line":70,"updated":"2015-11-17 10:04:41.000000000","message":"For an example of host string as part of URL see:\n\nhttps://github.com/openstack/cinder/blob/master/cinder/api/contrib/hosts.py#L197","commit_id":"7499c7f67d54bbb75f07c7c138cb08fec2347b41"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"e7c89afc2e708a8674c17c955c4a00f14a18bf42","unresolved":false,"context_lines":[{"line_number":67,"context_line":"* A host string such as host@backend or host@backend@pool"},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"Note: Need to verify whether it is acceptable for a host string"},{"line_number":70,"context_line":"to appear as part of the URL."},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Method: POST"},{"line_number":73,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"ba8a016a_ebc40e60","line":70,"in_reply_to":"ba8a016a_2aaeba2e","updated":"2015-11-17 22:10:43.000000000","message":"Good.  It\u0027s already being used.  Thanks!","commit_id":"7499c7f67d54bbb75f07c7c138cb08fec2347b41"},{"author":{"_account_id":14274,"name":"XinXiaohui","email":"xiaohui.xin@intel.com","username":"XinXiaohui"},"change_message_id":"3749927560ad59597be6c4166e853830dd948e31","unresolved":false,"context_lines":[{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Request body for enable replication:"},{"line_number":63,"context_line":"   {\u0027os-enable_replication\u0027:"},{"line_number":64,"context_line":"       {\u0027replication_granularity\u0027: \u0027volume\u0027|\u0027CG\u0027,"},{"line_number":65,"context_line":"       }"},{"line_number":66,"context_line":"   }"},{"line_number":67,"context_line":"If \u0027replication_granularity\u0027 is not present, \u0027volume\u0027 will be used"}],"source_content_type":"text/x-rst","patch_set":8,"id":"dae33548_0599ad1f","line":64,"range":{"start_line":64,"start_character":9,"end_line":64,"end_character":32},"updated":"2016-02-16 08:19:39.000000000","message":"If we make the \u0027replication_granularity\u0027 as the same name with \"replication_type\" in line 73, will it be more consistent in understanding?","commit_id":"fb364be673459f52d09a7c8c5e7c89bda57077a1"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"4ec747ca1e4257269e74ad7836042e02f6de8d2b","unresolved":false,"context_lines":[{"line_number":59,"context_line":""},{"line_number":60,"context_line":"Method: POST"},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Request body for enable replication:"},{"line_number":63,"context_line":"   {\u0027os-enable_replication\u0027:"},{"line_number":64,"context_line":"       {\u0027replication_granularity\u0027: \u0027volume\u0027|\u0027CG\u0027,"},{"line_number":65,"context_line":"       }"},{"line_number":66,"context_line":"   }"},{"line_number":67,"context_line":"If \u0027replication_granularity\u0027 is not present, \u0027volume\u0027 will be used"},{"line_number":68,"context_line":"as the default."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":71,"context_line":"replication:"}],"source_content_type":"text/x-rst","patch_set":8,"id":"7a5de9d1_543735b0","line":68,"range":{"start_line":62,"start_character":0,"end_line":68,"end_character":15},"updated":"2016-01-27 23:07:27.000000000","message":"Couldn\u0027t we just check to see what type it is with some lookups? Seems like the parameter could be removed easily with some automation in the api.","commit_id":"fb364be673459f52d09a7c8c5e7c89bda57077a1"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"4ec747ca1e4257269e74ad7836042e02f6de8d2b","unresolved":false,"context_lines":[{"line_number":70,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":71,"context_line":"replication:"},{"line_number":72,"context_line":"    \u0027replication_enabled\u0027: \u003cis\u003e True"},{"line_number":73,"context_line":"    \u0027replication_type\u0027: \u0027volume\u0027|\u0027CG\u0027"},{"line_number":74,"context_line":"If \u0027replication_type\u0027 is not present, \u0027volume\u0027 will be the default."},{"line_number":75,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":76,"context_line":"\u0027replication_type\u0027 as capabilities."}],"source_content_type":"text/x-rst","patch_set":8,"id":"7a5de9d1_94036db8","line":73,"updated":"2016-01-27 23:07:27.000000000","message":"Are we planning to allow for a volume to have both volume and CG replication separately? I would lean towards no, and make it only be one or the other, but either way we should probably make a note in here somewhere what the expected behavior will be.","commit_id":"fb364be673459f52d09a7c8c5e7c89bda57077a1"},{"author":{"_account_id":14274,"name":"XinXiaohui","email":"xiaohui.xin@intel.com","username":"XinXiaohui"},"change_message_id":"3749927560ad59597be6c4166e853830dd948e31","unresolved":false,"context_lines":[{"line_number":84,"context_line":""},{"line_number":85,"context_line":"0. Not to add support for replicating a group of volumes. Driver can"},{"line_number":86,"context_line":"add support by using volume types and extra specs. The only thing a"},{"line_number":87,"context_line":"driver cannot do it to update the statuses of all volumes in a group"},{"line_number":88,"context_line":"as the driver is not supposed to update db directly."},{"line_number":89,"context_line":""},{"line_number":90,"context_line":"1. Have a separate set of APIs for groups."}],"source_content_type":"text/x-rst","patch_set":8,"id":"dae33548_783dc0ba","line":87,"range":{"start_line":87,"start_character":20,"end_line":87,"end_character":22},"updated":"2016-02-16 08:19:39.000000000","message":"s/do it to/do it is to","commit_id":"fb364be673459f52d09a7c8c5e7c89bda57077a1"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"4ec747ca1e4257269e74ad7836042e02f6de8d2b","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Data model impact"},{"line_number":127,"context_line":"-----------------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"None"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"REST API impact"},{"line_number":132,"context_line":"---------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"7a5de9d1_54d3b538","line":129,"updated":"2016-01-27 23:07:27.000000000","message":"We probably need a new \u0027replication_status\u0027 kind of field for CG objects, right?","commit_id":"fb364be673459f52d09a7c8c5e7c89bda57077a1"},{"author":{"_account_id":14274,"name":"XinXiaohui","email":"xiaohui.xin@intel.com","username":"XinXiaohui"},"change_message_id":"3749927560ad59597be6c4166e853830dd948e31","unresolved":false,"context_lines":[{"line_number":226,"context_line":"Work Items"},{"line_number":227,"context_line":"----------"},{"line_number":228,"context_line":""},{"line_number":229,"context_line":"1. Change capability reporting."},{"line_number":230,"context_line":"2. Change replication APIs to support group level replication."},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"Dependencies"}],"source_content_type":"text/x-rst","patch_set":8,"id":"dae33548_18891427","line":229,"range":{"start_line":229,"start_character":21,"end_line":229,"end_character":30},"updated":"2016-02-16 08:19:39.000000000","message":"What\u0027s for the change of capability reporting here?","commit_id":"fb364be673459f52d09a7c8c5e7c89bda57077a1"}],"specs/newton/replication-group.rst":[{"author":{"_account_id":7198,"name":"Jay Bryant","email":"jungleboyj@electronicjungle.net","username":"jsbryant"},"change_message_id":"bbdfeed845e2e054f712081ddee272f70a6096f7","unresolved":false,"context_lines":[{"line_number":17,"context_line":"At the Mitaka mid-cycle meetup, Cinder team decided that after Cheesecake"},{"line_number":18,"context_line":"is implemented, Tiramisu will be the next step. Tiramisu is the code name"},{"line_number":19,"context_line":"for a new set of replication APIs with more granularity. It gives tenant"},{"line_number":20,"context_line":"more control on what should be replicated together, i.e., a volume or a"},{"line_number":21,"context_line":"group of volumes."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":10,"id":"da0c15f0_6c634cb2","line":20,"range":{"start_line":20,"start_character":13,"end_line":20,"end_character":15},"updated":"2016-03-25 19:04:02.000000000","message":"of","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"c37e22202c3a588cc48f058c64d57a345e8db03d","unresolved":false,"context_lines":[{"line_number":17,"context_line":"At the Mitaka mid-cycle meetup, Cinder team decided that after Cheesecake"},{"line_number":18,"context_line":"is implemented, Tiramisu will be the next step. Tiramisu is the code name"},{"line_number":19,"context_line":"for a new set of replication APIs with more granularity. It gives tenant"},{"line_number":20,"context_line":"more control on what should be replicated together, i.e., a volume or a"},{"line_number":21,"context_line":"group of volumes."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":10,"id":"ba0121b8_4b29c936","line":20,"range":{"start_line":20,"start_character":13,"end_line":20,"end_character":15},"in_reply_to":"da0c15f0_6c634cb2","updated":"2016-03-30 15:22:20.000000000","message":"Done","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":16708,"name":"Kendall Nelson","display_name":"Kendall (diablo_rojo)","email":"kennelson11@gmail.com","username":"kjnelson"},"change_message_id":"f148a7dc7721d0af324cd869c175bd4d5d26abb5","unresolved":false,"context_lines":[{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Replication v2.1, which was implement in Mitaka, was designed to solve"},{"line_number":27,"context_line":"a true DR scenario when the entire backend has to be failed over to"},{"line_number":28,"context_line":"the secondary site."},{"line_number":29,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_1973f581","line":26,"range":{"start_line":26,"start_character":28,"end_line":26,"end_character":37},"updated":"2016-03-22 20:38:58.000000000","message":"implement-\u003eimplemented","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"ffe87f6900d817d40cac59bdf8cacb46dd712535","unresolved":false,"context_lines":[{"line_number":23,"context_line":"Problem description"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Replication v2.1, which was implement in Mitaka, was designed to solve"},{"line_number":27,"context_line":"a true DR scenario when the entire backend has to be failed over to"},{"line_number":28,"context_line":"the secondary site."},{"line_number":29,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_c69c8ec2","line":26,"range":{"start_line":26,"start_character":28,"end_line":26,"end_character":37},"in_reply_to":"fa0719c6_1973f581","updated":"2016-03-23 15:08:44.000000000","message":"Done","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"98aacf86533599660214f208e2caaab306312a71","unresolved":false,"context_lines":[{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Replication v2.1, which was implement in Mitaka, was designed to solve"},{"line_number":27,"context_line":"a true DR scenario when the entire backend has to be failed over to"},{"line_number":28,"context_line":"the secondary site."},{"line_number":29,"context_line":""},{"line_number":30,"context_line":"However, customers also want to have more control over the granularity"},{"line_number":31,"context_line":"of replication. The ability to replicate a volume or a group of volumes"}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_da91f3f2","line":28,"range":{"start_line":28,"start_character":14,"end_line":28,"end_character":18},"updated":"2016-03-23 19:37:29.000000000","message":"s/site/device ?","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"c37e22202c3a588cc48f058c64d57a345e8db03d","unresolved":false,"context_lines":[{"line_number":25,"context_line":""},{"line_number":26,"context_line":"Replication v2.1, which was implement in Mitaka, was designed to solve"},{"line_number":27,"context_line":"a true DR scenario when the entire backend has to be failed over to"},{"line_number":28,"context_line":"the secondary site."},{"line_number":29,"context_line":""},{"line_number":30,"context_line":"However, customers also want to have more control over the granularity"},{"line_number":31,"context_line":"of replication. The ability to replicate a volume or a group of volumes"}],"source_content_type":"text/x-rst","patch_set":10,"id":"ba0121b8_4be0a9fa","line":28,"range":{"start_line":28,"start_character":14,"end_line":28,"end_character":18},"in_reply_to":"fa0719c6_da91f3f2","updated":"2016-03-30 15:22:20.000000000","message":"Done","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":7198,"name":"Jay Bryant","email":"jungleboyj@electronicjungle.net","username":"jsbryant"},"change_message_id":"bbdfeed845e2e054f712081ddee272f70a6096f7","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."}],"source_content_type":"text/x-rst","patch_set":10,"id":"da0c15f0_4c77683a","line":43,"range":{"start_line":43,"start_character":25,"end_line":43,"end_character":26},"updated":"2016-03-25 19:04:02.000000000","message":"has a database","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"c37e22202c3a588cc48f058c64d57a345e8db03d","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."}],"source_content_type":"text/x-rst","patch_set":10,"id":"ba0121b8_8bcda187","line":43,"range":{"start_line":43,"start_character":25,"end_line":43,"end_character":26},"in_reply_to":"da0c15f0_4c77683a","updated":"2016-03-30 15:22:20.000000000","message":"Done","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"98aacf86533599660214f208e2caaab306312a71","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"There is also a use case where the volumes are loosely grouped together"},{"line_number":49,"context_line":"and replicated together without ensuring data consistency."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Proposed change"},{"line_number":52,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_5d079d17","line":49,"range":{"start_line":43,"start_character":0,"end_line":49,"end_character":58},"updated":"2016-03-23 19:37:29.000000000","message":"This use case gives absolutely no reason why cheesecake is insufficient. There is no requirement here for any better granularity than per-backend.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"c37e22202c3a588cc48f058c64d57a345e8db03d","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"There is also a use case where the volumes are loosely grouped together"},{"line_number":49,"context_line":"and replicated together without ensuring data consistency."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Proposed change"},{"line_number":52,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"ba0121b8_0bbd11b7","line":49,"range":{"start_line":43,"start_character":0,"end_line":49,"end_character":58},"in_reply_to":"da0c15f0_6c7eace5","updated":"2016-03-30 15:22:20.000000000","message":"I rewrote the use case section to indicate that Cheesecake is admin facing and Tiramisu is tenant facing. Hopefully I\u0027ve addressed your concerns.