)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":1736,"name":"Ivan Kolodyazhny","email":"e0ne@e0ne.info","username":"e0ne"},"change_message_id":"3c5cc92e62dde737ae117901d93be62c47a23209","unresolved":false,"context_lines":[{"line_number":12,"context_line":"This is one of the aspects that are needed to support Active/Active in"},{"line_number":13,"context_line":"cinder."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"APIImpact"},{"line_number":16,"context_line":"Implements: blueprint cinder-volume-active-active-support"},{"line_number":17,"context_line":"Change-Id: Ieadf7979262142f03d3f337ea4399c54a2d08c52"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":5,"id":"7a740942_89fd383c","line":15,"updated":"2015-12-07 11:52:58.000000000","message":"Please, remove APIImpact flag from the spec commit message in the next patch","commit_id":"33c430cb30646a7cf63d1d13d07571dfe0ed4653"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"758f4e12e205a8c95e51f55fc95dfbc296eb4820","unresolved":false,"context_lines":[{"line_number":12,"context_line":"This is one of the aspects that are needed to support Active/Active in"},{"line_number":13,"context_line":"cinder."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"APIImpact"},{"line_number":16,"context_line":"Implements: blueprint cinder-volume-active-active-support"},{"line_number":17,"context_line":"Change-Id: Ieadf7979262142f03d3f337ea4399c54a2d08c52"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":5,"id":"7a740942_211e3f0c","line":15,"in_reply_to":"7a740942_89fd383c","updated":"2015-12-14 12:30:38.000000000","message":"Done","commit_id":"33c430cb30646a7cf63d1d13d07571dfe0ed4653"}],"specs/mitaka/ha-aa-job-distribution.rst":[{"author":{"_account_id":177,"name":"Alex Meade","email":"mr.alex.meade@gmail.com","username":"alex-meade"},"change_message_id":"99bf6d05c921756b7ba4df6734fad8b9d4296378","unresolved":false,"context_lines":[{"line_number":29,"context_line":"any of those nodes."},{"line_number":30,"context_line":""},{"line_number":31,"context_line":"Jobs are currently being distributed using a topic based message queue that is"},{"line_number":32,"context_line":"identified by the topic `host@gackend` and that\u0027s the mechanism used to send"},{"line_number":33,"context_line":"jobs to the Volume nodes regardless of the physical address of the node that is"},{"line_number":34,"context_line":"going to be handling the job, allowing an easier transition on failover."},{"line_number":35,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a2fa921_6af7588b","line":32,"updated":"2015-10-08 17:32:23.000000000","message":"s/gackend/backend","commit_id":"cb90c7b4b288f770e74d9769ef8c9261794798ff"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"8f950c0c9607ece99acb6cc289f721fcf2b5dea0","unresolved":false,"context_lines":[{"line_number":29,"context_line":"any of those nodes."},{"line_number":30,"context_line":""},{"line_number":31,"context_line":"Jobs are currently being distributed using a topic based message queue that is"},{"line_number":32,"context_line":"identified by the topic `host@gackend` and that\u0027s the mechanism used to send"},{"line_number":33,"context_line":"jobs to the Volume nodes regardless of the physical address of the node that is"},{"line_number":34,"context_line":"going to be handling the job, allowing an easier transition on failover."},{"line_number":35,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a2fa921_3f5addf2","line":32,"in_reply_to":"7a2fa921_6af7588b","updated":"2015-10-09 12:01:45.000000000","message":"Thanks.","commit_id":"cb90c7b4b288f770e74d9769ef8c9261794798ff"},{"author":{"_account_id":177,"name":"Alex Meade","email":"mr.alex.meade@gmail.com","username":"alex-meade"},"change_message_id":"99bf6d05c921756b7ba4df6734fad8b9d4296378","unresolved":false,"context_lines":[{"line_number":153,"context_line":""},{"line_number":154,"context_line":"- It will help us on job cleanup of failed nodes that do not come back up."},{"line_number":155,"context_line":"  Although cleanup is not part of this spec, it is good to keep it in mind and"},{"line_number":156,"context_line":"  facilitate it as much as possible."},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Data model impact"},{"line_number":159,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a2fa921_ca4b0c86","line":156,"updated":"2015-10-08 17:32:23.000000000","message":"Another alternative is not having the \u0027host\u0027 field at all, just let the backend name be the important part. Who cares if it goes to host1 or host2, they both manage the same backend. If I spin up a host3 to manage this backend then it would just start pulling requests for backend off of the queue with no changes/config needed.","commit_id":"cb90c7b4b288f770e74d9769ef8c9261794798ff"},{"author":{"_account_id":177,"name":"Alex Meade","email":"mr.alex.meade@gmail.com","username":"alex-meade"},"change_message_id":"bfe258ef32dbfb83a15718eb7219f8aa4c418064","unresolved":false,"context_lines":[{"line_number":153,"context_line":""},{"line_number":154,"context_line":"- It will help us on job cleanup of failed nodes that do not come back up."},{"line_number":155,"context_line":"  Although cleanup is not part of this spec, it is good to keep it in mind and"},{"line_number":156,"context_line":"  facilitate it as much as possible."},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Data model impact"},{"line_number":159,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a2fa921_9971897a","line":156,"in_reply_to":"7a2fa921_9fc77163","updated":"2015-10-12 14:06:54.000000000","message":"mm I see","commit_id":"cb90c7b4b288f770e74d9769ef8c9261794798ff"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"8f950c0c9607ece99acb6cc289f721fcf2b5dea0","unresolved":false,"context_lines":[{"line_number":153,"context_line":""},{"line_number":154,"context_line":"- It will help us on job cleanup of failed nodes that do not come back up."},{"line_number":155,"context_line":"  Although cleanup is not part of this spec, it is good to keep it in mind and"},{"line_number":156,"context_line":"  facilitate it as much as possible."},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Data model impact"},{"line_number":159,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a2fa921_9fc77163","line":156,"in_reply_to":"7a2fa921_ca4b0c86","updated":"2015-10-09 12:01:45.000000000","message":"Removing host configuration parameter would break backward compatibility, knowing the status of each node in the cluster is useful for operators and how do you expect to do proper cleanup if you don\u0027t know which host is working on the resources?","commit_id":"cb90c7b4b288f770e74d9769ef8c9261794798ff"},{"author":{"_account_id":177,"name":"Alex Meade","email":"mr.alex.meade@gmail.com","username":"alex-meade"},"change_message_id":"99bf6d05c921756b7ba4df6734fad8b9d4296378","unresolved":false,"context_lines":[{"line_number":189,"context_line":""},{"line_number":190,"context_line":"When ``cluster_detail`` is set to False, which tells us client knows of the new"},{"line_number":191,"context_line":"API options, we will be returning key ``cluster`` instead of key ``host`` and"},{"line_number":192,"context_line":"we\u0027ll return new fields:"},{"line_number":193,"context_line":""},{"line_number":194,"context_line":"- ``nodes``: Number of nodes that form the cluster."},{"line_number":195,"context_line":"- ``down_nodes``: Number of nodes that are down in the cluster."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a2fa921_4a0e1c3c","line":192,"updated":"2015-10-08 17:32:23.000000000","message":"Why have cluster_detail? I don\u0027t think there is any harm in just returning new fields + host all the time as they could be ignored. If there is need to know if the client understands the new fields, that should probably be done with a microversion","commit_id":"cb90c7b4b288f770e74d9769ef8c9261794798ff"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"8f950c0c9607ece99acb6cc289f721fcf2b5dea0","unresolved":false,"context_lines":[{"line_number":189,"context_line":""},{"line_number":190,"context_line":"When ``cluster_detail`` is set to False, which tells us client knows of the new"},{"line_number":191,"context_line":"API options, we will be returning key ``cluster`` instead of key ``host`` and"},{"line_number":192,"context_line":"we\u0027ll return new fields:"},{"line_number":193,"context_line":""},{"line_number":194,"context_line":"- ``nodes``: Number of nodes that form the cluster."},{"line_number":195,"context_line":"- ``down_nodes``: Number of nodes that are down in the cluster."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a2fa921_9f769104","line":192,"in_reply_to":"7a2fa921_4a0e1c3c","updated":"2015-10-09 12:01:45.000000000","message":"I thought of doing it that way but discarded it, the reason being that sent data needs to make sense.  If we send old data plus new data we would end up with:\n\nhost -\u003e Which is not semantically correct since this is the cluster but is needed to be backward compatible\ncluster -\u003e Which is repeated as has the same value as host, so why send it at all? But if we don\u0027t send it the data makes no sense.