)]}'
{"specs/rocky/support-placement-api.rst":[{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":10,"context_line":""},{"line_number":11,"context_line":"https://blueprints.launchpad.net/cinder/+spec/placeholder-for-link"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"This blueprint proposes to integrate Nova\u0027s placement service in cinder"},{"line_number":14,"context_line":"to support scheduler HA."},{"line_number":15,"context_line":""},{"line_number":16,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_32b8b9c6","line":13,"updated":"2018-04-10 09:42:54.000000000","message":"Just call it placement, it will be out of nova soon enough.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"5afc2249e4d27e977a0641c516bc71c921a2c665","unresolved":false,"context_lines":[{"line_number":10,"context_line":""},{"line_number":11,"context_line":"https://blueprints.launchpad.net/cinder/+spec/placeholder-for-link"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"This blueprint proposes to integrate Nova\u0027s placement service in cinder"},{"line_number":14,"context_line":"to support scheduler HA."},{"line_number":15,"context_line":""},{"line_number":16,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_0ed3113a","line":13,"in_reply_to":"bf659307_32b8b9c6","updated":"2018-04-10 12:08:01.000000000","message":"Will update in the next patch.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":5997,"name":"Walt","display_name":"Hemna","email":"waboring@hemna.com","username":"walter-boring","status":"SAP"},"change_message_id":"4de83882d1cff5a4071b5e169c6f095cfa97e1f0","unresolved":false,"context_lines":[{"line_number":19,"context_line":"Now cinder scheduler service can\u0027t be deployed in an HA mode. The reason"},{"line_number":20,"context_line":"is every scheduler service will keep a independent copy of backend status,"},{"line_number":21,"context_line":"thus there could be a data inconsistency among scheduler instances before"},{"line_number":22,"context_line":"volume services broadcast the new backend status."},{"line_number":23,"context_line":""},{"line_number":24,"context_line":"Use Cases"},{"line_number":25,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_432ebaa5","line":22,"updated":"2018-04-10 16:19:17.000000000","message":"have you spoke with Gorka about this?   It seems he has been tackling this issue for a while with inconsistent volume stats.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":5997,"name":"Walt","display_name":"Hemna","email":"waboring@hemna.com","username":"walter-boring","status":"SAP"},"change_message_id":"4de83882d1cff5a4071b5e169c6f095cfa97e1f0","unresolved":false,"context_lines":[{"line_number":27,"context_line":"TBD"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"Proposed change"},{"line_number":30,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"1. Provider \u0026 Inventory \u0026 Trait model"},{"line_number":33,"context_line":"-------------------------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_d50b5d21","line":30,"updated":"2018-04-10 16:19:17.000000000","message":"can you provide a description of what and why the placement server provides for us?  Are we going to have to deploy our own placement server along cinder services too now?","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":110,"context_line":"               |(RC:THICK_DISK_MIB)         |(RC:THIN_DISK_MIB)"},{"line_number":111,"context_line":"     +--------------------------+    +--------------------------+"},{"line_number":112,"context_line":"     |Inventory(if pool support)|    |Inventory(if pool support)|"},{"line_number":113,"context_line":"     +--------------------------+    +--------------------------+"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"**NOTE**:"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_1287b511","line":113,"updated":"2018-04-10 09:42:54.000000000","message":"Let me see if I\u0027m understanding this. At the top of the tree \"Provider: hostname\" is the system that gets told \"I\u0027m taking some storage from you\". It hosts multiple storage pools, each which can have different style of storage, yes?\n\nThe reason I ask is because in placement the expectation is that when representing nested providers, the thing which is the root of the tree of the providers is the target to which a request for resource use is dispatched (even if it is something within the tree on which the workload actually lands).\n\nThe other thing to note here is that it would be ideal to record inventory (the resource class)  in absolute terms. So rather than creating a new resource class that distinguishes between thick and thin, the DISK_GB resource class is used (floor-ing any fractional GBs) and a pool thick or thin-ness is distinguished by traits and manipulating the allocation ratio.\n\nThe \u0027total\u0027 field of inventory should change rarely as it should represent the real physical capacity of the storage media. When doing thin provisioning the allocation_ratio would be set pretty high.