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"95c4752503cb6c9d257b7380df0728028f6cc5d6","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"There is also a use case where the volumes are loosely grouped together"},{"line_number":49,"context_line":"and replicated together without ensuring data consistency."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Proposed change"},{"line_number":52,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_c3562a29","line":49,"range":{"start_line":43,"start_character":0,"end_line":49,"end_character":58},"in_reply_to":"fa0719c6_1d14059e","updated":"2016-03-23 20:34:06.000000000","message":"@Duncan: ISP with multiple tenants. Same backend primary array. One tenant has a database that requires to be in a consistency group - they have multiple database disks and a log disk - and they need a consistent DR replication of the database, but only on a per day basis. Another tenant has another database in a consistency group that needs consistent DR replication, but that has to be on a 2 hourly basis. Potentially the ISP may have multiple site and one  of the tenants needs the database to be replicated to multiple arrays (like Cheesecake can do for the whole array), but the other tenant only needs a single DR copy of the database.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"79a28685c54ed1ff8a9705f6dd1e956c9f145b44","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"There is also a use case where the volumes are loosely grouped together"},{"line_number":49,"context_line":"and replicated together without ensuring data consistency."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Proposed change"},{"line_number":52,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_49551bf8","line":49,"range":{"start_line":43,"start_character":0,"end_line":49,"end_character":58},"in_reply_to":"fa0719c6_1d14059e","updated":"2016-03-23 21:04:13.000000000","message":"I think the main takeaway I had on the discussion at the mid-cycle was that cheesecake was admin facing and tiramisu was user facing. In my mind this means that for a user of the cloud they would have control over failovers and that sort of thing. The main use-case I remember that was brought up was that the user (in this case the application admin) would potentially want to be able to failover volumes for test purposes without needing an openstack admin to do it, or affecting volumes from other tenants. This kind of ties into the cg\u0027s nicely since it would make sense that some app using a cg would want to failover the whole cg in a consistent way.\n\nTo Simon\u0027s point an app admin (openstack user) would have control, and maybe more importantly for compliance or SLA kind of stuff on an app by app basis. They could have much more knowledge about how volumes are replicated and whether or not they can failover the app (by actually testing it). This spec doesn\u0027t really have any additional granularity around the where, when, and how of replication as mentioned by Simon, but I do however think it becomes a natural extension of this to allow for user specified parameters like that (beyond the ability to pick from different volumes types where an admin set them up beforehand like we have today). Thats maybe something for the next dessert though once we have this infrastructure in place to handle groups and user facing replication api\u0027s.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":7198,"name":"Jay Bryant","email":"jungleboyj@electronicjungle.net","username":"jsbryant"},"change_message_id":"bbdfeed845e2e054f712081ddee272f70a6096f7","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"There is also a use case where the volumes are loosely grouped together"},{"line_number":49,"context_line":"and replicated together without ensuring data consistency."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Proposed change"},{"line_number":52,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"da0c15f0_6c7eace5","line":49,"range":{"start_line":43,"start_character":0,"end_line":49,"end_character":58},"in_reply_to":"fa0719c6_49551bf8","updated":"2016-03-25 19:04:02.000000000","message":"Patrick, I think you are correct that the intention of tiramisu is for the functionality to be more user facing.  Allow the app admin to group and manage their volumes and replication as they see fit.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"5b76cddcd2326675b57067627032caf7c32a1f8f","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"There is also a use case where the volumes are loosely grouped together"},{"line_number":49,"context_line":"and replicated together without ensuring data consistency."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Proposed change"},{"line_number":52,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_5df15d00","line":49,"range":{"start_line":43,"start_character":0,"end_line":49,"end_character":58},"in_reply_to":"fa0719c6_5d079d17","updated":"2016-03-23 19:42:34.000000000","message":"I think we are talking about consistency group replication plus the ability to replicate different volumes to different target arrays potentially. Additionally in an async environment you may need to have different volumes replicated with different async parameters. Some might only need to be replicated daily but some might need more frequent replication. cheesecake does not give this level of granualarity, hence the reason for Tiramisu.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"6c48c1e8c3915a1cab90572cf75036b67722a953","unresolved":false,"context_lines":[{"line_number":40,"context_line":"Use Cases"},{"line_number":41,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Assume an application has database stored on one volume and logs"},{"line_number":44,"context_line":"stored on another and other related files on other volumes. It is"},{"line_number":45,"context_line":"important to replicate all the volumes used by the application"},{"line_number":46,"context_line":"together to ensure data consistency."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"There is also a use case where the volumes are loosely grouped together"},{"line_number":49,"context_line":"and replicated together without ensuring data consistency."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Proposed change"},{"line_number":52,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_1d14059e","line":49,"range":{"start_line":43,"start_character":0,"end_line":49,"end_character":58},"in_reply_to":"fa0719c6_5df15d00","updated":"2016-03-23 19:52:26.000000000","message":"That is a statement of technology, not a use case. Please give an example application, and an example of what circumstances a failover would happen, and details of why different volumes need different replication parameters.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":4523,"name":"Eric Harney","email":"eharney@redhat.com","username":"eharney"},"change_message_id":"dce75e84c60bd892c9a0f0cf41b9eafc4234cf2e","unresolved":false,"context_lines":[{"line_number":88,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":89,"context_line":"* create_group"},{"line_number":90,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":91,"context_line":"* delete_group"},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Discussion on the following two choices:"},{"line_number":94,"context_line":"(1) Introduce a group type (that is different from the volume type) for the groups."}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_402ff0f5","line":91,"updated":"2016-03-22 21:12:55.000000000","message":"I don\u0027t see how this addresses one of the driving factors for groups, which is that some have to be driver-defined.\n\nWhat does a driver to do create one of these groups itself, rather than the admin creating it?","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"98aacf86533599660214f208e2caaab306312a71","unresolved":false,"context_lines":[{"line_number":88,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":89,"context_line":"* create_group"},{"line_number":90,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":91,"context_line":"* delete_group"},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Discussion on the following two choices:"},{"line_number":94,"context_line":"(1) Introduce a group type (that is different from the volume type) for the groups."}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_da2d73a9","line":91,"in_reply_to":"fa0719c6_402ff0f5","updated":"2016-03-23 19:37:29.000000000","message":"Eric: I\u0027ve never been clear on this driver group thing. Can you go into details, please?","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"ffe87f6900d817d40cac59bdf8cacb46dd712535","unresolved":false,"context_lines":[{"line_number":88,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":89,"context_line":"* create_group"},{"line_number":90,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":91,"context_line":"* delete_group"},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Discussion on the following two choices:"},{"line_number":94,"context_line":"(1) Introduce a group type (that is different from the volume type) for the groups."}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_a6d6ba4c","line":91,"in_reply_to":"fa0719c6_402ff0f5","updated":"2016-03-23 15:08:44.000000000","message":"There should be driver functions as well.  I added them.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"8b4c4ba278131aee813e05d15678b07bf0f44df7","unresolved":false,"context_lines":[{"line_number":94,"context_line":"(1) Introduce a group type (that is different from the volume type) for the groups."},{"line_number":95,"context_line":"(2) Just use volume types (a group can support multiple volume types)."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"Data model impact"},{"line_number":98,"context_line":"-----------------"},{"line_number":99,"context_line":""},{"line_number":100,"context_line":"A new table for the groups will be added."}],"source_content_type":"text/x-rst","patch_set":10,"id":"1af94dfe_0429138a","line":97,"range":{"start_line":97,"start_character":0,"end_line":97,"end_character":1},"updated":"2016-03-19 17:07:45.000000000","message":"Might be wrong but I suspect you need an \u0027Alternatives\u0027 section to stop the gate failing.","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"b4a24836c35450789fa2839532faa6e2b2a703a0","unresolved":false,"context_lines":[{"line_number":94,"context_line":"(1) Introduce a group type (that is different from the volume type) for the groups."},{"line_number":95,"context_line":"(2) Just use volume types (a group can support multiple volume types)."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"Data model impact"},{"line_number":98,"context_line":"-----------------"},{"line_number":99,"context_line":""},{"line_number":100,"context_line":"A new table for the groups will be added."}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_974edbbf","line":97,"range":{"start_line":97,"start_character":0,"end_line":97,"end_character":1},"in_reply_to":"1af94dfe_0429138a","updated":"2016-03-21 03:52:14.000000000","message":"I\u0027ll try that.  Thanks:).","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"ffe87f6900d817d40cac59bdf8cacb46dd712535","unresolved":false,"context_lines":[{"line_number":94,"context_line":"(1) Introduce a group type (that is different from the volume type) for the groups."},{"line_number":95,"context_line":"(2) Just use volume types (a group can support multiple volume types)."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"Data model impact"},{"line_number":98,"context_line":"-----------------"},{"line_number":99,"context_line":""},{"line_number":100,"context_line":"A new table for the groups will be added."}],"source_content_type":"text/x-rst","patch_set":10,"id":"fa0719c6_26ae2ae6","line":97,"range":{"start_line":97,"start_character":0,"end_line":97,"end_character":1},"in_reply_to":"fa0719c6_974edbbf","updated":"2016-03-23 15:08:44.000000000","message":"Done","commit_id":"3ecead6424c84d8a761fd1eb33ee6ee0326055df"},{"author":{"_account_id":16917,"name":"Alex O\u0027Rourke","email":"alex.orourke@hpe.com","username":"aorourke"},"change_message_id":"c7b209478a623b1d3d9c298d152a5a292002332d","unresolved":false,"context_lines":[{"line_number":58,"context_line":"Proposed change"},{"line_number":59,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"The following APIs will be introduced to support replication for a single"},{"line_number":62,"context_line":"volume or a group of volumes."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"* enable_replication"},{"line_number":65,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_3d2db3fe","line":62,"range":{"start_line":61,"start_character":0,"end_line":62,"end_character":29},"updated":"2016-03-30 16:54:44.000000000","message":"Will volume replication be triggered on volume create as it is now? Same with groups?","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"b4fed064fcd43934d3dcd71e0767d8385b9f5ee5","unresolved":false,"context_lines":[{"line_number":58,"context_line":"Proposed change"},{"line_number":59,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"The following APIs will be introduced to support replication for a single"},{"line_number":62,"context_line":"volume or a group of volumes."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"* enable_replication"},{"line_number":65,"context_line":"* disable_replication"}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_92c829aa","line":62,"range":{"start_line":61,"start_character":0,"end_line":62,"end_character":29},"in_reply_to":"ba0121b8_3d2db3fe","updated":"2016-03-30 17:32:57.000000000","message":"Yes. We just have additional commands enable_replication and disable_replication to give tenants more control on those operations.","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":16917,"name":"Alex O\u0027Rourke","email":"alex.orourke@hpe.com","username":"aorourke"},"change_message_id":"c7b209478a623b1d3d9c298d152a5a292002332d","unresolved":false,"context_lines":[{"line_number":61,"context_line":"The following APIs will be introduced to support replication for a single"},{"line_number":62,"context_line":"volume or a group of volumes."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"* enable_replication"},{"line_number":65,"context_line":"* disable_replication"},{"line_number":66,"context_line":"* failover_replication"},{"line_number":67,"context_line":"* list_replication_targets"},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_7ddbab4f","line":65,"range":{"start_line":64,"start_character":0,"end_line":65,"end_character":21},"updated":"2016-03-30 16:54:44.000000000","message":"These are not supported in cheesecake, does it makes sense to do this for individual/groups of volumes.","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"b4fed064fcd43934d3dcd71e0767d8385b9f5ee5","unresolved":false,"context_lines":[{"line_number":61,"context_line":"The following APIs will be introduced to support replication for a single"},{"line_number":62,"context_line":"volume or a group of volumes."