\nhosts -\u003e Is OK\n\nAlso, if the deployment is big you may not want to retrieve all the information and just a summary (this reason is lees relevant).","commit_id":"cb90c7b4b288f770e74d9769ef8c9261794798ff"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"f8ec1cc2b35f76fc51e0f81280b44a5bd3279fef","unresolved":false,"context_lines":[{"line_number":62,"context_line":"identify a group of Volume nodes that share the same storage backends and"},{"line_number":63,"context_line":"therefore can accept jobs for the same volumes."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"This new configuration option, unlike `host` option, can have the same value on"},{"line_number":66,"context_line":"multiple volume nodes, with the only requisite that all nodes that share the"},{"line_number":67,"context_line":"same value must also share the same storage backends."},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_f6bf805e","line":65,"updated":"2015-10-11 16:16:36.000000000","message":"The \u0027host\u0027 field absolutely can be set to the same value on multiple machines, and messages get round-robinned appropriately. There are some annoying side effects (the host setting is global, so it affects c-sch, c-bak etc, it makes a mess of cinder service-list output, and others) but it does work.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"d2955639a50e0141543c8710e8e3393a8e6e767f","unresolved":false,"context_lines":[{"line_number":62,"context_line":"identify a group of Volume nodes that share the same storage backends and"},{"line_number":63,"context_line":"therefore can accept jobs for the same volumes."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"This new configuration option, unlike `host` option, can have the same value on"},{"line_number":66,"context_line":"multiple volume nodes, with the only requisite that all nodes that share the"},{"line_number":67,"context_line":"same value must also share the same storage backends."},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a2ca52d_9daf35e3","line":65,"in_reply_to":"7a2fa921_838ee825","updated":"2015-10-13 16:30:51.000000000","message":"Like Michal said, we need different host fields to do the cleanup.  I may need to explain it better in the specs though.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a6b20f1298a960822945f6192f9fe2ec3be2affa","unresolved":false,"context_lines":[{"line_number":62,"context_line":"identify a group of Volume nodes that share the same storage backends and"},{"line_number":63,"context_line":"therefore can accept jobs for the same volumes."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"This new configuration option, unlike `host` option, can have the same value on"},{"line_number":66,"context_line":"multiple volume nodes, with the only requisite that all nodes that share the"},{"line_number":67,"context_line":"same value must also share the same storage backends."},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_838ee825","line":65,"in_reply_to":"7a2fa921_f6bf805e","updated":"2015-10-12 19:37:36.000000000","message":"https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L390-L428\n\nWith multiple c-vols with same host set volume can be errored out even if it is processed by another service. Similar situation happens with volumes in deleting state: https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L435-L447\n\nI know that Fuel 6 by default sets c-vol in A/A, but I wouldn\u0027t consider that safe.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a6b20f1298a960822945f6192f9fe2ec3be2affa","unresolved":false,"context_lines":[{"line_number":115,"context_line":"keep them updated with their stats, that way they can make informed decisions"},{"line_number":116,"context_line":"on where to perform operations."},{"line_number":117,"context_line":""},{"line_number":118,"context_line":"In a similar way to the Service state reporting we need to prevent concurrent"},{"line_number":119,"context_line":"access to the data structure when updating this information. Fortunately for us"},{"line_number":120,"context_line":"we are storing this information in a Python dictionary on the schedulers, and"},{"line_number":121,"context_line":"since we are using an eventlet executor for the rpc server we don’t have to"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_4e81d7fa","line":118,"updated":"2015-10-12 19:37:36.000000000","message":"When receiving capabilities update we\u0027re using just d[k] \u003d v operation (https://github.com/openstack/cinder/blob/master/cinder/scheduler/host_manager.py#L434), which is considered to be atomic in CPython. I don\u0027t think we\u0027re supporting other Python implementations, so whole paragraph makes sense to me.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a6b20f1298a960822945f6192f9fe2ec3be2affa","unresolved":false,"context_lines":[{"line_number":118,"context_line":"In a similar way to the Service state reporting we need to prevent concurrent"},{"line_number":119,"context_line":"access to the data structure when updating this information. Fortunately for us"},{"line_number":120,"context_line":"we are storing this information in a Python dictionary on the schedulers, and"},{"line_number":121,"context_line":"since we are using an eventlet executor for the rpc server we don’t have to"},{"line_number":122,"context_line":"worry about using locks, the inherent behavior of the executor will prevent"},{"line_number":123,"context_line":"concurrent access to the dictionary.  So no changes are needed there."},{"line_number":124,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_e32b1c93","line":121,"updated":"2015-10-12 19:37:36.000000000","message":"s/rpc/RPC","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a6b20f1298a960822945f6192f9fe2ec3be2affa","unresolved":false,"context_lines":[{"line_number":175,"context_line":"- ``host``: Primary key. Same meaning as it holds now"},{"line_number":176,"context_line":"- ``report_count``: Same meaning as it holds now"},{"line_number":177,"context_line":""},{"line_number":178,"context_line":""},{"line_number":179,"context_line":"REST API impact"},{"line_number":180,"context_line":"---------------"},{"line_number":181,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_8e304f09","line":178,"updated":"2015-10-12 19:37:36.000000000","message":"Some of this migrations are \"contracting\" and would break rolling upgrades in current state of Cinder. I haven\u0027t had communicated that yet, but I think that we should start caring about that - so I would love to see it explained in this context. Feel free to ask me for help on the topic. :)\n\nOnline schema migrations (expand-contract migrations) would solve that, but current state of that functionality in Nova is considered to be experimental and dangerous.\n\nGood insight on the topic (no, we cannot use versioned objects exactly like Nova - we don\u0027t have a conductor service): http://www.danplanet.com/blog/2015/10/07/upgrades-in-nova-database-migrations/\n\nP.S. Cinder haven\u0027t had introduced a contracting migration since Juno, so maybe it\u0027s possible to avoid them also in this spec. :)","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"d2955639a50e0141543c8710e8e3393a8e6e767f","unresolved":false,"context_lines":[{"line_number":175,"context_line":"- ``host``: Primary key. Same meaning as it holds now"},{"line_number":176,"context_line":"- ``report_count``: Same meaning as it holds now"},{"line_number":177,"context_line":""},{"line_number":178,"context_line":""},{"line_number":179,"context_line":"REST API impact"},{"line_number":180,"context_line":"---------------"},{"line_number":181,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a2ca52d_18ffb302","line":178,"in_reply_to":"7a2fa921_8e304f09","updated":"2015-10-13 16:30:51.000000000","message":"I remember discussing with you how rolling upgrades will not work as we have everything now and the missing conductor service.  And until we have an specification that explains how we intend to do it I think discussing it in this spec would just make it even more confusing.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a6b20f1298a960822945f6192f9fe2ec3be2affa","unresolved":false,"context_lines":[{"line_number":182,"context_line":"To report status of services we will add an optional parameter called"},{"line_number":183,"context_line":"``cluster_detail`` which will report the node breakdown."},{"line_number":184,"context_line":""},{"line_number":185,"context_line":"When ``cluster_detail`` is not set we will report exactly as we are doing now,"},{"line_number":186,"context_line":"to be backward compatible with clients, and we\u0027ll return ``cluster`` field in"},{"line_number":187,"context_line":"the place of ``host`` and in ``state`` we\u0027ll send if *any* of the hosts forming"},{"line_number":188,"context_line":"the cluster are up."