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"5afc2249e4d27e977a0641c516bc71c921a2c665","unresolved":false,"context_lines":[{"line_number":110,"context_line":"               |(RC:THICK_DISK_MIB)         |(RC:THIN_DISK_MIB)"},{"line_number":111,"context_line":"     +--------------------------+    +--------------------------+"},{"line_number":112,"context_line":"     |Inventory(if pool support)|    |Inventory(if pool support)|"},{"line_number":113,"context_line":"     +--------------------------+    +--------------------------+"},{"line_number":114,"context_line":""},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"**NOTE**:"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_ce82f9c8","line":113,"in_reply_to":"bf659307_1287b511","updated":"2018-04-10 12:08:01.000000000","message":"Thanks, so the root provider should be the storage backend provider rather than volume service\u0027s provider, right? Regarding THICK and THIN pool, I personally guess that is a little complicated that one storage pool may both support thin and thick provisioning. Does placement support provider to have two inventorys which have identical resouce class? I mean one for thick and another for thin. Also because of the same reason I just mentioned, the \u0027total\u0027 field for thick_disk_gb or thin_disk_gb inventory would change if opposite kind of volume is created, and what\u0027s more administrator can configure the same backend storage pool for different cinder/openstack systems.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":116,"context_line":"**NOTE**:"},{"line_number":117,"context_line":"1. Inventory ``THICK_DISK_MIB`` and ``THIN_DISK_MIB`` will be added only"},{"line_number":118,"context_line":"when backend announces that is supported via ``thin_provisioning_support``"},{"line_number":119,"context_line":"and ``thick_provisioning_support``."},{"line_number":120,"context_line":"2. GiB is the first choice, but considering Placement doesn\u0027t support"},{"line_number":121,"context_line":"float while our backend driver will report capacity within float,"},{"line_number":122,"context_line":"we can use MiB instead."}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_b27da91d","line":119,"updated":"2018-04-10 09:42:54.000000000","message":"See above. And also:\n\nIntroducing a new resource class will upset the existing resource class handling already happening in nova, which presumably will need to interoperate with any disk handling managed by cinder.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"5afc2249e4d27e977a0641c516bc71c921a2c665","unresolved":false,"context_lines":[{"line_number":116,"context_line":"**NOTE**:"},{"line_number":117,"context_line":"1. Inventory ``THICK_DISK_MIB`` and ``THIN_DISK_MIB`` will be added only"},{"line_number":118,"context_line":"when backend announces that is supported via ``thin_provisioning_support``"},{"line_number":119,"context_line":"and ``thick_provisioning_support``."},{"line_number":120,"context_line":"2. GiB is the first choice, but considering Placement doesn\u0027t support"},{"line_number":121,"context_line":"float while our backend driver will report capacity within float,"},{"line_number":122,"context_line":"we can use MiB instead."}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_2e65b545","line":119,"in_reply_to":"bf659307_b27da91d","updated":"2018-04-10 12:08:01.000000000","message":"This worths discussing :)","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":131,"context_line":"    3. ``min_unit``: 1024"},{"line_number":132,"context_line":"    4. ``reserved``: Floor(``total_capacity_gb`` * Float(``reserved_percentage``/100))*1024"},{"line_number":133,"context_line":"    5. ``step_size``: 1024 in default"},{"line_number":134,"context_line":"    6. ``total``: Floor(``provisioned_capacity_gb(allocated_capacity_gb)`` + ``free_capacity_gb``)*1024"},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"2. For **THIN** pool"},{"line_number":137,"context_line":"    1. ``allocation_ratio``: ``max_over_subscription_ratio``, also it\u0027s can be calculated"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_d2cb7d47","line":134,"updated":"2018-04-10 09:42:54.000000000","message":"I don\u0027t have enough knowledge of the various fields on the right hand side here, but the main thing to keep in mind is that most of the time these value usually should only change when the physical capacity of the storage pool changes and should represent the real physical capacity of the media.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":154,"context_line":""},{"line_number":155,"context_line":"**Question**:"},{"line_number":156,"context_line":"How to support the case when driver reports the ``total_capacity_gb`` is"},{"line_number":157,"context_line":"``infinite`` or ``unknown`` rather than digital number?"},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Regarding traits, we would extract our common capabilities to traits"},{"line_number":160,"context_line":"when reporting:"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_722151fb","line":157,"updated":"2018-04-10 09:42:54.000000000","message":"Although at one point I tried to convince other placement people that there should be infinite resource classes \u003chttps://anticdent.org/simple-resource-provision.html\u003e that didn\u0027t fly, so at this stage a big number is probably the way to go. The max limit for total is currently 2147483647.