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"* enable_replication"},{"line_number":65,"context_line":"* disable_replication"},{"line_number":66,"context_line":"* failover_replication"},{"line_number":67,"context_line":"* list_replication_targets"},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_2f38daee","line":65,"range":{"start_line":64,"start_character":0,"end_line":65,"end_character":21},"in_reply_to":"ba0121b8_7ddbab4f","updated":"2016-03-30 17:32:57.000000000","message":"For groups, the tenant wants to have more controls and wants to be able to run tests.  So it makes sense to have these functions.","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":16917,"name":"Alex O\u0027Rourke","email":"alex.orourke@hpe.com","username":"aorourke"},"change_message_id":"c7b209478a623b1d3d9c298d152a5a292002332d","unresolved":false,"context_lines":[{"line_number":86,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":87,"context_line":"replication:"},{"line_number":88,"context_line":"    \u0027replication_enabled\u0027: \u003cis\u003e True"},{"line_number":89,"context_line":"    \u0027replication_type\u0027: \u0027volume\u0027|\u0027group\u0027"},{"line_number":90,"context_line":"If \u0027replication_type\u0027 is not present, \u0027volume\u0027 will be the default."},{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_7dfc4b72","line":89,"range":{"start_line":89,"start_character":4,"end_line":89,"end_character":40},"updated":"2016-03-30 16:54:44.000000000","message":"Why do we need to specify based on the group or the volume? Why can\u0027t we replicate individual volumes and groups with the same volume type?","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"b4fed064fcd43934d3dcd71e0767d8385b9f5ee5","unresolved":false,"context_lines":[{"line_number":86,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":87,"context_line":"replication:"},{"line_number":88,"context_line":"    \u0027replication_enabled\u0027: \u003cis\u003e True"},{"line_number":89,"context_line":"    \u0027replication_type\u0027: \u0027volume\u0027|\u0027group\u0027"},{"line_number":90,"context_line":"If \u0027replication_type\u0027 is not present, \u0027volume\u0027 will be the default."},{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_d26aa1b2","line":89,"range":{"start_line":89,"start_character":4,"end_line":89,"end_character":40},"in_reply_to":"ba0121b8_7dfc4b72","updated":"2016-03-30 17:32:57.000000000","message":"If we want to support volume replication and group replication, we need to somehow differentiate them as the code path to handle them will be different. Alternatively we can just support group replication and a single volume can be a special case of a group with one volume.  Personally I think group replication is more important than single volume replication.","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":16917,"name":"Alex O\u0027Rourke","email":"alex.orourke@hpe.com","username":"aorourke"},"change_message_id":"c7b209478a623b1d3d9c298d152a5a292002332d","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":98,"context_line":"* delete_group"},{"line_number":99,"context_line":"Note: There will be driver interfaces for create_group, update_group,"},{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_7d0a8b7a","line":98,"range":{"start_line":94,"start_character":0,"end_line":98,"end_character":14},"updated":"2016-03-30 16:54:44.000000000","message":"These groups are 100% independent from CGs? So you cannot snapshot a replication group, only the currently existing CG?","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"b4fed064fcd43934d3dcd71e0767d8385b9f5ee5","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":98,"context_line":"* delete_group"},{"line_number":99,"context_line":"Note: There will be driver interfaces for create_group, update_group,"},{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":12,"id":"ba0121b8_2f51fa36","line":98,"range":{"start_line":94,"start_character":0,"end_line":98,"end_character":14},"in_reply_to":"ba0121b8_7d0a8b7a","updated":"2016-03-30 17:32:57.000000000","message":"If we decide to create a new group construct and not use CG, this will be independent from CG.  As indicated in Alternatives, we can use CG without creating a new group construct if there is no need to support a loosely replicated group that does not need data consistency across volumes in the group.","commit_id":"e46404c10e7e5c8c48c310fc780ebde200e1b2ba"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"3546514b1d70b15c1aca944dc6b070296c4faa3e","unresolved":false,"context_lines":[{"line_number":30,"context_line":"However, customers also want to have more control over the granularity"},{"line_number":31,"context_line":"of replication. The ability to replicate a volume or a group of volumes"},{"line_number":32,"context_line":"has been a desired feature. A tenant facing API is needed to meet this"},{"line_number":33,"context_line":"requirement."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Some backends can replicate a group of volumes, but not a single volume."},{"line_number":36,"context_line":"Some backends can replicate either a group of volumes or a single volume."}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_1eb87bc7","line":33,"range":{"start_line":33,"start_character":0,"end_line":33,"end_character":11},"updated":"2016-04-07 03:15:07.000000000","message":"Can we write up a specific use case for this one?  What exactly do we want to provide and why?\n\nFWIW, you can already do group based replication with Cheesecake, by adding replication info to a type and creating the volumes of that type you get replicated volumes.  We allow that granularity already, we just don\u0027t allow failing over a single volume (and I don\u0027t know for sure that I think we should).","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":20531,"name":"Jim Holl","email":"jimholl@netapp.com","username":"jimholl"},"change_message_id":"3c6175570d131a0dcfa8e83cbd0a0be2a668d132","unresolved":false,"context_lines":[{"line_number":30,"context_line":"However, customers also want to have more control over the granularity"},{"line_number":31,"context_line":"of replication. The ability to replicate a volume or a group of volumes"},{"line_number":32,"context_line":"has been a desired feature. A tenant facing API is needed to meet this"},{"line_number":33,"context_line":"requirement."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Some backends can replicate a group of volumes, but not a single volume."},{"line_number":36,"context_line":"Some backends can replicate either a group of volumes or a single volume."}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_5e080ada","line":33,"range":{"start_line":33,"start_character":0,"end_line":33,"end_character":11},"in_reply_to":"9a061dce_1eb87bc7","updated":"2016-04-07 15:50:40.000000000","message":"I agree we should be more explicit about the use case. I think tenant control means two things:\n1) the ability to perform a failover test\n2) the ability to decide when and whether to failover after an outage or wait for the primary to recover","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":12112,"name":"Tom Swanson","email":"tom_swanson@dell.com","username":"tomswanson"},"change_message_id":"fc5136034ebb4f9fce2acf5c089a33d0063bc9ca","unresolved":false,"context_lines":[{"line_number":30,"context_line":"However, customers also want to have more control over the granularity"},{"line_number":31,"context_line":"of replication. The ability to replicate a volume or a group of volumes"},{"line_number":32,"context_line":"has been a desired feature. A tenant facing API is needed to meet this"},{"line_number":33,"context_line":"requirement."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Some backends can replicate a group of volumes, but not a single volume."},{"line_number":36,"context_line":"Some backends can replicate either a group of volumes or a single volume."}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_43503732","line":33,"range":{"start_line":33,"start_character":0,"end_line":33,"end_character":11},"in_reply_to":"9a061dce_1eb87bc7","updated":"2016-04-07 15:33:20.000000000","message":"I like the use of volume types but I can see the case for wanting to failover that volume type.\n\nLike a repl_type to just enable replication and a critical_repl_type for vols that need replication and that the user might want higher availability for.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"004d131004a22099569db20301ce96dcb5690192","unresolved":false,"context_lines":[{"line_number":30,"context_line":"However, customers also want to have more control over the granularity"},{"line_number":31,"context_line":"of replication. The ability to replicate a volume or a group of volumes"},{"line_number":32,"context_line":"has been a desired feature. A tenant facing API is needed to meet this"},{"line_number":33,"context_line":"requirement."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Some backends can replicate a group of volumes, but not a single volume."},{"line_number":36,"context_line":"Some backends can replicate either a group of volumes or a single volume."}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_665ed777","line":33,"range":{"start_line":33,"start_character":0,"end_line":33,"end_character":11},"in_reply_to":"9a061dce_3fd1b423","updated":"2016-04-08 21:55:03.000000000","message":"I\u0027d like to understand the interaction between Cheesecake \u0026 Tiramisu.  If a twitchy tenant performs a failover early in an outage, and the admin later decides it\u0027s best to failover the entire backend, what happens?","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"2afafdf36e8c0b1c7b7a2ab1189232a2c62c6c44","unresolved":false,"context_lines":[{"line_number":30,"context_line":"However, customers also want to have more control over the granularity"},{"line_number":31,"context_line":"of replication. The ability to replicate a volume or a group of volumes"},{"line_number":32,"context_line":"has been a desired feature. A tenant facing API is needed to meet this"},{"line_number":33,"context_line":"requirement."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Some backends can replicate a group of volumes, but not a single volume."},{"line_number":36,"context_line":"Some backends can replicate either a group of volumes or a single volume."}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_3fd1b423","line":33,"range":{"start_line":33,"start_character":0,"end_line":33,"end_character":11},"in_reply_to":"9a061dce_5e080ada","updated":"2016-04-08 03:36:11.000000000","message":"Sure, I\u0027ll add a specific use case.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"3546514b1d70b15c1aca944dc6b070296c4faa3e","unresolved":false,"context_lines":[{"line_number":53,"context_line":"and require different replication and failover policies. Therefore, a tenant"},{"line_number":54,"context_line":"wants to be able to set up a group of volumes with its own replication policy"},{"line_number":55,"context_line":"and be able to fail over this group of volumes without affecting the entire"},{"line_number":56,"context_line":"backend."},{"line_number":57,"context_line":""},{"line_number":58,"context_line":"Proposed change"},{"line_number":59,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_3e01ff6a","line":56,"range":{"start_line":56,"start_character":0,"end_line":56,"end_character":7},"updated":"2016-04-07 03:15:07.000000000","message":"The crux of the issue for me however is what the use case is for a failover here that would be different than a DR scenario?  I\u0027m all for a logical grouping of volumes (and have proposed it quite a few times), and certainly having drivers do fancy replication things with those groups is fair game and certainly cool.\n\nThe failover use case however is what I\u0027m stuck on.  We\u0027ve learned (myself included) that replicating and failing over a single volume to a secondary while leaving a primary running with the remaining volumes gets rather complex and convoluted very quickly.   I\u0027m also unclear on what the valid use case is for this scenario.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":12112,"name":"Tom Swanson","email":"tom_swanson@dell.com","username":"tomswanson"},"change_message_id":"fc5136034ebb4f9fce2acf5c089a33d0063bc9ca","unresolved":false,"context_lines":[{"line_number":53,"context_line":"and require different replication and failover policies. Therefore, a tenant"},{"line_number":54,"context_line":"wants to be able to set up a group of volumes with its own replication policy"},{"line_number":55,"context_line":"and be able to fail over this group of volumes without affecting the entire"},{"line_number":56,"context_line":"backend."},{"line_number":57,"context_line":""},{"line_number":58,"context_line":"Proposed change"},{"line_number":59,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_8337ef4b","line":56,"range":{"start_line":56,"start_character":0,"end_line":56,"end_character":7},"in_reply_to":"9a061dce_3e01ff6a","updated":"2016-04-07 15:33:20.000000000","message":"Power went out in lab a. Due back in an hour. An hour is too long for server x but y and z are not absolutely critical. Their data needs protecting but they do not need to be up all the time.  Failover x\u0027s volumes and let y and z\u0027s come back when lab a comes back.\n\nFailback isn\u0027t clean or even supported in some drivers.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"2afafdf36e8c0b1c7b7a2ab1189232a2c62c6c44","unresolved":false,"context_lines":[{"line_number":53,"context_line":"and require different replication and failover policies. Therefore, a tenant"},{"line_number":54,"context_line":"wants to be able to set up a group of volumes with its own replication policy"},{"line_number":55,"context_line":"and be able to fail over this group of volumes without affecting the entire"},{"line_number":56,"context_line":"backend."},{"line_number":57,"context_line":""},{"line_number":58,"context_line":"Proposed change"},{"line_number":59,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_df27d860","line":56,"range":{"start_line":56,"start_character":0,"end_line":56,"end_character":7},"in_reply_to":"9a061dce_8337ef4b","updated":"2016-04-08 03:36:11.000000000","message":"Failing over a group of volumes is an important use case for me.  There are critical applications that a tenant wants to make sure he/she can test failover and failback easily and does not need to failover the entire backend.\n\nI am okay not to worry about failing over a single volume for now.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"004d131004a22099569db20301ce96dcb5690192","unresolved":false,"context_lines":[{"line_number":53,"context_line":"and require different replication and failover policies. Therefore, a tenant"},{"line_number":54,"context_line":"wants to be able to set up a group of volumes with its own replication policy"},{"line_number":55,"context_line":"and be able to fail over this group of volumes without affecting the entire"},{"line_number":56,"context_line":"backend."