}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_a9d42547","line":185,"updated":"2015-10-12 19:37:36.000000000","message":"I\u0027m not sure, so just to ask more experienced reviewers - is it a good idea to differentiate between \"not-set\", \"false\", \"true\" in the REST API? Does this fit into microversion (maybe is, but probably we shouldn\u0027t abuse that concept?).","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"d2955639a50e0141543c8710e8e3393a8e6e767f","unresolved":false,"context_lines":[{"line_number":182,"context_line":"To report status of services we will add an optional parameter called"},{"line_number":183,"context_line":"``cluster_detail`` which will report the node breakdown."},{"line_number":184,"context_line":""},{"line_number":185,"context_line":"When ``cluster_detail`` is not set we will report exactly as we are doing now,"},{"line_number":186,"context_line":"to be backward compatible with clients, and we\u0027ll return ``cluster`` field in"},{"line_number":187,"context_line":"the place of ``host`` and in ``state`` we\u0027ll send if *any* of the hosts forming"},{"line_number":188,"context_line":"the cluster are up."}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a2ca52d_9872e37b","line":185,"in_reply_to":"7a2fa921_a9d42547","updated":"2015-10-13 16:30:51.000000000","message":"This is just an option, we could also change this to have 2 options: `cluster_mode` and `detail` instead of a tristate `detail`.\n\nOr just have one (`cluster_mode`) and do like Alex mentioned and always return all details when it is set to True.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a6b20f1298a960822945f6192f9fe2ec3be2affa","unresolved":false,"context_lines":[{"line_number":198,"context_line":"will contain a list of dictionaries with ``state`` and ``host`` keys that will"},{"line_number":199,"context_line":"contain individual values for each of the cluster hosts."},{"line_number":200,"context_line":""},{"line_number":201,"context_line":"On the update we should allow to receive ``host`` or ``cluster`` and they will"},{"line_number":202,"context_line":"be treated in the same way."},{"line_number":203,"context_line":""},{"line_number":204,"context_line":"Security impact"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_a99f45b3","line":201,"updated":"2015-10-12 19:37:36.000000000","message":"So we\u0027re disabling whole cluster when disabling a service? I guess that when sending disable with individual host value, we\u0027re failing? That probably should be clarified here.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"d2955639a50e0141543c8710e8e3393a8e6e767f","unresolved":false,"context_lines":[{"line_number":198,"context_line":"will contain a list of dictionaries with ``state`` and ``host`` keys that will"},{"line_number":199,"context_line":"contain individual values for each of the cluster hosts."},{"line_number":200,"context_line":""},{"line_number":201,"context_line":"On the update we should allow to receive ``host`` or ``cluster`` and they will"},{"line_number":202,"context_line":"be treated in the same way."},{"line_number":203,"context_line":""},{"line_number":204,"context_line":"Security impact"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a2ca52d_38314fa3","line":201,"in_reply_to":"7a2fa921_a99f45b3","updated":"2015-10-13 16:30:51.000000000","message":"Yes, me and my assumption that everyone can read my thoughts and know what I want to do.  ;-)","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a6b20f1298a960822945f6192f9fe2ec3be2affa","unresolved":false,"context_lines":[{"line_number":250,"context_line":"  Gorka Eguileor (geguileo)"},{"line_number":251,"context_line":""},{"line_number":252,"context_line":"Other contributors:"},{"line_number":253,"context_line":"  Anyone is welcome to help"},{"line_number":254,"context_line":""},{"line_number":255,"context_line":"Work Items"},{"line_number":256,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a2fa921_695ccded","line":253,"updated":"2015-10-12 19:37:36.000000000","message":"I\u0027m still considering other approach to the problem, but if we\u0027ll decide to go that way - count me in.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"d2955639a50e0141543c8710e8e3393a8e6e767f","unresolved":false,"context_lines":[{"line_number":250,"context_line":"  Gorka Eguileor (geguileo)"},{"line_number":251,"context_line":""},{"line_number":252,"context_line":"Other contributors:"},{"line_number":253,"context_line":"  Anyone is welcome to help"},{"line_number":254,"context_line":""},{"line_number":255,"context_line":"Work Items"},{"line_number":256,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a2ca52d_b81d5f11","line":253,"in_reply_to":"7a2fa921_695ccded","updated":"2015-10-13 16:30:51.000000000","message":"Awesome!!  On both ends, other solutions and offering to help.","commit_id":"d2f71b92ffe012e2abdbf685ac6611e2f349113e"},{"author":{"_account_id":5997,"name":"Walt","display_name":"Hemna","email":"waboring@hemna.com","username":"walter-boring","status":"SAP"},"change_message_id":"eb7126b1fb86b0750971da9c171cbd07505798d4","unresolved":false,"context_lines":[{"line_number":96,"context_line":""},{"line_number":97,"context_line":"On Active/Active configurations it is down if there is no valid heartbeat from"},{"line_number":98,"context_line":"any of the hosts that constitute the cluster, and it is up is there is at least"},{"line_number":99,"context_line":"one valid heartbeat."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Since we are moving from a 1 to 1 relationship between hosts and services to a"},{"line_number":102,"context_line":"n to 1 relationship we will need to make some changes to the DB services table"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3a29b11f_f9405aa7","line":99,"updated":"2015-10-20 18:44:46.000000000","message":"We already have something in the respect of a heartbeat.   Every c-vol service reports period stats updates that happens every 60 seconds.   How are we going to avoid getting multiple reports back from every c-vol service that is the same backend?","commit_id":"fabb9cf7ebc676d402138188b73a8e7bec2202e5"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"2d094d4c2ef1f5ba7f0bd36cc6274dcc00ee137d","unresolved":false,"context_lines":[{"line_number":96,"context_line":""},{"line_number":97,"context_line":"On Active/Active configurations it is down if there is no valid heartbeat from"},{"line_number":98,"context_line":"any of the hosts that constitute the cluster, and it is up is there is at least"},{"line_number":99,"context_line":"one valid heartbeat."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Since we are moving from a 1 to 1 relationship between hosts and services to a"},{"line_number":102,"context_line":"n to 1 relationship we will need to make some changes to the DB services table"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3a29b11f_6f89dc25","line":99,"in_reply_to":"3a29b11f_4f80a046","updated":"2015-10-20 19:18:54.000000000","message":"Okay, this may get more interesting with capacity handling. Do all of the c-vols have the same view of the capacity? Each of them seems to keep a local copy of data [1]. I\u0027m not sure if we can consider that this data is synced between c-vols.\n\n[1] https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L1858","commit_id":"fabb9cf7ebc676d402138188b73a8e7bec2202e5"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"0e2e59ebeb7781196fc532fb224764d62abf77c0","unresolved":false,"context_lines":[{"line_number":96,"context_line":""},{"line_number":97,"context_line":"On Active/Active configurations it is down if there is no valid heartbeat from"},{"line_number":98,"context_line":"any of the hosts that constitute the cluster, and it is up is there is at least"},{"line_number":99,"context_line":"one valid heartbeat."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Since we are moving from a 1 to 1 relationship between hosts and services to a"},{"line_number":102,"context_line":"n to 1 relationship we will need to make some changes to the DB services table"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3a29b11f_e7d96f23","line":99,"in_reply_to":"3a29b11f_6f89dc25","updated":"2015-10-21 13:34:50.000000000","message":"Like dulek said, we are talking about DB heartbeats.  Capacity does not need to be synchronize between volume nodes, just between schedulers, which is not assured as code is now.  I\u0027ll add a line on this in the spec.","commit_id":"fabb9cf7ebc676d402138188b73a8e7bec2202e5"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"755322bf9884710dbd3261de89972e330a1e745d","unresolved":false,"context_lines":[{"line_number":96,"context_line":""},{"line_number":97,"context_line":"On Active/Active configurations it is down if there is no valid heartbeat from"},{"line_number":98,"context_line":"any of the hosts that constitute the cluster, and it is up is there is at least"},{"line_number":99,"context_line":"one valid heartbeat."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Since we are moving from a 1 to 1 relationship between hosts and services to a"},{"line_number":102,"context_line":"n to 1 relationship we will need to make some changes to the DB services table"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3a29b11f_4f80a046","line":99,"in_reply_to":"3a29b11f_f9405aa7","updated":"2015-10-20 19:10:03.000000000","message":"Here we\u0027re thinking of DB heartbeats - that is bumping up updated_at column in the services table.