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":164,"context_line":"3. ``CUSTOM_MULTIATTACH``: corresponding to ``multiattach``"},{"line_number":165,"context_line":"4. ``CUSTOM_PROTOCOL_ISCSI``: corresponding to ``storage_protocol\u003diSCSI``"},{"line_number":166,"context_line":"5. ``CUSTOM_PROTOCOL_DRBD``: corresponding to ``storage_protocol\u003dDRBD``"},{"line_number":167,"context_line":"6. Other storage custom protocol traits"},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"2. Update in volume service"},{"line_number":170,"context_line":"---------------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_52652d44","line":167,"updated":"2018-04-10 09:42:54.000000000","message":"It\u0027s quite likely some of these should be standard traits and added to os-traits: https://pypi.org/project/os-traits/\n\nThere\u0027s a storage section in there.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"5afc2249e4d27e977a0641c516bc71c921a2c665","unresolved":false,"context_lines":[{"line_number":164,"context_line":"3. ``CUSTOM_MULTIATTACH``: corresponding to ``multiattach``"},{"line_number":165,"context_line":"4. ``CUSTOM_PROTOCOL_ISCSI``: corresponding to ``storage_protocol\u003diSCSI``"},{"line_number":166,"context_line":"5. ``CUSTOM_PROTOCOL_DRBD``: corresponding to ``storage_protocol\u003dDRBD``"},{"line_number":167,"context_line":"6. Other storage custom protocol traits"},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"2. Update in volume service"},{"line_number":170,"context_line":"---------------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_cee99980","line":167,"in_reply_to":"bf659307_52652d44","updated":"2018-04-10 12:08:01.000000000","message":"Yes, will add some of them to the sandard traits when we agreed on the standard trait list.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":170,"context_line":"---------------------------"},{"line_number":171,"context_line":""},{"line_number":172,"context_line":"1. New Class ``ProviderTree`` will be added to represent service\u0027s"},{"line_number":173,"context_line":"   backend provider tree and cache Placement info."},{"line_number":174,"context_line":"2. New Class ``ReportClient`` will be added which acts as a Placement"},{"line_number":175,"context_line":"   proxy to create/delete new allocations, get candidates, update"},{"line_number":176,"context_line":"   inventories and etc."}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_72e4f1ac","line":173,"updated":"2018-04-10 09:42:54.000000000","message":"Despite its use in nova, I\u0027m not confident that the use of caching in the ProviderTree is universally useful. It adds a lot of complexity that can be avoided by simply asking placement to provide the information again.\n\nThe notion of having a class to manage the hierarchy is useful, though, if there nested structure you\u0027ve drawn above is the norm (rather than lots of disparate single providers of storage).","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"73ffb93f3d64295f5679f444ee8198f41c475621","unresolved":false,"context_lines":[{"line_number":196,"context_line":"1. Scheduler will now ask Placement for provider (backend) candidates"},{"line_number":197,"context_line":"   before requiring plugins to filter and weigh backends."},{"line_number":198,"context_line":"2. Scheduler will now create allocations right after backend is"},{"line_number":199,"context_line":"   chosen."},{"line_number":200,"context_line":"3. ``consume_from_volume`` in scheduler will be removed."},{"line_number":201,"context_line":"4. CapacityFilter plugin will be updated since capacity related comparision"},{"line_number":202,"context_line":"   will be moved to Placement service."}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_12a2554c","line":199,"updated":"2018-04-10 09:42:54.000000000","message":"Another step here is if, after the backend is chosen the backend is unable to accept the workload, the allocations needs to be cleared up.\n\nThis can happen semi-automatically in the case where a different backend is chosen and that one works: rewriting the allocations with the same consumer id but different targets will remove the previous allocations.","commit_id":"6837d9222ff971925df6259515a891d705580c47"},{"author":{"_account_id":23083,"name":"TommyLike","email":"tommylikehu@gmail.com","username":"TommyLike"},"change_message_id":"5afc2249e4d27e977a0641c516bc71c921a2c665","unresolved":false,"context_lines":[{"line_number":196,"context_line":"1. Scheduler will now ask Placement for provider (backend) candidates"},{"line_number":197,"context_line":"   before requiring plugins to filter and weigh backends."},{"line_number":198,"context_line":"2. Scheduler will now create allocations right after backend is"},{"line_number":199,"context_line":"   chosen."},{"line_number":200,"context_line":"3. ``consume_from_volume`` in scheduler will be removed."},{"line_number":201,"context_line":"4. CapacityFilter plugin will be updated since capacity related comparision"},{"line_number":202,"context_line":"   will be moved to Placement service."}],"source_content_type":"text/x-rst","patch_set":2,"id":"bf659307_aeac6535","line":199,"in_reply_to":"bf659307_12a2554c","updated":"2018-04-10 12:08:01.000000000","message":"Is this related to rescheduling? It\u0027s mentioned in the volume service part","commit_id":"6837d9222ff971925df6259515a891d705580c47"}]}