},{"line_number":57,"context_line":""},{"line_number":58,"context_line":"Proposed change"},{"line_number":59,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_c601a33b","line":56,"range":{"start_line":56,"start_character":0,"end_line":56,"end_character":7},"in_reply_to":"9a061dce_df27d860","updated":"2016-04-08 21:55:03.000000000","message":"I think that for replication, a group should be atomic; failing over a subset of a group is asking for trouble.  (In fairness, I didn\u0027t see that implied in the original text.)","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"004d131004a22099569db20301ce96dcb5690192","unresolved":false,"context_lines":[{"line_number":61,"context_line":"The following APIs will be introduced to support replication for a single"},{"line_number":62,"context_line":"volume or a group of volumes."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"* enable_replication"},{"line_number":65,"context_line":"* disable_replication"},{"line_number":66,"context_line":"* failover_replication"},{"line_number":67,"context_line":"* list_replication_targets"},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_261cdf76","line":65,"range":{"start_line":64,"start_character":0,"end_line":65,"end_character":21},"updated":"2016-04-08 21:55:03.000000000","message":"What do these APIs do?  Isn\u0027t it sufficient to know from a volume type or volume group type that replication is supported?  Or are you assuming the backend will somehow know where it should replicate to and create replicas automatically?  Is that true even if there are several AZs that may contain the replicas, and how many replicas are created?  I think this spec needs more detail about how the feature works from a user experience perspective and how the APIs create that experience.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"eac3aaf91c48deeb3cefc1e9acbeaf448bfd5c28","unresolved":false,"context_lines":[{"line_number":80,"context_line":"       {\u0027replication_granularity\u0027: \u0027volume\u0027|\u0027group\u0027,"},{"line_number":81,"context_line":"       }"},{"line_number":82,"context_line":"   }"},{"line_number":83,"context_line":"If \u0027replication_granularity\u0027 is not present, \u0027volume\u0027 will be used"},{"line_number":84,"context_line":"as the default."},{"line_number":85,"context_line":""},{"line_number":86,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":87,"context_line":"replication:"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_5d897d74","line":84,"range":{"start_line":83,"start_character":0,"end_line":84,"end_character":15},"updated":"2016-03-30 20:57:40.000000000","message":"Can\u0027t we just detect if the resource_id was a volume or group? With at most two db queries (one of which we always have to doanyway) we would have the info and don\u0027t have to make any assumptions. We could even drop the replication_granularity parameter altogether if folks are ok with an extra db query.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":80,"context_line":"       {\u0027replication_granularity\u0027: \u0027volume\u0027|\u0027group\u0027,"},{"line_number":81,"context_line":"       }"},{"line_number":82,"context_line":"   }"},{"line_number":83,"context_line":"If \u0027replication_granularity\u0027 is not present, \u0027volume\u0027 will be used"},{"line_number":84,"context_line":"as the default."},{"line_number":85,"context_line":""},{"line_number":86,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":87,"context_line":"replication:"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_b304e414","line":84,"range":{"start_line":83,"start_character":0,"end_line":84,"end_character":15},"in_reply_to":"ba0121b8_5d897d74","updated":"2016-04-01 15:45:10.000000000","message":"Sure.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"eac3aaf91c48deeb3cefc1e9acbeaf448bfd5c28","unresolved":false,"context_lines":[{"line_number":86,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":87,"context_line":"replication:"},{"line_number":88,"context_line":"    \u0027replication_enabled\u0027: \u003cis\u003e True"},{"line_number":89,"context_line":"    \u0027replication_type\u0027: \u0027volume\u0027|\u0027group\u0027"},{"line_number":90,"context_line":"If \u0027replication_type\u0027 is not present, \u0027volume\u0027 will be the default."},{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_7d2d61fd","line":89,"updated":"2016-03-30 20:57:40.000000000","message":"Some folks use replication_type for sync/async and stuff, we might have to use a different keyword.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":86,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":87,"context_line":"replication:"},{"line_number":88,"context_line":"    \u0027replication_enabled\u0027: \u003cis\u003e True"},{"line_number":89,"context_line":"    \u0027replication_type\u0027: \u0027volume\u0027|\u0027group\u0027"},{"line_number":90,"context_line":"If \u0027replication_type\u0027 is not present, \u0027volume\u0027 will be the default."},{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_d3287888","line":89,"in_reply_to":"ba0121b8_0100e264","updated":"2016-04-01 15:45:10.000000000","message":"Ok.  I\u0027ll change to replication_granularity.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"7b51db6700871da9b0da94cee0e5352b91d1c17b","unresolved":false,"context_lines":[{"line_number":86,"context_line":"Volume type needs to have the following extra_spec to support"},{"line_number":87,"context_line":"replication:"},{"line_number":88,"context_line":"    \u0027replication_enabled\u0027: \u003cis\u003e True"},{"line_number":89,"context_line":"    \u0027replication_type\u0027: \u0027volume\u0027|\u0027group\u0027"},{"line_number":90,"context_line":"If \u0027replication_type\u0027 is not present, \u0027volume\u0027 will be the default."},{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_0100e264","line":89,"in_reply_to":"ba0121b8_7d2d61fd","updated":"2016-04-01 14:15:03.000000000","message":"Good point Patrick. If we are using replication_granularity, then maybe we could call this \u0027replication_grain\u0027 instead :)","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"3546514b1d70b15c1aca944dc6b070296c4faa3e","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_fe31475b","line":94,"range":{"start_line":94,"start_character":0,"end_line":94,"end_character":66},"updated":"2016-04-07 03:15:07.000000000","message":"I\u0027m certainly in favor of this type of construct, maybe worth implementing as an independent thing first?","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"2afafdf36e8c0b1c7b7a2ab1189232a2c62c6c44","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_ffa9fc90","line":94,"range":{"start_line":94,"start_character":0,"end_line":94,"end_character":66},"in_reply_to":"9a061dce_fe31475b","updated":"2016-04-08 03:36:11.000000000","message":"Sure.  Replication group and groups with other features can be dependent on the basic group construct.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"004d131004a22099569db20301ce96dcb5690192","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_a1f579cc","line":94,"range":{"start_line":94,"start_character":0,"end_line":94,"end_character":66},"in_reply_to":"9a061dce_ffa9fc90","updated":"2016-04-08 21:55:03.000000000","message":"+1, I\u0027d like to see generic groups come before group replication.  So maybe do generic groups and Tiramisu for single volumes in Newton, and do Tiramisu for groups in Ocata.  Based on our experience with replication in Manila, I doubt all of the above will fit in one release.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"eac3aaf91c48deeb3cefc1e9acbeaf448bfd5c28","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":98,"context_line":"* delete_group"},{"line_number":99,"context_line":"Note: There will be driver interfaces for create_group, update_group,"},{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_1d7d45b1","line":98,"range":{"start_line":94,"start_character":0,"end_line":98,"end_character":14},"updated":"2016-03-30 20:57:40.000000000","message":"Would it make sense to have the api\u0027s be more like create_volume_group or create_vol_group?\n\nI think we would also need a like show/get and list API for groups, right?","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"2afafdf36e8c0b1c7b7a2ab1189232a2c62c6c44","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":98,"context_line":"* delete_group"},{"line_number":99,"context_line":"Note: There will be driver interfaces for create_group, update_group,"},{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_bf2dc4ff","line":98,"range":{"start_line":94,"start_character":0,"end_line":98,"end_character":14},"in_reply_to":"9a061dce_2358a355","updated":"2016-04-08 03:36:11.000000000","message":"When the group is created, the scheduler will choose a backend that can support the specified volume types (or group type if we decide to add that).  When we create a volume, we provide group uuid and volume type.  That should pick the same backend where the group is created on.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":10221,"name":"Edward Balduf","email":"ed.balduf@solidfire.com","username":"ebalduf"},"change_message_id":"9932e6b28c817bd572d62739c6a4deb14c26a734","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":98,"context_line":"* delete_group"},{"line_number":99,"context_line":"Note: There will be driver interfaces for create_group, update_group,"},{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_bde51239","line":98,"range":{"start_line":94,"start_character":0,"end_line":98,"end_character":14},"in_reply_to":"ba0121b8_1d7d45b1","updated":"2016-03-31 14:24:44.000000000","message":"I agree with Patrick, get/show and list API calls are important.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":20531,"name":"Jim Holl","email":"jimholl@netapp.com","username":"jimholl"},"change_message_id":"3c6175570d131a0dcfa8e83cbd0a0be2a668d132","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":98,"context_line":"* delete_group"},{"line_number":99,"context_line":"Note: There will be driver interfaces for create_group, update_group,"},{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_2358a355","line":98,"range":{"start_line":94,"start_character":0,"end_line":98,"end_character":14},"in_reply_to":"ba0121b8_73452c41","updated":"2016-04-07 15:50:40.000000000","message":"Will there be a way to specify a group when a volume is created? Adding a volume to a group after creation might fail or require migration. Perhaps groups need a mutability attribute.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Driver also needs to report \u0027replication_enabled\u0027 and"},{"line_number":92,"context_line":"\u0027replication_type\u0027 as capabilities."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"To support a group of volumes, a new VolumeGroup construct will be"},{"line_number":95,"context_line":"introduced. The following APIs are needed for the group:"},{"line_number":96,"context_line":"* create_group"},{"line_number":97,"context_line":"* update_group (add volume to the group or remove volume from the group)"},{"line_number":98,"context_line":"* delete_group"},{"line_number":99,"context_line":"Note: There will be driver interfaces for create_group, update_group,"},{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_73452c41","line":98,"range":{"start_line":94,"start_character":0,"end_line":98,"end_character":14},"in_reply_to":"ba0121b8_bde51239","updated":"2016-04-01 15:45:10.000000000","message":"Sure.  I kind of assumed get/show/list are in by default, but it\u0027s better to call them out.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"eac3aaf91c48deeb3cefc1e9acbeaf448bfd5c28","unresolved":false,"context_lines":[{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"Discussions on whether a group type is needed:"},{"line_number":103,"context_line":"(1) Introduce a group type (that is different from the volume type) for"},{"line_number":104,"context_line":"    the groups."},{"line_number":105,"context_line":"(2) Just use volume types (a group can support multiple volume types)."},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_3df009c9","line":104,"range":{"start_line":103,"start_character":0,"end_line":104,"end_character":15},"updated":"2016-03-30 20:57:40.000000000","message":"+1 for group types, I am a fan of decoupling these a little bit. I don\u0027t really like the idea of having group specifications on the individual volumes. As an example, it lets us keep the door open for volumes being replicated with cheesecake (using replication_enabled\u003dTrue) *and* replicated as part of a user facing CG or something (using a group type that has replication_enable\u003dTrue) without the need to try and cobble together some volume type thats replication_type\u003dvolume \u0026 group. Does that make sense?","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"004d131004a22099569db20301ce96dcb5690192","unresolved":false,"context_lines":[{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"Discussions on whether a group type is needed:"},{"line_number":103,"context_line":"(1) Introduce a group type (that is different from the volume type) for"},{"line_number":104,"context_line":"    the groups."},{"line_number":105,"context_line":"(2) Just use volume types (a group can support multiple volume types)."},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_81567504","line":104,"range":{"start_line":103,"start_character":0,"end_line":104,"end_character":15},"in_reply_to":"9a061dce_c3c28784","updated":"2016-04-08 21:55:03.000000000","message":"+1 for group types.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":20531,"name":"Jim Holl","email":"jimholl@netapp.com","username":"jimholl"},"change_message_id":"3c6175570d131a0dcfa8e83cbd0a0be2a668d132","unresolved":false,"context_lines":[{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"Discussions on whether a group type is needed:"},{"line_number":103,"context_line":"(1) Introduce a group type (that is different from the volume type) for"},{"line_number":104,"context_line":"    the groups."},{"line_number":105,"context_line":"(2) Just use volume types (a group can support multiple volume types)."},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_c3c28784","line":104,"range":{"start_line":103,"start_character":0,"end_line":104,"end_character":15},"in_reply_to":"ba0121b8_33e57408","updated":"2016-04-07 15:50:40.000000000","message":"+1 for group types. I think combining volume and group attributes into volumeType might be confusing. The scheduler will need to know the required volumesTypes at group create time.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":100,"context_line":"and delete_group as well."