\n\nTechnically all the cluster members should report same capabilities, so this is not a problem (besides having more messages on the broker). This is how it works now in Fuel deployment, which starts c-vol in A/A.","commit_id":"fabb9cf7ebc676d402138188b73a8e7bec2202e5"},{"author":{"_account_id":7173,"name":"Scott DAngelo","email":"scott.dangelo@gmail.com","username":"scottda"},"change_message_id":"ff05cf5581881f102e8c9dd096fbc60c678b48e6","unresolved":false,"context_lines":[{"line_number":257,"context_line":"Primary assignee:"},{"line_number":258,"context_line":"  Gorka Eguileor (geguileo)"},{"line_number":259,"context_line":""},{"line_number":260,"context_line":"Other contributors:"},{"line_number":261,"context_line":"  Michal Dulko (dulek)"},{"line_number":262,"context_line":"  Anyone is welcome to help"},{"line_number":263,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a2ca52d_56309756","line":260,"updated":"2015-10-13 18:50:40.000000000","message":"I\u0027d like to help with this as well.","commit_id":"fabb9cf7ebc676d402138188b73a8e7bec2202e5"},{"author":{"_account_id":11904,"name":"Sean McGinnis","email":"sean.mcginnis@gmail.com","username":"SeanM"},"change_message_id":"d68d3d5ef91b04eb9eb62048529be20269d25204","unresolved":false,"context_lines":[{"line_number":96,"context_line":"heartbeats are reported in the DB in ``services`` table."},{"line_number":97,"context_line":""},{"line_number":98,"context_line":"On Active/Active configurations it is down if there is no valid heartbeat from"},{"line_number":99,"context_line":"any of the hosts that constitute the cluster, and it is up is there is at least"},{"line_number":100,"context_line":"one valid heartbeat."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"Since we are moving from a 1 to 1 relationship between hosts and services to a"}],"source_content_type":"text/x-rst","patch_set":5,"id":"7a740942_5f07bf17","line":99,"updated":"2015-12-13 23:51:59.000000000","message":"s/up is there/up if there/","commit_id":"33c430cb30646a7cf63d1d13d07571dfe0ed4653"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"758f4e12e205a8c95e51f55fc95dfbc296eb4820","unresolved":false,"context_lines":[{"line_number":96,"context_line":"heartbeats are reported in the DB in ``services`` table."},{"line_number":97,"context_line":""},{"line_number":98,"context_line":"On Active/Active configurations it is down if there is no valid heartbeat from"},{"line_number":99,"context_line":"any of the hosts that constitute the cluster, and it is up is there is at least"},{"line_number":100,"context_line":"one valid heartbeat."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"Since we are moving from a 1 to 1 relationship between hosts and services to a"}],"source_content_type":"text/x-rst","patch_set":5,"id":"7a740942_01230357","line":99,"in_reply_to":"7a740942_5f07bf17","updated":"2015-12-14 12:30:38.000000000","message":"Done","commit_id":"33c430cb30646a7cf63d1d13d07571dfe0ed4653"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"7fdd98ef7913bbc96c02c1d14f8b3f85fe3ca8de","unresolved":false,"context_lines":[{"line_number":148,"context_line":"when each node processes them we can have greenthreads execution in different"},{"line_number":149,"context_line":"order on different nodes thus ending up with different data on each node."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":"Since this is a very rare case we can ignore it, since discrepancies between"},{"line_number":152,"context_line":"schedulers are almost impossible and even when if they happen will be minimum."},{"line_number":153,"context_line":"On `Alternatives`_ there is a proposal on how to have this fixed."},{"line_number":154,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa80f949_8a810a92","line":151,"updated":"2015-11-03 18:38:31.000000000","message":"Well… We don\u0027t know if it\u0027s a rare case. See what\u0027s Nova\u0027s dealing with in [1]. But assuming we\u0027re borrowing the architecture from Nova and they doesn\u0027t have a clear solution to the problem, I think it\u0027s fair to rely on eventual consistency and rescheduling.\n\nBTW - this is a great explanation of the general issue.\n\n[1] http://lists.openstack.org/pipermail/openstack-dev/2015-October/076373.html","commit_id":"33c430cb30646a7cf63d1d13d07571dfe0ed4653"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"1401292533705afeb4626d9c3d424f18d61c6b9c","unresolved":false,"context_lines":[{"line_number":148,"context_line":"when each node processes them we can have greenthreads execution in different"},{"line_number":149,"context_line":"order on different nodes thus ending up with different data on each node."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":"Since this is a very rare case we can ignore it, since discrepancies between"},{"line_number":152,"context_line":"schedulers are almost impossible and even when if they happen will be minimum."},{"line_number":153,"context_line":"On `Alternatives`_ there is a proposal on how to have this fixed."},{"line_number":154,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"ba8a016a_980cb93c","line":151,"in_reply_to":"ba8a016a_3264e520","updated":"2015-11-23 17:27:34.000000000","message":"Timestamping would need to be done by storage backend itself, as it is the source of truth on capabilities.\n\nWe should probably test it, but in general it seems we\u0027ll achieve have eventual consistency here, so I\u0027m fine.","commit_id":"33c430cb30646a7cf63d1d13d07571dfe0ed4653"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"aca4c40872caa564481a24d856b0542748aec152","unresolved":false,"context_lines":[{"line_number":148,"context_line":"when each node processes them we can have greenthreads execution in different"},{"line_number":149,"context_line":"order on different nodes thus ending up with different data on each node."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":"Since this is a very rare case we can ignore it, since discrepancies between"},{"line_number":152,"context_line":"schedulers are almost impossible and even when if they happen will be minimum."},{"line_number":153,"context_line":"On `Alternatives`_ there is a proposal on how to have this fixed."},{"line_number":154,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"ba8a016a_3264e520","line":151,"in_reply_to":"fa80f949_8a810a92","updated":"2015-11-17 12:50:41.000000000","message":"Thanks for the info, I have read it and while I think it is related in the sense that both are talking about Scheduler problems, but it is not what I\u0027m talking about here and the part that is related to Cinder problems can be solved with conditional updates.\n\nWhat I\u0027m talking about here is 2 schedulers having different capabilities for a cluster at a given time.  But we have a very easy solution for this as explained in L188-L192.","commit_id":"33c430cb30646a7cf63d1d13d07571dfe0ed4653"},{"author":{"_account_id":6491,"name":"xing-yang","email":"xingyang105@gmail.com","username":"xing-yang"},"change_message_id":"3d57959bf775563617601fbac290e791ee04e04e","unresolved":false,"context_lines":[{"line_number":206,"context_line":"- Change ``host`` field for cluster field"},{"line_number":207,"context_line":"- Removal of ``report_count`` field"},{"line_number":208,"context_line":"- Removal of ``modified_at`` field since we can go back to using ``updated_at``"},{"line_number":209,"context_line":"  field now that heartbeats will be reported on another table."},{"line_number":210,"context_line":""},{"line_number":211,"context_line":"New ``service_heartbeats`` table will have following fields:"},{"line_number":212,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"1a7b0d38_c66dac05","line":209,"updated":"2015-12-23 23:15:07.000000000","message":"How does this affect rolling upgrade?  Removal of columns cannot be done in a single release based on the online-schema-upgrades spec:\n\nhttps://review.openstack.org/#/c/245976/","commit_id":"80db6b585dc37acd25b4bcb05de64a0ad2328bcc"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"adad4d17884602d8a4f972306c9ee99a0717afb0","unresolved":false,"context_lines":[{"line_number":206,"context_line":"- Change ``host`` field for cluster field"},{"line_number":207,"context_line":"- Removal of ``report_count`` field"},{"line_number":208,"context_line":"- Removal of ``modified_at`` field since we can go back to using ``updated_at``"},{"line_number":209,"context_line":"  field now that heartbeats will be reported on another table."},{"line_number":210,"context_line":""},{"line_number":211,"context_line":"New ``service_heartbeats`` table will have following fields:"},{"line_number":212,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"da6ed579_0a251caa","line":209,"in_reply_to":"1a7b0d38_c66dac05","updated":"2016-01-12 12:00:14.000000000","message":"We can start using updated_at in this release, and ignore modified_at field, and remove the field in the next release.","commit_id":"80db6b585dc37acd25b4bcb05de64a0ad2328bcc"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"fff67f00a9823c5e25bbdd8ebbe9945b884823a9","unresolved":false,"context_lines":[{"line_number":206,"context_line":"- Change ``host`` field for cluster field"},{"line_number":207,"context_line":"- Removal of ``report_count`` field"},{"line_number":208,"context_line":"- Removal of ``modified_at`` field since we can go back to using ``updated_at``"},{"line_number":209,"context_line":"  field now that heartbeats will be reported on another table."},{"line_number":210,"context_line":""},{"line_number":211,"context_line":"New ``service_heartbeats`` table will have following fields:"},{"line_number":212,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"da6ed579_6eddc713","line":209,"in_reply_to":"da6ed579_0a251caa","updated":"2016-01-12 13:50:44.000000000","message":"Yay, complicated programming puzzles!\n\nSome assumptions first:\n* The order of upgrades is considered to be c-api-\u003ec-sch-\u003ec-vol. c-bak can be upgraded just after c-api.\n* We want Liberty to work with Mitaka\u0027s DB schema.\n* Good news is that `host` column is never modified in Cinder. If it changes, a service will create a new DB entry. This is *very* important in my further reasoning.\n\nWe certainly don\u0027t want to drop any columns in Mitaka as we need to make sure that Liberty is able to work with M\u0027s DB schema. Same thing for renaming \"host\" field. We can and should of course add a new \"cluster\" field.\n\nservice-enable, service-disable and service-list actions are just hitting the DB from c-api. We would need to make sure they\u0027re able to read from `host` column if `cluster` column is empty. This is easy to do with o.vo layer. Please note that we\u0027re able to do that only because `host` column cannot be modified.\n\nc-sch still uses only `host` or `cluster` field to schedule volumes, so same principle applies here - we just need to read from `host` if `cluster` is empty.\n\nI understand that `updated_at` is useless now for heartbeating and when upgrading c-vol\u0027s we should have all c-sch and c-api in Mitaka versions, so they can understand new c-vol reporting format. That means that new versions of services can report only in service_heartbeats table.\n\nBut we need to make sure that M\u0027s c-sch and c-api can understand how L services heartbeat. This seems trivial - when deciding if service is up check service_heartbeats and if it returns false fallback and check updated_at in services table. DB impact can be limited by joining two queries.\n\nDoes this make sense to you? I understand that this is extremely complicated so I offer my help in getting this right.","commit_id":"80db6b585dc37acd25b4bcb05de64a0ad2328bcc"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"7aa22ba50eadb796587fb84df371718ea4c8af66","unresolved":false,"context_lines":[{"line_number":206,"context_line":"- Change ``host`` field for cluster field"},{"line_number":207,"context_line":"- Removal of ``report_count`` field"},{"line_number":208,"context_line":"- Removal of ``modified_at`` field since we can go back to using ``updated_at``"},{"line_number":209,"context_line":"  field now that heartbeats will be reported on another table."},{"line_number":210,"context_line":""},{"line_number":211,"context_line":"New ``service_heartbeats`` table will have following fields:"},{"line_number":212,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"da6ed579_df585a8c","line":209,"in_reply_to":"da6ed579_33140ff4","updated":"2016-01-15 13:50:06.000000000","message":"Your explanation was great, you even made it look easy  :-)\n\nHost field can be modified during migration and we don\u0027t create a new row in the DB.\n\nWe can\u0027t write only to cluster, because for example when we are doing the upgrades of the schedulers we could have an M scheduler create a volume (sets cluster and leaves host empty) and then sends the request to the c-vol nodes that are still in L and don\u0027t know anything about cluster field and are using host in the creation flow.\n\nA more complicated problem is if we do a migration/retype with migration in the same situation, because L c-vol node will change the host field but will leave the cluster field of the old node, so we\u0027ll end up with a discrepancy.","commit_id":"80db6b585dc37acd25b4bcb05de64a0ad2328bcc"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"26d9400d7ff47cc77a6e23b82d73d65649ddf637","unresolved":false,"context_lines":[{"line_number":206,"context_line":"- Change ``host`` field for cluster field"},{"line_number":207,"context_line":"- Removal of ``report_count`` field"},{"line_number":208,"context_line":"- Removal of ``modified_at`` field since we can go back to using ``updated_at``"},{"line_number":209,"context_line":"  field now that heartbeats will be reported on another table."},{"line_number":210,"context_line":""},{"line_number":211,"context_line":"New ``service_heartbeats`` table will have following fields:"},{"line_number":212,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"da6ed579_33140ff4","line":209,"in_reply_to":"da6ed579_6eddc713","updated":"2016-01-12 16:16:37.000000000","message":"Ah, one more thing to clarify. I think M\u0027s services don\u0027t need to write to `host` column and need only write to `cluster` one in our case. This is because of the upgrade order.\n\nOf course it won\u0027t hurt if they\u0027ll do both just in case. It should be easy to implement in o.vo layer.","commit_id":"80db6b585dc37acd25b4bcb05de64a0ad2328bcc"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"3661f924a7babbcd766a2f60198681fbe185a678","unresolved":false,"context_lines":[{"line_number":227,"context_line":""},{"line_number":228,"context_line":"- ``consistencygroups``"},{"line_number":229,"context_line":"- ``volumes``"},{"line_number":230,"context_line":"- ``backups``"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"There are more fields called ``host`` in cinder model, but on those case they"},{"line_number":233,"context_line":"don\u0027t need to be changed:"}],"source_content_type":"text/x-rst","patch_set":7,"id":"ba63e191_0ff30675","line":230,"range":{"start_line":230,"start_character":0,"end_line":230,"end_character":13},"updated":"2016-01-18 19:35:49.000000000","message":"I\u0027m not sure if this will be required once [1] get merged? It may be handy, but I think in that model we should run multiple instances of c-bak in same cluster.\n\n[1] https://review.openstack.org/#/c/262395/","commit_id":"6a995b5b34a708b87a1994edd2d9d674be4b5964"},{"author":{"_account_id":15961,"name":"lisali","email":"xiaoyan.li@intel.com","username":"lisali"},"change_message_id":"dc7b4eda0c2b32e91f3811b8611e90b66f2bf0dd","unresolved":false,"context_lines":[{"line_number":227,"context_line":""},{"line_number":228,"context_line":"- ``consistencygroups``"},{"line_number":229,"context_line":"- ``volumes``"},{"line_number":230,"context_line":"- ``backups``"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"There are more fields called ``host`` in cinder model, but on those case they"},{"line_number":233,"context_line":"don\u0027t need to be changed:"}],"source_content_type":"text/x-rst","patch_set":7,"id":"9a68dd71_db1a987c","line":230,"range":{"start_line":230,"start_character":0,"end_line":230,"end_character":13},"in_reply_to":"9a68dd71_30ae97e6","updated":"2016-01-26 05:27:23.000000000","message":"In scaling backup service, c-bak service can allow multiple backup backends. The differences are that c-bak can be active-active if supporting only 1 storage backends, but it is not active-active if multiple storage backends. \n\nPlease look into _get_available_backup_service_host() in https://review.openstack.org/#/c/262395/13/cinder/backup/api.py","commit_id":"6a995b5b34a708b87a1994edd2d9d674be4b5964"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"be363a2d15cb9f6b548095be174538d6ada7283d","unresolved":false,"context_lines":[{"line_number":227,"context_line":""},{"line_number":228,"context_line":"- ``consistencygroups``"},{"line_number":229,"context_line":"- ``volumes``"},{"line_number":230,"context_line":"- ``backups``"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"There are more fields called ``host`` in cinder model, but on those case they"},{"line_number":233,"context_line":"don\u0027t need to be changed:"}],"source_content_type":"text/x-rst","patch_set":7,"id":"9a68dd71_30ae97e6","line":230,"range":{"start_line":230,"start_character":0,"end_line":230,"end_character":13},"in_reply_to":"9a68dd71_bed3485f","updated":"2016-01-25 13:17:43.000000000","message":"I have new specs locally that I need to submit which completely change the approach, and I\u0027ll try to have a look at the cinder backup specs and patch, but from what I\u0027m hearing 1) c-back will only allow 1 storage backend in *ALL* c-back nodes in the deployment and 2) c-back will work Active-Active like c-api and c-sch do.  