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"Discussions on whether a group type is needed:"},{"line_number":103,"context_line":"(1) Introduce a group type (that is different from the volume type) for"},{"line_number":104,"context_line":"    the groups."},{"line_number":105,"context_line":"(2) Just use volume types (a group can support multiple volume types)."},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_33e57408","line":104,"range":{"start_line":103,"start_character":0,"end_line":104,"end_character":15},"in_reply_to":"ba0121b8_3df009c9","updated":"2016-04-01 15:45:10.000000000","message":"Personally I prefer less than more.  In the existing CG implementation, the schedule will choose a backend that supports all volume types requested by the user.  I think we can do the same for this generic group.  I\u0027ll leave this open and see what is the consensus.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":2243,"name":"John Griffith","email":"john.griffith8@gmail.com","username":"john-griffith"},"change_message_id":"3546514b1d70b15c1aca944dc6b070296c4faa3e","unresolved":false,"context_lines":[{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_1e623b3e","line":113,"range":{"start_line":113,"start_character":8,"end_line":113,"end_character":10},"updated":"2016-04-07 03:15:07.000000000","message":"I agree there\u0027s some risk of making things a mess depending on how this is implemented.  I was actually thinking at one point that CG\u0027s as an entity go away as a unique thing and are some specially defined grouping.  That would\u0027ve made the original implementation of that much easier anyway IMO.\n\nThe trouble is everybody wants a special API for things.  My view is don\u0027t do that.  Create a group, let the driver figure out what that means and do the right thing.  Keep base functionality simple and robust, let the drivers perform hackery and tricks if they so desire.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":10221,"name":"Edward Balduf","email":"ed.balduf@solidfire.com","username":"ebalduf"},"change_message_id":"9932e6b28c817bd572d62739c6a4deb14c26a734","unresolved":false,"context_lines":[{"line_number":107,"context_line":"Alternatives"},{"line_number":108,"context_line":"------------"},{"line_number":109,"context_line":""},{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_5dd036f6","line":113,"range":{"start_line":110,"start_character":0,"end_line":113,"end_character":10},"updated":"2016-03-31 14:24:44.000000000","message":"If this can be built into ConsitencyGroups, it would be good.  To many types of groups could be confusing.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":20531,"name":"Jim Holl","email":"jimholl@netapp.com","username":"jimholl"},"change_message_id":"3c6175570d131a0dcfa8e83cbd0a0be2a668d132","unresolved":false,"context_lines":[{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_60d3b160","line":113,"range":{"start_line":113,"start_character":8,"end_line":113,"end_character":10},"in_reply_to":"9a061dce_1e623b3e","updated":"2016-04-07 15:50:40.000000000","message":"I think making consistency an attribute of a generic group construct has several advantages. Besides being able to replicate a set of inconsistent volumes, we could be more explicit about what is meant by consistency. For example, an attribute meaning the group is replicated in a way that preserves write ordering vs. point-in-time snapshots spanning volumes.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":10221,"name":"Edward Balduf","email":"ed.balduf@solidfire.com","username":"ebalduf"},"change_message_id":"d20e65705ffde8ab39c9e820e682ef9d49f25e23","unresolved":false,"context_lines":[{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_59bf8460","line":113,"range":{"start_line":113,"start_character":8,"end_line":113,"end_character":10},"in_reply_to":"9a061dce_1e623b3e","updated":"2016-04-07 15:53:00.000000000","message":"In speaking with the NTAP folks here, they have some additional needs for Generic groups. They will comment with details, therefore; my comments about doing all in one \u0027group\u0027 construct is still appropriate, but there is a need for replication groups and consistency groups.  Lets combine into a generic group construct. I agree with John\u0027s thoughts that the \u0027\u0027group\u0027 work is probably a separate spec/project.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"2afafdf36e8c0b1c7b7a2ab1189232a2c62c6c44","unresolved":false,"context_lines":[{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_5ff6285d","line":113,"range":{"start_line":113,"start_character":8,"end_line":113,"end_character":10},"in_reply_to":"9a061dce_59bf8460","updated":"2016-04-08 03:36:11.000000000","message":"Giving all the feedback from the reviewers so far, I\u0027ll create one spec for a generic group, and another spec for replication group.\n\nAfter we are done with the above, we can add support for CG snapshot using the generic group.  The existing CG APIs can be redirected to use the generic groups but we need to make sure we don\u0027t break rolling upgrades (they can be deprecated when the time is right).\n\nFeel free to add more comments.  I\u0027ll also sync up with Manila folks as they are also designing the generic groups in Manila.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"004d131004a22099569db20301ce96dcb5690192","unresolved":false,"context_lines":[{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"9a061dce_61f7e1f5","line":113,"range":{"start_line":113,"start_character":8,"end_line":113,"end_character":10},"in_reply_to":"9a061dce_5ff6285d","updated":"2016-04-08 21:55:03.000000000","message":"+1 to a separate spec for groups.  And I agree with John\u0027s comments above.  Generic groups with a group type of \u0027consistent snapshots\u0027 move most of the CG complexity into the drivers, and the core grouping construct is then reusable for other things like replication, and the API stays clean and uniform.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"7b51db6700871da9b0da94cee0e5352b91d1c17b","unresolved":false,"context_lines":[{"line_number":107,"context_line":"Alternatives"},{"line_number":108,"context_line":"------------"},{"line_number":109,"context_line":""},{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_a1765603","line":113,"range":{"start_line":110,"start_character":0,"end_line":113,"end_character":10},"in_reply_to":"ba0121b8_5dd036f6","updated":"2016-04-01 14:15:03.000000000","message":"I agree we should try to keep this as simple as possible - it\u0027s already pretty complex, so linking with CGs would be my preferred way to go as well.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":107,"context_line":"Alternatives"},{"line_number":108,"context_line":"------------"},{"line_number":109,"context_line":""},{"line_number":110,"context_line":"Do not add a VolumeGroup construct. Instead, just use the existing consistency"},{"line_number":111,"context_line":"group to implement group replication. Do we need to support replicating a group"},{"line_number":112,"context_line":"of volumes that does not ensure data consistency (volumes loosed grouped"},{"line_number":113,"context_line":"together)?"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"Data model impact"},{"line_number":116,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_9348c0cd","line":113,"range":{"start_line":110,"start_character":0,"end_line":113,"end_character":10},"in_reply_to":"ba0121b8_a1765603","updated":"2016-04-01 15:45:10.000000000","message":"Actually I prefer to use the existing CG without adding this generic group construct, but it seems that people at the mid-cycle prefer to have this.\n\nIf we decide to add this generic group, CG can be a special case of this group and we can add a foreign key of the generic group UUID in the CG table to link them together.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"eac3aaf91c48deeb3cefc1e9acbeaf448bfd5c28","unresolved":false,"context_lines":[{"line_number":130,"context_line":"* list_replication_targets"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"URL for the APIs:"},{"line_number":133,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replications/\u003cresource_id\u003e/action\u0027"},{"line_number":134,"context_line":""},{"line_number":135,"context_line":"resource_id can be one of the following:"},{"line_number":136,"context_line":"* UUID of a volume"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_5d67bd4c","line":133,"range":{"start_line":133,"start_character":2,"end_line":133,"end_character":4},"updated":"2016-03-30 20:57:40.000000000","message":"v3 now right? :D","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":130,"context_line":"* list_replication_targets"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"URL for the APIs:"},{"line_number":133,"context_line":"\u0027/v2/\u003ctenant_id\u003e/replications/\u003cresource_id\u003e/action\u0027"},{"line_number":134,"context_line":""},{"line_number":135,"context_line":"resource_id can be one of the following:"},{"line_number":136,"context_line":"* UUID of a volume"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_93760004","line":133,"range":{"start_line":133,"start_character":2,"end_line":133,"end_character":4},"in_reply_to":"ba0121b8_5d67bd4c","updated":"2016-04-01 15:45:10.000000000","message":"I guess so:).","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":10221,"name":"Edward Balduf","email":"ed.balduf@solidfire.com","username":"ebalduf"},"change_message_id":"9932e6b28c817bd572d62739c6a4deb14c26a734","unresolved":false,"context_lines":[{"line_number":203,"context_line":"---------------"},{"line_number":204,"context_line":""},{"line_number":205,"context_line":""},{"line_number":206,"context_line":"Notifications impact"},{"line_number":207,"context_line":"--------------------"},{"line_number":208,"context_line":""},{"line_number":209,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_1d5d1e79","line":206,"range":{"start_line":206,"start_character":0,"end_line":206,"end_character":20},"updated":"2016-03-31 14:24:44.000000000","message":"enable_replication and disable_replication calls need to emit notifications for Ceilometer to account for these changes.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":203,"context_line":"---------------"},{"line_number":204,"context_line":""},{"line_number":205,"context_line":""},{"line_number":206,"context_line":"Notifications impact"},{"line_number":207,"context_line":"--------------------"},{"line_number":208,"context_line":""},{"line_number":209,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_b37ea4e1","line":206,"range":{"start_line":206,"start_character":0,"end_line":206,"end_character":20},"in_reply_to":"ba0121b8_1d5d1e79","updated":"2016-04-01 15:45:10.000000000","message":"Sure.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":10221,"name":"Edward Balduf","email":"ed.balduf@solidfire.com","username":"ebalduf"},"change_message_id":"9932e6b28c817bd572d62739c6a4deb14c26a734","unresolved":false,"context_lines":[{"line_number":220,"context_line":""},{"line_number":221,"context_line":"* create_group"},{"line_number":222,"context_line":"* update_group"},{"line_number":223,"context_line":"* delete_group"},{"line_number":224,"context_line":""},{"line_number":225,"context_line":"Performance Impact"},{"line_number":226,"context_line":"------------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_5d8836c5","line":223,"range":{"start_line":223,"start_character":0,"end_line":223,"end_character":14},"updated":"2016-03-31 14:24:44.000000000","message":"list_group,\nshow_group","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"855b38fbcadfbe82ee33c3ff727f4508f9db4233","unresolved":false,"context_lines":[{"line_number":220,"context_line":""},{"line_number":221,"context_line":"* create_group"},{"line_number":222,"context_line":"* update_group"},{"line_number":223,"context_line":"* delete_group"},{"line_number":224,"context_line":""},{"line_number":225,"context_line":"Performance Impact"},{"line_number":226,"context_line":"------------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"ba0121b8_731a4cc7","line":223,"range":{"start_line":223,"start_character":0,"end_line":223,"end_character":14},"in_reply_to":"ba0121b8_5d8836c5","updated":"2016-04-01 15:45:10.000000000","message":"Sure.","commit_id":"f2352bf7d92ae14a0c0db9c38b6598b409f902a3"},{"author":{"_account_id":11904,"name":"Sean McGinnis","email":"sean.mcginnis@gmail.com","username":"SeanM"},"change_message_id":"c4756c6e35c208f84b859ceba9771f04fff1e4f1","unresolved":false,"context_lines":[{"line_number":30,"context_line":"There are other cases where the customers also want to have more control"},{"line_number":31,"context_line":"over the granularity of replication. The ability to replicate a group"},{"line_number":32,"context_line":"of volumes and test failover without affecting the entire backend is"},{"line_number":33,"context_line":"also a desired feature. A tenant facing API is needed to meet this"},{"line_number":34,"context_line":"requirement because only a tenant knows what volumes should be grouped"},{"line_number":35,"context_line":"together for his/her application."},{"line_number":36,"context_line":""}],"source_content_type":"text/x-rst","patch_set":16,"id":"5a18252c_b42d27aa","line":33,"updated":"2016-04-14 21:02:07.000000000","message":"I\u0027m a little worried (OK, a lot) about making this a tenant API until or unless we have some integration with Nova. Failover at the Cinder level is certainly possible for the tenant to control, but it may not be obvious that there isn\u0027t any integration and their VMs will crash unless they perform a lot of manual operations first to move things over.","commit_id":"fee0947ebd5e9515e5d6588ffdf282c1d5511186"},{"author":{"_account_id":7198,"name":"Jay Bryant","email":"jungleboyj@electronicjungle.net","username":"jsbryant"},"change_message_id":"2b691885f160faed513f961fc45211828a0dcf02","unresolved":false,"context_lines":[{"line_number":30,"context_line":"There are other cases where the customers also want to have more control"},{"line_number":31,"context_line":"over the granularity of replication. The ability to replicate a group"},{"line_number":32,"context_line":"of volumes and test failover without affecting the entire backend is"},{"line_number":33,"context_line":"also a desired feature. A tenant facing API is needed to meet this"},{"line_number":34,"context_line":"requirement because only a tenant knows what volumes should be grouped"},{"line_number":35,"context_line":"together for his/her application."},{"line_number":36,"context_line":""}],"source_content_type":"text/x-rst","patch_set":16,"id":"1a122d0e_ce87c7d5","line":33,"in_reply_to":"5a18252c_b42d27aa","updated":"2016-04-18 20:55:52.000000000","message":"Agree that this is a concern.  Also don\u0027t think we want to completely hold up tiramisu behind this.  Can we keep moving forward with this concept and have the API \u0027admin only\u0027 by default and let administrators configure it to be tenant facing if they so choose?  