And in my opinion that is not the right approach...","commit_id":"6a995b5b34a708b87a1994edd2d9d674be4b5964"},{"author":{"_account_id":15961,"name":"lisali","email":"xiaoyan.li@intel.com","username":"lisali"},"change_message_id":"26a06f01d48a9c2d56b79623097c2e8e18bc9deb","unresolved":false,"context_lines":[{"line_number":227,"context_line":""},{"line_number":228,"context_line":"- ``consistencygroups``"},{"line_number":229,"context_line":"- ``volumes``"},{"line_number":230,"context_line":"- ``backups``"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"There are more fields called ``host`` in cinder model, but on those case they"},{"line_number":233,"context_line":"don\u0027t need to be changed:"}],"source_content_type":"text/x-rst","patch_set":7,"id":"9a68dd71_bed3485f","line":230,"range":{"start_line":230,"start_character":0,"end_line":230,"end_character":13},"in_reply_to":"ba63e191_0ff30675","updated":"2016-01-25 07:25:47.000000000","message":"In my opinion, once the above patch is merged, we don\u0027t need to change host to cluster in backups table. \n\nIt may need in future if we support multiple backup backends. For example, some c-bak nodes back data to ceph, some c-bak nodes back data to swift.","commit_id":"6a995b5b34a708b87a1994edd2d9d674be4b5964"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"58a8aef2204f95bddb25a3759a8525dc7b1831f2","unresolved":false,"context_lines":[{"line_number":72,"context_line":"Volume nodes that share the same storage backends and therefore can accept jobs"},{"line_number":73,"context_line":"for the same volumes interchangeably."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":"This means that now `host` can have can have the same value on multiple volume"},{"line_number":76,"context_line":"nodes, with the only requisite that all nodes that share the same value must"},{"line_number":77,"context_line":"also share the same storage backends and configuration."},{"line_number":78,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"5a5ae5dd_3d6935bf","line":75,"range":{"start_line":75,"start_character":27,"end_line":75,"end_character":44},"updated":"2016-02-05 16:00:21.000000000","message":"double \"can have\"","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"98bac9ad3761136786449e315c7674c0c3727658","unresolved":false,"context_lines":[{"line_number":72,"context_line":"Volume nodes that share the same storage backends and therefore can accept jobs"},{"line_number":73,"context_line":"for the same volumes interchangeably."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":"This means that now `host` can have can have the same value on multiple volume"},{"line_number":76,"context_line":"nodes, with the only requisite that all nodes that share the same value must"},{"line_number":77,"context_line":"also share the same storage backends and configuration."},{"line_number":78,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a57f1b5_1a747f33","line":75,"range":{"start_line":75,"start_character":27,"end_line":75,"end_character":44},"in_reply_to":"5a5ae5dd_3d6935bf","updated":"2016-02-10 15:16:50.000000000","message":"But it really can have can have it.  ;-)","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":6804,"name":"bruce-benjamin","email":"bruce.benjamin@jhuapl.edu","username":"bruce-benjamin"},"change_message_id":"1c688da55017e33ec18737e6acef77c81fd66f1e","unresolved":false,"context_lines":[{"line_number":127,"context_line":"---------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"We need to allow individual node disabling in the same manner that we now allow"},{"line_number":130,"context_line":"disabling just one node."},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"By adding this individual disabling possibility to the group disabling option"},{"line_number":133,"context_line":"we will now have multiple ways of disabling a service, by disabling it as a"}],"source_content_type":"text/x-rst","patch_set":8,"id":"5a5ae5dd_e7de2d10","line":130,"range":{"start_line":130,"start_character":23,"end_line":130,"end_character":24},"updated":"2016-02-05 22:20:49.000000000","message":"I think the following update is needed:\ns/just one node/the whole cluster/","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"98bac9ad3761136786449e315c7674c0c3727658","unresolved":false,"context_lines":[{"line_number":127,"context_line":"---------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"We need to allow individual node disabling in the same manner that we now allow"},{"line_number":130,"context_line":"disabling just one node."},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"By adding this individual disabling possibility to the group disabling option"},{"line_number":133,"context_line":"we will now have multiple ways of disabling a service, by disabling it as a"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a57f1b5_b4953ae5","line":130,"range":{"start_line":130,"start_character":23,"end_line":130,"end_character":24},"in_reply_to":"5a5ae5dd_e7de2d10","updated":"2016-02-10 15:16:50.000000000","message":"Maybe the phrasing was not as clear as it should have been, but the sentence was correct.\n\nAnyway, next patch changes all these paragraphs, I hope it is more clear there.","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"58a8aef2204f95bddb25a3759a8525dc7b1831f2","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Disabling"},{"line_number":127,"context_line":"---------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"We need to allow individual node disabling in the same manner that we now allow"},{"line_number":130,"context_line":"disabling just one node."},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"By adding this individual disabling possibility to the group disabling option"},{"line_number":133,"context_line":"we will now have multiple ways of disabling a service, by disabling it as a"},{"line_number":134,"context_line":"whole or by disabling all its individual nodes."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"Capabilities"},{"line_number":137,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"5a5ae5dd_59d617f2","line":134,"range":{"start_line":129,"start_character":0,"end_line":134,"end_character":47},"updated":"2016-02-05 16:00:21.000000000","message":"How do you plan to implement disabling individual node? Right now the disabling of a host will just exclude it from scheduling and you cannot exclude a single node from scheduling (scheduling chooses a host, not a node).","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"98bac9ad3761136786449e315c7674c0c3727658","unresolved":false,"context_lines":[{"line_number":126,"context_line":"Disabling"},{"line_number":127,"context_line":"---------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"We need to allow individual node disabling in the same manner that we now allow"},{"line_number":130,"context_line":"disabling just one node."},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"By adding this individual disabling possibility to the group disabling option"},{"line_number":133,"context_line":"we will now have multiple ways of disabling a service, by disabling it as a"},{"line_number":134,"context_line":"whole or by disabling all its individual nodes."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"Capabilities"},{"line_number":137,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a57f1b5_5a93d75b","line":134,"range":{"start_line":129,"start_character":0,"end_line":134,"end_character":47},"in_reply_to":"5a5ae5dd_59d617f2","updated":"2016-02-10 15:16:50.000000000","message":"You are correct, and I had forgotten that disabling of the nodes was never meant to have the same meaning as disabling of the host group in the first place.","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"58a8aef2204f95bddb25a3759a8525dc7b1831f2","unresolved":false,"context_lines":[{"line_number":298,"context_line":"REST API impact"},{"line_number":299,"context_line":"---------------"},{"line_number":300,"context_line":""},{"line_number":301,"context_line":"To report status of services we will add an optional parameter called"},{"line_number":302,"context_line":"``node_detail`` which will report the node breakdown."},{"line_number":303,"context_line":""},{"line_number":304,"context_line":"When ``node_detail`` is not set we will report exactly as we are doing now, to"},{"line_number":305,"context_line":"be backward compatible with clients, so we\u0027ll be sending the service is up if"},{"line_number":306,"context_line":"*any* the nodes forming the host set is up and sending that it is disabled if"},{"line_number":307,"context_line":"the service is globally disabled or all the nodes of the service are disabled."},{"line_number":308,"context_line":""},{"line_number":309,"context_line":"When ``node_detail`` is set to False, which tells us client knows of the new"},{"line_number":310,"context_line":"API options, we\u0027ll return new fields:"}],"source_content_type":"text/x-rst","patch_set":8,"id":"5a5ae5dd_1932df5f","line":307,"range":{"start_line":301,"start_character":0,"end_line":307,"end_character":78},"updated":"2016-02-05 16:00:21.000000000","message":"I believe instead of differentiate not setting it, you should create a new microversion.","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"98bac9ad3761136786449e315c7674c0c3727658","unresolved":false,"context_lines":[{"line_number":298,"context_line":"REST API impact"},{"line_number":299,"context_line":"---------------"},{"line_number":300,"context_line":""},{"line_number":301,"context_line":"To report status of services we will add an optional parameter called"},{"line_number":302,"context_line":"``node_detail`` which will report the node breakdown."},{"line_number":303,"context_line":""},{"line_number":304,"context_line":"When ``node_detail`` is not set we will report exactly as we are doing now, to"},{"line_number":305,"context_line":"be backward compatible with clients, so we\u0027ll be sending the service is up if"},{"line_number":306,"context_line":"*any* the nodes forming the host set is up and sending that it is disabled if"},{"line_number":307,"context_line":"the service is globally disabled or all the nodes of the service are disabled."},{"line_number":308,"context_line":""},{"line_number":309,"context_line":"When ``node_detail`` is set to False, which tells us client knows of the new"},{"line_number":310,"context_line":"API options, we\u0027ll return new fields:"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a57f1b5_ba044b7e","line":307,"range":{"start_line":301,"start_character":0,"end_line":307,"end_character":78},"in_reply_to":"5a5ae5dd_1932df5f","updated":"2016-02-10 15:16:50.000000000","message":"I don\u0027t see the benefit of adding a microversion for a backward compatible API change.","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"58a8aef2204f95bddb25a3759a8525dc7b1831f2","unresolved":false,"context_lines":[{"line_number":377,"context_line":"Work Items"},{"line_number":378,"context_line":"----------"},{"line_number":379,"context_line":""},{"line_number":380,"context_line":"- Ad the new ``service_nodes`` table."},{"line_number":381,"context_line":""},{"line_number":382,"context_line":"- Add `node_name` configuration option."},{"line_number":383,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"5a5ae5dd_5944d7b5","line":380,"range":{"start_line":380,"start_character":2,"end_line":380,"end_character":4},"updated":"2016-02-05 16:00:21.000000000","message":"s/Ad/Add","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"98bac9ad3761136786449e315c7674c0c3727658","unresolved":false,"context_lines":[{"line_number":377,"context_line":"Work Items"},{"line_number":378,"context_line":"----------"},{"line_number":379,"context_line":""},{"line_number":380,"context_line":"- Ad the new ``service_nodes`` table."},{"line_number":381,"context_line":""},{"line_number":382,"context_line":"- Add `node_name` configuration option."},{"line_number":383,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a57f1b5_ba18cb85","line":380,"range":{"start_line":380,"start_character":2,"end_line":380,"end_character":4},"in_reply_to":"5a5ae5dd_5944d7b5","updated":"2016-02-10 15:16:50.000000000","message":"Done","commit_id":"c406b93d607e174b447bf7a9a5cf8bb1e01d11df"},{"author":{"_account_id":14274,"name":"XinXiaohui","email":"xiaohui.xin@intel.com","username":"XinXiaohui"},"change_message_id":"6b942a121f452b39afcf99c97e12ee14fbaca9ed","unresolved":false,"context_lines":[{"line_number":130,"context_line":"only disabling a host as a whole, but also disabling individual nodes in the"},{"line_number":131,"context_line":"host group."},{"line_number":132,"context_line":""},{"line_number":133,"context_line":"Disabing a host group and disabling a node have different purposes and will"},{"line_number":134,"context_line":"have different effects on the cinder-volume nodes."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"Disabling a host group is meant to prevent the host from being used by the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dae33548_bcbf45fa","line":133,"range":{"start_line":133,"start_character":0,"end_line":133,"end_character":8},"updated":"2016-02-19 00:52:53.000000000","message":"s/Disabing/Disabling","commit_id":"fa7858b7e9de41075ea9170743637122821af8a4"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"cfd0cf3d598772bbb42e4a074931d69ff5814c5b","unresolved":false,"context_lines":[{"line_number":137,"context_line":"scheduler in volume creations and migrations and will continue to have this"},{"line_number":138,"context_line":"behavior."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"Disabling a specific node on the other hand is done to stop it from working on"},{"line_number":141,"context_line":"anything, not just creation, but also attaching, detaching, etc.  The use of"},{"line_number":142,"context_line":"this kind of disabling operation is meant to effectively exclude a node from"},{"line_number":143,"context_line":"all operations, for example because we are seeing too many errors on the node"},{"line_number":144,"context_line":"or because we want to temporary reduce the workload on the server."},{"line_number":145,"context_line":""},{"line_number":146,"context_line":"To implement this we will have a periodic task that will check in the DB if the"},{"line_number":147,"context_line":"node we are running in is enabled or disabled and stop/start the RPC server"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dae33548_f4928f56","line":144,"range":{"start_line":140,"start_character":0,"end_line":144,"end_character":66},"updated":"2016-02-15 20:03:38.000000000","message":"This is clearly an operator-facing API. What is the difference for them between this and simply sending SIGKILL to the service process?\n\nService disable API was historically implemented in Nova to be able to execute host maintenance:\n\n1. Disable the host A so no new VMs will get scheduled on it.\n2. Live-migrate VMs out of A. In the meanwhile all the API requests made on the VMs still on A are working fine.\n3. Once A have no VMs - patch it, reboot it, do anything.\n\nThis pattern seems useful also in Cinder - if someone is running LVM backend.\n\nGiven that I don\u0027t see much value in implementing it for individual HA nodes, as stopping the service process has almost the same effect (it even stops in a graceful way).\n\nThis \"almost\" is there because in case you\u0027re describing we\u0027re able to turn off individual backends on a multi-backend node.\n\nSumming up - I don\u0027t think this needs to be a vital part of this spec. If operators would have interest in such feature we can add it later.","commit_id":"fa7858b7e9de41075ea9170743637122821af8a4"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"7cf46cd2578f2f992758b165faf3618e50628a7a","unresolved":false,"context_lines":[{"line_number":137,"context_line":"scheduler in volume creations and migrations and will continue to have this"},{"line_number":138,"context_line":"behavior."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"Disabling a specific node on the other hand is done to stop it from working on"},{"line_number":141,"context_line":"anything, not just creation, but also attaching, detaching, etc.  The use of"},{"line_number":142,"context_line":"this kind of disabling operation is meant to effectively exclude a node from"},{"line_number":143,"context_line":"all operations, for example because we are seeing too many errors on the node"},{"line_number":144,"context_line":"or because we want to temporary reduce the workload on the server."},{"line_number":145,"context_line":""},{"line_number":146,"context_line":"To implement this we will have a periodic task that will check in the DB if the"},{"line_number":147,"context_line":"node we are running in is enabled or disabled and stop/start the RPC server"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dae33548_daddce15","line":144,"range":{"start_line":140,"start_character":0,"end_line":144,"end_character":66},"in_reply_to":"dae33548_f4928f56","updated":"2016-02-16 14:10:45.000000000","message":"The reason why this is in the specs is because this was requested in Tokyo\u0027s session and the other day I asked this during the cinder meeting to make sure that we really wanted this and I was told again that this was something we needed.\n\nYou are right that we now have graceful shutdown in Cinder and it will wait until all operations are completed if it is properly setup and the request is properly done.\n\nThe only 2 differences that we would have between that method and this one are:\n\n1- With this one it would always stop as expected independent of Cinder\u0027s configuration and we don\u0027t have to worry about sending the kill signal to the right process.\n\n2- You will be able to make the node start receiving operations again directly since the service will not be stopped.