Once the issues with Nova are addressed we can change the default policy.","commit_id":"fee0947ebd5e9515e5d6588ffdf282c1d5511186"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"925e5c39e6a572fc1b423b236a36086dad2cd65b","unresolved":false,"context_lines":[{"line_number":227,"context_line":"   group_replication_enabled."},{"line_number":228,"context_line":"2. Add group actions to support group replication."},{"line_number":229,"context_line":""},{"line_number":230,"context_line":"Dependencies"},{"line_number":231,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":232,"context_line":"This depends on the Generic Volume Group spec."},{"line_number":233,"context_line":""}],"source_content_type":"text/x-rst","patch_set":16,"id":"1a122d0e_9f7e52e2","line":230,"range":{"start_line":230,"start_character":0,"end_line":230,"end_character":12},"updated":"2016-04-27 13:43:58.000000000","message":"I think it was the right thing to split out groups into a separate spec.  But I wonder if a phased approach to Tiramisu makes sense.  You could implement Tiramisu for single volumes in parallel with building the grouping construct, and then enable Tiramisu for groups in the next release.  It\u0027s great to be ambitious, but I think doing groups and replication for groups in one release is a stretch.","commit_id":"fee0947ebd5e9515e5d6588ffdf282c1d5511186"},{"author":{"_account_id":11865,"name":"Clinton Knight","email":"cknight@netapp.com","username":"cknight"},"change_message_id":"925e5c39e6a572fc1b423b236a36086dad2cd65b","unresolved":false,"context_lines":[{"line_number":231,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":232,"context_line":"This depends on the Generic Volume Group spec."},{"line_number":233,"context_line":""},{"line_number":234,"context_line":"Testing"},{"line_number":235,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":236,"context_line":""},{"line_number":237,"context_line":"New unit tests will be added to test the changed code."}],"source_content_type":"text/x-rst","patch_set":16,"id":"1a122d0e_fffae675","line":234,"range":{"start_line":234,"start_character":0,"end_line":234,"end_character":7},"updated":"2016-04-27 13:43:58.000000000","message":"I think you should cover Tempest testing as well.  This feature should have support in a 1st-party driver with associated Tempest tests.","commit_id":"fee0947ebd5e9515e5d6588ffdf282c1d5511186"},{"author":{"_account_id":2023,"name":"Saggi Mizrahi","email":"ficoos+osreview@gmail.com","username":"ficoos"},"change_message_id":"b5065f60b9064ce6d7638b7a4453b56e45edd07f","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"dab17558_12de2511","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"updated":"2016-05-08 12:01:27.000000000","message":"Is there a way for external services (eg. Smaug) to manage the groups for the tenant.\n\nAs an example. If we detect that there is a VM with 2 volumes and they are not part of a consistency group but they user wants to replicate them. We should be able to set it up for the tenant.\n\nFurther more, since we will depend on their existence for the validity of our protection plan. Can we hide\\namespace groups so that they can be modified\\removed only through a service?\n\nIn order to be able to perform operation across different OpenStack services we need to be aware or even prevent changes to consistency groups.\n\nFor example, if the user disables a consistency group through Cinder, the user will also have to disable the protection plan in Smaug so that it doesn\u0027t keep failing due to replication not being set up.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"f2ebb9b3c248d6f4d834946c8f6030251c7ffafc","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"bab6814e_37f345a1","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"bab6814e_4421f9fb","updated":"2016-05-25 09:17:18.000000000","message":"I\u0027d be very, very resistant to things like reference count hacks being added to the cinder API. If we add the ability for a service like smaug to have partial ownership of cinder entities, then we need to design it to be explicit, inspectable and obvious.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":19346,"name":"yinwei","email":"yinweiishere@icloud.com","username":"yinwei"},"change_message_id":"1494036664a0f05e19fa78a0308c858236154e99","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"bab6814e_f6406da7","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"bab6814e_576889ba","updated":"2016-05-26 02:55:36.000000000","message":"Thanks, Thomas! I didn\u0027t intend to introduce this change into tiramisu, which are two different issues. We will propose a separate BP to describe the requirement and changes in detail.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"08ea9f5a6c28943be14e6c0f2ed3a25b05dc02c6","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"3aaa91ec_479f8d50","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"bab6814e_f6406da7","updated":"2016-06-27 15:09:26.000000000","message":"I modified the spec to indicate that Smaug can use Cinder APIs to manage the groups.  Also added a note about the requirement to hide/lock the namespace and that the Smaug team will submit a separate spec to discuss about it.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"6373bbf574afc8eed4e0d3a584a6f288e5120f55","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"dab17558_f708abdb","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"dab17558_12de2511","updated":"2016-05-09 03:34:23.000000000","message":"This spec is built upon the Generic Volume Group spec: https://review.openstack.org/#/c/303893/.\n\nIn that spec, there are APIs to create/delete/update/list/show a group. You can use those APIs to manage a group.\n\nYou are asking if the group can be modified only by a special tenant like a service? We can consider that. The group construct is a pretty generic one, so it could be used for other purposes. I wonder if making this restriction will limit what we can with the groups in other situations. When Smaug uses the group APIs to manage the groups, will it be acting as the Smaug service or a regular tenant?\n\nIf the user disables the replication group, Smaug can detect that by checking the group status.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"f2ebb9b3c248d6f4d834946c8f6030251c7ffafc","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"bab6814e_576889ba","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"dab17558_3202000c","updated":"2016-05-25 09:17:18.000000000","message":"Hiding things from the tenant view is not something we currently support, and is a very significant change to the way our UI works - it is not something that should be bought in via the back door or without careful consideration.\n\nIf smaug has requirements or desires for this, they should definitely come in via a separate spec, with details on the usecases, so that we can think them through carefully.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"8fd5d1bfea1a56ee43398947497f05ff537cd787","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"dab17558_806704b6","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"dab17558_3202000c","updated":"2016-05-15 17:34:55.000000000","message":"There will be a \"updated_at\" field in the group table.\n\nSo Smaug is impersonating the tenant, right?  The volumes are created by the tenant using Cinder APIs, not through Smaug, right?  If after the tenant adds the volumes he/she created to a group using Smaug, he/she can no longer see the volumes, that seems unexpected by the tenant?  I wonder if we can have a special mode, i.e., maintenance mode.  If the tenant uses Smaug to create the group and add volumes to the group, Smaug sets the group to be in maintenance mode.  After that only admin can modify the group but the group should still be visible to the tenant.\n\nI wonder how Heat handles this. Tenant launches a stack in Heat which has VMs, volumes, etc. If the tenant deletes the volume in Cinder, what happens to the stack in Heat?  We don\u0027t have any special code in Cinder to prevent that from happening.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":19346,"name":"yinwei","email":"yinweiishere@icloud.com","username":"yinwei"},"change_message_id":"10f3230fcc92384436eb4dc91c9fa9430801c893","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"bab6814e_4421f9fb","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"dab17558_806704b6","updated":"2016-05-25 09:01:06.000000000","message":"Yes, resource being protected by smaug may get deleted anonymously. This is an issue we actually have met. \n\nMaybe not related to CG replication. Wihtout checking the existence of resources, maybe we could refer to the case how to delete a volume in-use.  As one volume attached to a server, volume status will turn to be in-use to prevent deletion.  Is it possible for cinder to open another API for other services to update reference counters to one volume?  In this way, we will have a reference counting for safe resource deletion.","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":2023,"name":"Saggi Mizrahi","email":"ficoos+osreview@gmail.com","username":"ficoos"},"change_message_id":"ab3e209113a9ba06073a196089adb6c72f755950","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or a similar project can be the consumer of the replication API in"},{"line_number":130,"context_line":"Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."},{"line_number":133,"context_line":"* Admin sets up group type and volume types in Cinder."}],"source_content_type":"text/x-rst","patch_set":17,"id":"dab17558_3202000c","line":130,"range":{"start_line":129,"start_character":0,"end_line":130,"end_character":29},"in_reply_to":"dab17558_f708abdb","updated":"2016-05-10 10:38:51.000000000","message":"It will be performed by Smaug as the tenant. So this might be some sort of service\\tenant combo. In a more general sense. It\u0027s a tenant entity but it should only be managed through Smaug. We would like to hide\\lock\\namespace it so that when the tenant looks in horizon it isn\u0027t confused by all the groups it didn\u0027t manually create.\n\nWRT you suggestion, checking the group status can allow us to report an error in case of group deletion but group modification might cause subtle errors and we would prefer not keeping a replica of the group to check if it didn\u0027t change.\n\nThere is a \u0027created_at\u0027 field on a group already. Are there any plans to add a \u0027modified_at\u0027? Should this be another spec?","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":7198,"name":"Jay Bryant","email":"jungleboyj@electronicjungle.net","username":"jsbryant"},"change_message_id":"67ceb5d2311df305c08b7d9c93ecf3e86b89208d","unresolved":false,"context_lines":[{"line_number":137,"context_line":"* In Smaug, a tenant chooses what resources to put in a protection group."},{"line_number":138,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":139,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":140,"context_line":"  group. For example, Smaug can check whether a Cinde volume group has"},{"line_number":141,"context_line":"  ``group_replication_enabled`` or ``consistent_group_snapshot_enabled``"},{"line_number":142,"context_line":"  in its group specs."},{"line_number":143,"context_line":"* In Smaug, a protection action can be executed depending on the"}],"source_content_type":"text/x-rst","patch_set":17,"id":"dab17558_6f4fdb92","line":140,"range":{"start_line":140,"start_character":48,"end_line":140,"end_character":53},"updated":"2016-05-06 21:55:38.000000000","message":"Cinder","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"08ea9f5a6c28943be14e6c0f2ed3a25b05dc02c6","unresolved":false,"context_lines":[{"line_number":137,"context_line":"* In Smaug, a tenant chooses what resources to put in a protection group."},{"line_number":138,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":139,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":140,"context_line":"  group. For example, Smaug can check whether a Cinde volume group has"},{"line_number":141,"context_line":"  ``group_replication_enabled`` or ``consistent_group_snapshot_enabled``"},{"line_number":142,"context_line":"  in its group specs."},{"line_number":143,"context_line":"* In Smaug, a protection action can be executed depending on the"}],"source_content_type":"text/x-rst","patch_set":17,"id":"3aaa91ec_b64dc59b","line":140,"range":{"start_line":140,"start_character":48,"end_line":140,"end_character":53},"in_reply_to":"dab17558_6f4fdb92","updated":"2016-06-27 15:09:26.000000000","message":"Done","commit_id":"4346154ac27a0f3566e246be7e13f9153d9e8278"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"f21737e3408a1045f2dacba347e8b446ebc16ada","unresolved":false,"context_lines":[{"line_number":141,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":142,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":143,"context_line":"  group. For example, Smaug can check whether a Cinder volume group has"},{"line_number":144,"context_line":"  ``group_replication_enabled`` or ``consistent_group_snapshot_enabled``"},{"line_number":145,"context_line":"  in its group specs."},{"line_number":146,"context_line":""},{"line_number":147,"context_line":"* There are APIs to create/delete/update/list/show a group in the Generic"}],"source_content_type":"text/x-rst","patch_set":18,"id":"3aaa91ec_dd350ca9","line":144,"updated":"2016-06-27 15:30:44.000000000","message":"consistent_group_snapshot_enabled or consistent_group_replication_enabled???","commit_id":"4e54cbbf7320e1540328df41abd7525b5a514ba6"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"edcf7a9e9585180f47bc87c77cf6b32e3a75c619","unresolved":false,"context_lines":[{"line_number":141,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":142,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":143,"context_line":"  group. For example, Smaug can check whether a Cinder volume group has"},{"line_number":144,"context_line":"  ``group_replication_enabled`` or ``consistent_group_snapshot_enabled``"},{"line_number":145,"context_line":"  in its group specs."},{"line_number":146,"context_line":""},{"line_number":147,"context_line":"* There are APIs to create/delete/update/list/show a group in the Generic"}],"source_content_type":"text/x-rst","patch_set":18,"id":"3aaa91ec_fd01e86b","line":144,"in_reply_to":"3aaa91ec_dd350ca9","updated":"2016-06-27 15:34:46.000000000","message":"Oops:).  Good catch!  Fixed it.","commit_id":"4e54cbbf7320e1540328df41abd7525b5a514ba6"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"e0f2b1387b13d9face383581743d2702019fcc3b","unresolved":false,"context_lines":[{"line_number":113,"context_line":""},{"line_number":114,"context_line":"If a volume is \"in-use\", a \"allow-attached-volume\" flag needs to set to"},{"line_number":115,"context_line":"True for failover_replication to work. This is to prevent the tenant"},{"line_number":116,"context_line":"from abusing the failover API and causing problems when forcifully"},{"line_number":117,"context_line":"failing over an attached volume. Either the tenant or the admin should"},{"line_number":118,"context_line":"detach the volume before failing over, or it will be handled by a higher"},{"line_number":119,"context_line":"layer project such as Smaug."}],"source_content_type":"text/x-rst","patch_set":19,"id":"1aa78d24_028b2ac4","line":116,"range":{"start_line":116,"start_character":56,"end_line":116,"end_character":66},"updated":"2016-07-12 04:37:08.000000000","message":"forcefully or \u0027failing over an attached volume with the ``force`` flag\u0027.","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":113,"context_line":""},{"line_number":114,"context_line":"If a volume is \"in-use\", a \"allow-attached-volume\" flag needs to set to"},{"line_number":115,"context_line":"True for failover_replication to work. This is to prevent the tenant"},{"line_number":116,"context_line":"from abusing the failover API and causing problems when forcifully"},{"line_number":117,"context_line":"failing over an attached volume. Either the tenant or the admin should"},{"line_number":118,"context_line":"detach the volume before failing over, or it will be handled by a higher"},{"line_number":119,"context_line":"layer project such as Smaug."}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_0c7be97b","line":116,"range":{"start_line":116,"start_character":56,"end_line":116,"end_character":66},"in_reply_to":"1aa78d24_028b2ac4","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"dad027eeae6cd2fd59eff9193738c38187a3c12e","unresolved":false,"context_lines":[{"line_number":116,"context_line":"from abusing the failover API and causing problems when forcifully"},{"line_number":117,"context_line":"failing over an attached volume. Either the tenant or the admin should"},{"line_number":118,"context_line":"detach the volume before failing over, or it will be handled by a higher"},{"line_number":119,"context_line":"layer project such as Smaug."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"Group type needs to have the following extra_spec to support"},{"line_number":122,"context_line":"replication:"}],"source_content_type":"text/x-rst","patch_set":19,"id":"9a629dbe_713418de","line":119,"range":{"start_line":119,"start_character":22,"end_line":119,"end_character":27},"updated":"2016-11-07 10:56:41.000000000","message":"Karbor:)","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":116,"context_line":"from abusing the failover API and causing problems when forcifully"},{"line_number":117,"context_line":"failing over an attached volume. Either the tenant or the admin should"},{"line_number":118,"context_line":"detach the volume before failing over, or it will be handled by a higher"},{"line_number":119,"context_line":"layer project such as Smaug."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"Group type needs to have the following extra_spec to support"},{"line_number":122,"context_line":"replication:"}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_2c7eed69","line":119,"range":{"start_line":119,"start_character":22,"end_line":119,"end_character":27},"in_reply_to":"9a629dbe_713418de","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":2023,"name":"Saggi Mizrahi","email":"ficoos+osreview@gmail.com","username":"ficoos"},"change_message_id":"f3943efe97fbaa432cba899404ca04aea6254534","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or another data protection solution can be the consumer of the"},{"line_number":130,"context_line":"replication API in Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."}],"source_content_type":"text/x-rst","patch_set":19,"id":"3ac371cc_8cd2bd5a","line":129,"range":{"start_line":129,"start_character":0,"end_line":129,"end_character":5},"updated":"2016-08-17 08:53:57.000000000","message":"Karbor :)","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Driver also needs to report ``group_replication_enabled`` or"},{"line_number":127,"context_line":"``consistent_group_replication_enabled`` as capabilities."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"Smaug or another data protection solution can be the consumer of the"},{"line_number":130,"context_line":"replication API in Cinder. The flow is like this:"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"* Admin configures cinder.conf."}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_4c9961c3","line":129,"range":{"start_line":129,"start_character":0,"end_line":129,"end_character":5},"in_reply_to":"3ac371cc_8cd2bd5a","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":2023,"name":"Saggi Mizrahi","email":"ficoos+osreview@gmail.com","username":"ficoos"},"change_message_id":"f3943efe97fbaa432cba899404ca04aea6254534","unresolved":false,"context_lines":[{"line_number":137,"context_line":""},{"line_number":138,"context_line":"* Tenant creates volumes and puts into the group in Cinder."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"* In Smaug, a tenant chooses what resources to put in a protection group."},{"line_number":141,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":142,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":143,"context_line":"  group. For example, Smaug can check whether a Cinder volume group has"}],"source_content_type":"text/x-rst","patch_set":19,"id":"3ac371cc_6c3e618d","line":140,"range":{"start_line":140,"start_character":5,"end_line":140,"end_character":10},"updated":"2016-08-17 08:53:57.000000000","message":"Karbor","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":137,"context_line":""},{"line_number":138,"context_line":"* Tenant creates volumes and puts into the group in Cinder."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"* In Smaug, a tenant chooses what resources to put in a protection group."},{"line_number":141,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":142,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":143,"context_line":"  group. For example, Smaug can check whether a Cinder volume group has"}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_6c94658a","line":140,"range":{"start_line":140,"start_character":5,"end_line":140,"end_character":10},"in_reply_to":"3ac371cc_6c3e618d","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":2023,"name":"Saggi Mizrahi","email":"ficoos+osreview@gmail.com","username":"ficoos"},"change_message_id":"f3943efe97fbaa432cba899404ca04aea6254534","unresolved":false,"context_lines":[{"line_number":138,"context_line":"* Tenant creates volumes and puts into the group in Cinder."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"* In Smaug, a tenant chooses what resources to put in a protection group."},{"line_number":141,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":142,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":143,"context_line":"  group. For example, Smaug can check whether a Cinder volume group has"},{"line_number":144,"context_line":"  ``group_replication_enabled`` or ``consistent_group_replication_enabled``"}],"source_content_type":"text/x-rst","patch_set":19,"id":"3ac371cc_0c4ecd3a","line":141,"range":{"start_line":141,"start_character":21,"end_line":141,"end_character":26},"updated":"2016-08-17 08:53:57.000000000","message":"Karbor","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":138,"context_line":"* Tenant creates volumes and puts into the group in Cinder."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"* In Smaug, a tenant chooses what resources to put in a protection group."},{"line_number":141,"context_line":"  For Cinder volume, Smaug can detect whether a selected volume belongs"},{"line_number":142,"context_line":"  to a replication group and automatically selects all volumes in that"},{"line_number":143,"context_line":"  group. For example, Smaug can check whether a Cinder volume group has"},{"line_number":144,"context_line":"  ``group_replication_enabled`` or ``consistent_group_replication_enabled``"}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_8c8fd973","line":141,"range":{"start_line":141,"start_character":21,"end_line":141,"end_character":26},"in_reply_to":"3ac371cc_0c4ecd3a","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"dad027eeae6cd2fd59eff9193738c38187a3c12e","unresolved":false,"context_lines":[{"line_number":145,"context_line":"  in its group specs."},{"line_number":146,"context_line":""},{"line_number":147,"context_line":"* There are APIs to create/delete/update/list/show a group in the Generic"},{"line_number":148,"context_line":"  Volume Group spec [1]. Smaug or another data protection solution can use"},{"line_number":149,"context_line":"  those APIs to manage the groups for the tenant."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":"* In Smaug, a protection action can be executed depending on the"}],"source_content_type":"text/x-rst","patch_set":19,"id":"9a629dbe_3168e0ba","line":148,"range":{"start_line":148,"start_character":25,"end_line":148,"end_character":30},"updated":"2016-11-07 10:56:41.000000000","message":"Karbor:)","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":145,"context_line":"  in its group specs."},{"line_number":146,"context_line":""},{"line_number":147,"context_line":"* There are APIs to create/delete/update/list/show a group in the Generic"},{"line_number":148,"context_line":"  Volume Group spec [1]. Smaug or another data protection solution can use"},{"line_number":149,"context_line":"  those APIs to manage the groups for the tenant."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":"* In Smaug, a protection action can be executed depending on the"}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_ac8add61","line":148,"range":{"start_line":148,"start_character":25,"end_line":148,"end_character":30},"in_reply_to":"9a629dbe_3168e0ba","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"dad027eeae6cd2fd59eff9193738c38187a3c12e","unresolved":false,"context_lines":[{"line_number":148,"context_line":"  Volume Group spec [1]. Smaug or another data protection solution can use"},{"line_number":149,"context_line":"  those APIs to manage the groups for the tenant."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":"* In Smaug, a protection action can be executed depending on the"},{"line_number":152,"context_line":"  protection policies. If the protection action triggers a fail over,"},{"line_number":153,"context_line":"  Smaug can call Cinder\u0027s replication API to failover the replication"},{"line_number":154,"context_line":"  group."}],"source_content_type":"text/x-rst","patch_set":19,"id":"9a629dbe_915fec0d","line":151,"range":{"start_line":151,"start_character":5,"end_line":151,"end_character":10},"updated":"2016-11-07 10:56:41.000000000","message":"ditto","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":148,"context_line":"  Volume Group spec [1]. Smaug or another data protection solution can use"},{"line_number":149,"context_line":"  those APIs to manage the groups for the tenant."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":"* In Smaug, a protection action can be executed depending on the"},{"line_number":152,"context_line":"  protection policies. If the protection action triggers a fail over,"},{"line_number":153,"context_line":"  Smaug can call Cinder\u0027s replication API to failover the replication"},{"line_number":154,"context_line":"  group."}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_cceb9111","line":151,"range":{"start_line":151,"start_character":5,"end_line":151,"end_character":10},"in_reply_to":"9a629dbe_915fec0d","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":2023,"name":"Saggi Mizrahi","email":"ficoos+osreview@gmail.com","username":"ficoos"},"change_message_id":"f3943efe97fbaa432cba899404ca04aea6254534","unresolved":false,"context_lines":[{"line_number":153,"context_line":"  Smaug can call Cinder\u0027s replication API to failover the replication"},{"line_number":154,"context_line":"  group."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Note: The Smaug team has a requirement for Cinder to hide/lock the namespace"},{"line_number":157,"context_line":"of a group and its volumes once they are under Smaug\u0027s protection so that"},{"line_number":158,"context_line":"they will not be deleted/modified without Smaug\u0027s knowledge. Since Cinder"},{"line_number":159,"context_line":"does not currently hide things from the tenant, this is very different from"}],"source_content_type":"text/x-rst","patch_set":19,"id":"3ac371cc_2c68e986","line":156,"range":{"start_line":156,"start_character":10,"end_line":156,"end_character":15},"updated":"2016-08-17 08:53:57.000000000","message":"Karbor\n\nAnd so on...","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":153,"context_line":"  Smaug can call Cinder\u0027s replication API to failover the replication"},{"line_number":154,"context_line":"  group."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Note: The Smaug team has a requirement for Cinder to hide/lock the namespace"},{"line_number":157,"context_line":"of a group and its volumes once they are under Smaug\u0027s protection so that"},{"line_number":158,"context_line":"they will not be deleted/modified without Smaug\u0027s knowledge. Since Cinder"},{"line_number":159,"context_line":"does not currently hide things from the tenant, this is very different from"}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_0cf20906","line":156,"range":{"start_line":156,"start_character":10,"end_line":156,"end_character":15},"in_reply_to":"3ac371cc_2c68e986","updated":"2017-02-15 22:39:10.000000000","message":"Done","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":2023,"name":"Saggi Mizrahi","email":"ficoos+osreview@gmail.com","username":"ficoos"},"change_message_id":"f3943efe97fbaa432cba899404ca04aea6254534","unresolved":false,"context_lines":[{"line_number":153,"context_line":"  Smaug can call Cinder\u0027s replication API to failover the replication"},{"line_number":154,"context_line":"  group."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Note: The Smaug team has a requirement for Cinder to hide/lock the namespace"},{"line_number":157,"context_line":"of a group and its volumes once they are under Smaug\u0027s protection so that"},{"line_number":158,"context_line":"they will not be deleted/modified without Smaug\u0027s knowledge. Since Cinder"},{"line_number":159,"context_line":"does not currently hide things from the tenant, this is very different from"},{"line_number":160,"context_line":"what Cinder currently support. Also the issue that a resource being protected"},{"line_number":161,"context_line":"by Smaug might be deleted anonymously is not specific to Tiramisu. We"},{"line_number":162,"context_line":"recommend the Smaug team to submit a separate spec to discuss about this"},{"line_number":163,"context_line":"requirement."