\n\nThe reason why I asked this was because I also think we don\u0027t need it, since nodes can be drained manually using a normal signal, and adding this makes Cinder code a lot more complicated (I have most of this functionality done and it adds a lot of complexity).","commit_id":"fa7858b7e9de41075ea9170743637122821af8a4"},{"author":{"_account_id":14274,"name":"XinXiaohui","email":"xiaohui.xin@intel.com","username":"XinXiaohui"},"change_message_id":"6b942a121f452b39afcf99c97e12ee14fbaca9ed","unresolved":false,"context_lines":[{"line_number":406,"context_line":"Other contributors:"},{"line_number":407,"context_line":"  Michal Dulko (dulek)"},{"line_number":408,"context_line":"  Scott DAngelo (scottda)"},{"line_number":409,"context_line":"  Anyone is welcome to help"},{"line_number":410,"context_line":""},{"line_number":411,"context_line":"Work Items"},{"line_number":412,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dae33548_0759fa8e","line":409,"range":{"start_line":409,"start_character":23,"end_line":409,"end_character":27},"updated":"2016-02-19 00:52:53.000000000","message":"I\u0027d like to work with the capability report part in this spec. Not sure what\u0027s the current status for this?","commit_id":"fa7858b7e9de41075ea9170743637122821af8a4"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"682833edc3a45c64f293802b222d968295be2f92","unresolved":false,"context_lines":[{"line_number":406,"context_line":"Other contributors:"},{"line_number":407,"context_line":"  Michal Dulko (dulek)"},{"line_number":408,"context_line":"  Scott DAngelo (scottda)"},{"line_number":409,"context_line":"  Anyone is welcome to help"},{"line_number":410,"context_line":""},{"line_number":411,"context_line":"Work Items"},{"line_number":412,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"7af24918_747ce520","line":409,"range":{"start_line":409,"start_character":23,"end_line":409,"end_character":27},"in_reply_to":"dae33548_0759fa8e","updated":"2016-03-01 13:57:51.000000000","message":"Ok, I have implemented everything else and left the capability reporting part to you.  You should probably add your patch after the last patch of the dependency chain: https://review.openstack.org/#/c/286601/","commit_id":"fa7858b7e9de41075ea9170743637122821af8a4"}],"specs/newton/ha-aa-job-distribution.rst":[{"author":{"_account_id":6804,"name":"bruce-benjamin","email":"bruce.benjamin@jhuapl.edu","username":"bruce-benjamin"},"change_message_id":"e88b43af1c568b2cc64a086e3ec42c44280ae43f","unresolved":false,"context_lines":[{"line_number":140,"context_line":"`graceful_shutdown_timeout` configuration option in our [DEFAULT] configuration"},{"line_number":141,"context_line":"section or the 60 seconds default timeout will be used, and we could end up"},{"line_number":142,"context_line":"abruptly stopping the service if any of the operations take longer than 60"},{"line_number":143,"context_line":"seconds."},{"line_number":144,"context_line":""},{"line_number":145,"context_line":"Capabilities"},{"line_number":146,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"7af24918_7341910b","line":143,"range":{"start_line":143,"start_character":0,"end_line":143,"end_character":8},"updated":"2016-03-01 16:08:59.000000000","message":"I have a concern here about the default value possibly being too low.  If there is a temporary network outage, you could lose the service and some user data.  To reduce the chance of that, wouldn\u0027t it be better to have a higher default value (2 to 5 mins?) with the option to optimize the timeout when it seems appropriate? Just something to consider.","commit_id":"354234bd99546b6ae604a4f59b2e85fbbd1852e9"},{"author":{"_account_id":9535,"name":"Gorka Eguileor","email":"geguileo@redhat.com","username":"Gorka"},"change_message_id":"e69978f3dc4f1ad581312a446e331b9bd621d086","unresolved":false,"context_lines":[{"line_number":140,"context_line":"`graceful_shutdown_timeout` configuration option in our [DEFAULT] configuration"},{"line_number":141,"context_line":"section or the 60 seconds default timeout will be used, and we could end up"},{"line_number":142,"context_line":"abruptly stopping the service if any of the operations take longer than 60"},{"line_number":143,"context_line":"seconds."},{"line_number":144,"context_line":""},{"line_number":145,"context_line":"Capabilities"},{"line_number":146,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"7af24918_82467e21","line":143,"range":{"start_line":143,"start_character":0,"end_line":143,"end_character":8},"in_reply_to":"7af24918_7341910b","updated":"2016-03-01 17:48:08.000000000","message":"I wrote those 2 paragraphs to provide context for people who don\u0027t know that the current code already supports graceful shutdown and for people who will try to use it so that they know that there\u0027s a configuration option that is very important for a proper graceful shutdown.\n\nI guess I will have to make it more clear that this is NOT going to be changed by this spec and that it is something that we already have in Cinder and this spec or implementation is not going to change in any way.","commit_id":"354234bd99546b6ae604a4f59b2e85fbbd1852e9"},{"author":{"_account_id":11904,"name":"Sean McGinnis","email":"sean.mcginnis@gmail.com","username":"SeanM"},"change_message_id":"91e74dc037d1d8a094965dbcac705f125db7abb0","unresolved":false,"context_lines":[{"line_number":193,"context_line":"expensive operation and in Active-Active configurations we\u0027ll have multiple"},{"line_number":194,"context_line":"nodes requesting the same capabilities with the same frequency for the same"},{"line_number":195,"context_line":"back-end, so capability caching could be a good solution to decrease the cost"},{"line_number":196,"context_line":"of the gathering on the backend."},{"line_number":197,"context_line":""},{"line_number":198,"context_line":"Alternatives"},{"line_number":199,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"5a18252c_530b04d6","line":196,"updated":"2016-04-14 22:57:41.000000000","message":"+1","commit_id":"cfd1d7377e582496a858543b5374e8d13303dcfb"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"736b91dd82c240a1e18e9f11fa8a217baa55d247","unresolved":false,"context_lines":[{"line_number":275,"context_line":"- Removal of ``report_count`` field"},{"line_number":276,"context_line":"- Removal of ``modified_at`` field since we can go back to using ``updated_at``"},{"line_number":277,"context_line":"  field now that heartbeats will be reported on another table."},{"line_number":278,"context_line":"- Removal of all version fields ``rpc_current_version``,"},{"line_number":279,"context_line":"  ``object_current_version`` as they will be moved to ``service_nodes`` table."},{"line_number":280,"context_line":""},{"line_number":281,"context_line":"New ``service_nodes`` table will have following fields:"},{"line_number":282,"context_line":""}],"source_content_type":"text/x-rst","patch_set":13,"id":"7af24918_23843e5c","line":279,"range":{"start_line":278,"start_character":0,"end_line":279,"end_character":78},"updated":"2016-03-03 14:35:46.000000000","message":"Ah, I haven\u0027t thought of that. Yeah…","commit_id":"cfd1d7377e582496a858543b5374e8d13303dcfb"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"736b91dd82c240a1e18e9f11fa8a217baa55d247","unresolved":false,"context_lines":[{"line_number":311,"context_line":"In O release:"},{"line_number":312,"context_line":""},{"line_number":313,"context_line":"- Service availability will only be done checking the ``service_nodes`` table."},{"line_number":314,"context_line":"- Minimum version detection will only be done checking the ``service_nodes``"},{"line_number":315,"context_line":"  table."},{"line_number":316,"context_line":""},{"line_number":317,"context_line":"- As a final step of the rolling upgrade we\u0027ll drop fields ``report_count``,"},{"line_number":318,"context_line":"  ``modified_at``, ``rpc_current_version``, and ``object_current_version`` from"}],"source_content_type":"text/x-rst","patch_set":13,"id":"7af24918_234c3e9d","line":315,"range":{"start_line":314,"start_character":0,"end_line":315,"end_character":8},"updated":"2016-03-03 14:35:46.000000000","message":"This is possible only because only a service can modify it\u0027s own row. Anyway - good use of the assumption to simplify things. :)","commit_id":"cfd1d7377e582496a858543b5374e8d13303dcfb"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"736b91dd82c240a1e18e9f11fa8a217baa55d247","unresolved":false,"context_lines":[{"line_number":314,"context_line":"- Minimum version detection will only be done checking the ``service_nodes``"},{"line_number":315,"context_line":"  table."},{"line_number":316,"context_line":""},{"line_number":317,"context_line":"- As a final step of the rolling upgrade we\u0027ll drop fields ``report_count``,"},{"line_number":318,"context_line":"  ``modified_at``, ``rpc_current_version``, and ``object_current_version`` from"},{"line_number":319,"context_line":"  ``services`` table as they have been moved to ``service_nodes`` table."},{"line_number":320,"context_line":""},{"line_number":321,"context_line":"REST API impact"},{"line_number":322,"context_line":"---------------"}],"source_content_type":"text/x-rst","patch_set":13,"id":"7af24918_8639e0d9","line":319,"range":{"start_line":317,"start_character":0,"end_line":319,"end_character":72},"updated":"2016-03-03 14:35:46.000000000","message":"I sign up to write the infrastructure for post-upgrade migrations - finally…","commit_id":"cfd1d7377e582496a858543b5374e8d13303dcfb"}]}