},{"line_number":164,"context_line":""},{"line_number":165,"context_line":"Alternatives"},{"line_number":166,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":19,"id":"3ac371cc_ace059fb","line":163,"range":{"start_line":156,"start_character":0,"end_line":163,"end_character":12},"updated":"2016-08-17 08:53:57.000000000","message":"As discussed in it\u0027s own BP, if a hiding\\locking feature is ever added it will have to be added to all\\most entities and not just replication-groups so it can be omitted from this BP.","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"dad027eeae6cd2fd59eff9193738c38187a3c12e","unresolved":false,"context_lines":[{"line_number":153,"context_line":"  Smaug can call Cinder\u0027s replication API to failover the replication"},{"line_number":154,"context_line":"  group."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Note: The Smaug team has a requirement for Cinder to hide/lock the namespace"},{"line_number":157,"context_line":"of a group and its volumes once they are under Smaug\u0027s protection so that"},{"line_number":158,"context_line":"they will not be deleted/modified without Smaug\u0027s knowledge. Since Cinder"},{"line_number":159,"context_line":"does not currently hide things from the tenant, this is very different from"},{"line_number":160,"context_line":"what Cinder currently support. Also the issue that a resource being protected"},{"line_number":161,"context_line":"by Smaug might be deleted anonymously is not specific to Tiramisu. We"},{"line_number":162,"context_line":"recommend the Smaug team to submit a separate spec to discuss about this"},{"line_number":163,"context_line":"requirement."},{"line_number":164,"context_line":""},{"line_number":165,"context_line":"Alternatives"},{"line_number":166,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":19,"id":"9a629dbe_b1e87018","line":163,"range":{"start_line":156,"start_character":0,"end_line":163,"end_character":12},"in_reply_to":"3ac371cc_ace059fb","updated":"2016-11-07 10:56:41.000000000","message":"Any bp\\spec link here?","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"a5dac05d69691ac71f6c6c2916691664182c3999","unresolved":false,"context_lines":[{"line_number":153,"context_line":"  Smaug can call Cinder\u0027s replication API to failover the replication"},{"line_number":154,"context_line":"  group."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Note: The Smaug team has a requirement for Cinder to hide/lock the namespace"},{"line_number":157,"context_line":"of a group and its volumes once they are under Smaug\u0027s protection so that"},{"line_number":158,"context_line":"they will not be deleted/modified without Smaug\u0027s knowledge. Since Cinder"},{"line_number":159,"context_line":"does not currently hide things from the tenant, this is very different from"},{"line_number":160,"context_line":"what Cinder currently support. Also the issue that a resource being protected"},{"line_number":161,"context_line":"by Smaug might be deleted anonymously is not specific to Tiramisu. We"},{"line_number":162,"context_line":"recommend the Smaug team to submit a separate spec to discuss about this"},{"line_number":163,"context_line":"requirement."},{"line_number":164,"context_line":""},{"line_number":165,"context_line":"Alternatives"},{"line_number":166,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":19,"id":"fa31d9ce_acb81d12","line":163,"range":{"start_line":156,"start_character":0,"end_line":163,"end_character":12},"in_reply_to":"9a629dbe_b1e87018","updated":"2017-02-15 22:39:10.000000000","message":"Removed this note from this bp.\n\nHere\u0027s the spec submitted by Saggi:\nhttps://review.openstack.org/#/c/348469/","commit_id":"f979d6863ab9cfde8f9598ba0eed035a07377505"}],"specs/pike/replication-group.rst":[{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"90d3dd336d0d65f98cfe81ec700bcacdc7ec3651","unresolved":false,"context_lines":[{"line_number":228,"context_line":"            \u0027failover_replication\u0027:"},{"line_number":229,"context_line":"            {"},{"line_number":230,"context_line":"                \u0027allow-attached-volume\u0027: False,"},{"line_number":231,"context_line":"                \u0027secondary_backend_id\u0027: \u0027xxxxxxxx\u0027,"},{"line_number":232,"context_line":"            }"},{"line_number":233,"context_line":"        }"},{"line_number":234,"context_line":""}],"source_content_type":"text/x-rst","patch_set":23,"id":"ba2be162_67e1379a","line":231,"range":{"start_line":231,"start_character":0,"end_line":231,"end_character":51},"updated":"2017-03-03 03:35:21.000000000","message":"Where can the tenant get this ID from?\n\nFor cheesecake, the secondary_backend_id is the replication target (\"replication_device\") configured by the administrator. These can be backend specific names. In the NetApp driver, the backend_id corresponds to a configuration section where the secondary backend\u0027s configuration lives. Typically this is hidden away from the tenant..","commit_id":"b03608d45c9ed9ffd12bc47995a65569aa1f3a2d"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"8635b6b13ceffd46f70313cd5b145bfc646ac433","unresolved":false,"context_lines":[{"line_number":228,"context_line":"            \u0027failover_replication\u0027:"},{"line_number":229,"context_line":"            {"},{"line_number":230,"context_line":"                \u0027allow-attached-volume\u0027: False,"},{"line_number":231,"context_line":"                \u0027secondary_backend_id\u0027: \u0027xxxxxxxx\u0027,"},{"line_number":232,"context_line":"            }"},{"line_number":233,"context_line":"        }"},{"line_number":234,"context_line":""}],"source_content_type":"text/x-rst","patch_set":23,"id":"ba2be162_4924335f","line":231,"range":{"start_line":231,"start_character":0,"end_line":231,"end_character":51},"in_reply_to":"ba2be162_05ea33db","updated":"2017-03-03 14:38:15.000000000","message":"I was never expecting the tenants to fully understand the replication configurations. The admin is the one configuring the replication targets for backends and volume types. That is their responsibility. The tenant can just see which are available for a given group type - the names of these can be anything defined by the admin. So they can use domain A, domain B, whatever.","commit_id":"b03608d45c9ed9ffd12bc47995a65569aa1f3a2d"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"4d997913b431634e9d6ba73419395cf960dba46c","unresolved":false,"context_lines":[{"line_number":228,"context_line":"            \u0027failover_replication\u0027:"},{"line_number":229,"context_line":"            {"},{"line_number":230,"context_line":"                \u0027allow-attached-volume\u0027: False,"},{"line_number":231,"context_line":"                \u0027secondary_backend_id\u0027: \u0027xxxxxxxx\u0027,"},{"line_number":232,"context_line":"            }"},{"line_number":233,"context_line":"        }"},{"line_number":234,"context_line":""}],"source_content_type":"text/x-rst","patch_set":23,"id":"ba2be162_05ea33db","line":231,"range":{"start_line":231,"start_character":0,"end_line":231,"end_character":51},"in_reply_to":"ba2be162_0a2962e6","updated":"2017-03-03 04:44:59.000000000","message":"I think end users don\u0027t/shouldn\u0027t know/care about backends. They only need to know that their replication targets are not in the same failure domain as the primary. I think we should invent/expose something a little more abstract:\n\n create volume group\n enable replication to failure-domain-A\n enable replication to failure-domain-B\n failover volume group to failure-domain-B\n\nHow about adding a configuration option in the primary backend that will allow the administrator to define what these failure domains are going to be called? They\u0027re much like names of Availability Zones, only, they make sense per back end.","commit_id":"b03608d45c9ed9ffd12bc47995a65569aa1f3a2d"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"1b3c41b9e7e291bc2a7bf845ca6a4a0115f8a1d5","unresolved":false,"context_lines":[{"line_number":228,"context_line":"            \u0027failover_replication\u0027:"},{"line_number":229,"context_line":"            {"},{"line_number":230,"context_line":"                \u0027allow-attached-volume\u0027: False,"},{"line_number":231,"context_line":"                \u0027secondary_backend_id\u0027: \u0027xxxxxxxx\u0027,"},{"line_number":232,"context_line":"            }"},{"line_number":233,"context_line":"        }"},{"line_number":234,"context_line":""}],"source_content_type":"text/x-rst","patch_set":23,"id":"ba2be162_e7d339d3","line":231,"range":{"start_line":231,"start_character":0,"end_line":231,"end_character":51},"in_reply_to":"ba2be162_4924335f","updated":"2017-03-03 21:58:07.000000000","message":"I\u0027m planning to use the config options already available for Cheesecake.  We have the following for Cheesecake:\n\nreplication_device \u003d backend_id:vendor-id-1,unique_key:val….\n\nbackend_id is a required key.  We can either use the backend_id as the secondary_backend_id above, or add another required field for Tiramisu:\n\nreplication_device \u003d failure_domain:domain-A,backend_id:vendor-id-1,unique_key:val….\n\nIn this case, \"failure_domain\" will replace \"secondary_backend_id\" in the API.  Internally we still know which backend_id is corresponding to the failure_domain.\n\nWhen admin runs list replication targets, he/she will get all the details.  When non-admin runs list replication targets, he/she will just get the failure_domain list.","commit_id":"b03608d45c9ed9ffd12bc47995a65569aa1f3a2d"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"c9ac90182f2fd00f3a16f46e388f881052e12e7b","unresolved":false,"context_lines":[{"line_number":228,"context_line":"            \u0027failover_replication\u0027:"},{"line_number":229,"context_line":"            {"},{"line_number":230,"context_line":"                \u0027allow-attached-volume\u0027: False,"},{"line_number":231,"context_line":"                \u0027secondary_backend_id\u0027: \u0027xxxxxxxx\u0027,"},{"line_number":232,"context_line":"            }"},{"line_number":233,"context_line":"        }"},{"line_number":234,"context_line":""}],"source_content_type":"text/x-rst","patch_set":23,"id":"ba2be162_0a2962e6","line":231,"range":{"start_line":231,"start_character":0,"end_line":231,"end_character":51},"in_reply_to":"ba2be162_67e1379a","updated":"2017-03-03 04:03:39.000000000","message":"I\u0027m assuming the list_replication_targets API will provide this information. It should provide a list of configured secondary_backed_id\u0027s for each backend. \nPresumably the tenant can then select an appropriate id which will be used with the enable_replication API with some checks to ensure the volume_group and secondary_backend_id are linked to the same backend","commit_id":"b03608d45c9ed9ffd12bc47995a65569aa1f3a2d"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"90d3dd336d0d65f98cfe81ec700bcacdc7ec3651","unresolved":false,"context_lines":[{"line_number":241,"context_line":"            \u0027list_replication_targets\u0027:"},{"line_number":242,"context_line":"            {"},{"line_number":243,"context_line":"            }"},{"line_number":244,"context_line":"        }"},{"line_number":245,"context_line":""},{"line_number":246,"context_line":"Security impact"},{"line_number":247,"context_line":"---------------"}],"source_content_type":"text/x-rst","patch_set":23,"id":"ba2be162_c758238d","line":244,"range":{"start_line":244,"start_character":0,"end_line":244,"end_character":9},"updated":"2017-03-03 03:35:21.000000000","message":"What will the response of this API look like?","commit_id":"b03608d45c9ed9ffd12bc47995a65569aa1f3a2d"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"c9ac90182f2fd00f3a16f46e388f881052e12e7b","unresolved":false,"context_lines":[{"line_number":241,"context_line":"            \u0027list_replication_targets\u0027:"},{"line_number":242,"context_line":"            {"},{"line_number":243,"context_line":"            }"},{"line_number":244,"context_line":"        }"},{"line_number":245,"context_line":""},{"line_number":246,"context_line":"Security impact"},{"line_number":247,"context_line":"---------------"}],"source_content_type":"text/x-rst","patch_set":23,"id":"ba2be162_2a6d4631","line":244,"range":{"start_line":244,"start_character":0,"end_line":244,"end_character":9},"in_reply_to":"ba2be162_c758238d","updated":"2017-03-03 04:03:39.000000000","message":"I think this question could be asked of all the APIs defined in this spec.","commit_id":"b03608d45c9ed9ffd12bc47995a65569aa1f3a2d"},{"author":{"_account_id":13425,"name":"Simon Dodsley","email":"simon@everpuredata.com","username":"sdodsley"},"change_message_id":"d736ade1377bb94e5dd38a0fa602eb3ace2e8317","unresolved":false,"context_lines":[{"line_number":272,"context_line":"               }"},{"line_number":273,"context_line":"            ]"},{"line_number":274,"context_line":"       }"},{"line_number":275,"context_line":""},{"line_number":276,"context_line":"  ** Response example for non-admin::"},{"line_number":277,"context_line":"       {"},{"line_number":278,"context_line":"           \u0027replication_targets\u0027: ["}],"source_content_type":"text/x-rst","patch_set":25,"id":"ba2be162_fc2e5541","line":275,"updated":"2017-03-05 21:30:16.000000000","message":"Do we need to add the additional failure_domain key into the replication config options? Seems like an unnecessary complication. The backend_id key in the replication setting doesn\u0027t actually have any bearing on a configured backend, given this can be something outside of Cinder\u0027s control. It\u0027s just a name that is being used. Yes, if that replication array is managed already by Cinder then it\u0027s helpful to call it the same, but if not then it\u0027s still just an arbitrary name. Differentiate the admin and tenant responses to the API so that the tenant sees the term failure_domain as that is more meaningful to them, but the admin knows what is happening - or at least they should as they set it up...","commit_id":"9e954601947ae42ee740d7eb9faf0fde54eaa659"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"faf1f328c7d584a3c24f15241e09ec618ef0095f","unresolved":false,"context_lines":[{"line_number":272,"context_line":"               }"},{"line_number":273,"context_line":"            ]"},{"line_number":274,"context_line":"       }"},{"line_number":275,"context_line":""},{"line_number":276,"context_line":"  ** Response example for non-admin::"},{"line_number":277,"context_line":"       {"},{"line_number":278,"context_line":"           \u0027replication_targets\u0027: ["}],"source_content_type":"text/x-rst","patch_set":25,"id":"ba2be162_2b16cc4b","line":275,"in_reply_to":"ba2be162_fc2e5541","updated":"2017-03-06 14:07:54.000000000","message":"Alright, removed \"failure_domain\" and let admin handle what is stored in \"backend_id\".  I still want to display different results for admin and non-admin.  non-admin can only see backend_id.  admin can see all details of the replication_targets.  This is similar to the case where only admin can see extra specs of a volume type.","commit_id":"9e954601947ae42ee740d7eb9faf0fde54eaa659"}]}
