)]}'
{"specs/queens/cinder-storage-aggregates.rst":[{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"79e0852bbcf48146c58f9c5bab68a0669eb4d880","unresolved":false,"context_lines":[{"line_number":14,"context_line":"Problem description\r"},{"line_number":15,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"},{"line_number":16,"context_line":"\r"},{"line_number":17,"context_line":"In a large-scale cloud, there are thousands of storage backends with a lot of\r"},{"line_number":18,"context_line":"types, such as Ceph RBD, EMC, IBM and so on.  As we usually create volume with\r"},{"line_number":19,"context_line":"a specific volume type, one volume type for one storage backend may be\r"},{"line_number":20,"context_line":"inconvenience, so one volume type for multi-backends is recommended. In the\r"},{"line_number":21,"context_line":"case, storage aggregate is recommended to schedule storage backends.\r"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5f201791_238e0d2e","line":18,"range":{"start_line":17,"start_character":0,"end_line":18,"end_character":44},"updated":"2017-06-23 10:56:10.000000000","message":"Is there evidence of this? Certainly there are clouds with many backends, but I\u0027m not aware of any system with more than 3 backend types, and that was an outlier, built that way to consume previously bought storage.","commit_id":"48234038781891cf5b34b0f4513ee6a30a61fc5c"},{"author":{"_account_id":18529,"name":"Dongdong Pei","email":"peidongdong121@163.com","username":"dongdongpei"},"change_message_id":"9304e5f442911fb319b852814834f7c399e52816","unresolved":false,"context_lines":[{"line_number":14,"context_line":"Problem description\r"},{"line_number":15,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"},{"line_number":16,"context_line":"\r"},{"line_number":17,"context_line":"In a large-scale cloud, there are thousands of storage backends with a lot of\r"},{"line_number":18,"context_line":"types, such as Ceph RBD, EMC, IBM and so on.  As we usually create volume with\r"},{"line_number":19,"context_line":"a specific volume type, one volume type for one storage backend may be\r"},{"line_number":20,"context_line":"inconvenience, so one volume type for multi-backends is recommended. In the\r"},{"line_number":21,"context_line":"case, storage aggregate is recommended to schedule storage backends.\r"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5f201791_7dddb06e","line":18,"range":{"start_line":17,"start_character":0,"end_line":18,"end_character":44},"in_reply_to":"5f201791_238e0d2e","updated":"2017-06-24 08:06:06.000000000","message":"Thanks for you review. My company is using more than three backend types with different performance and available for different scenarios.That\u0027s why I\u0027m thinking about it.","commit_id":"48234038781891cf5b34b0f4513ee6a30a61fc5c"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"79e0852bbcf48146c58f9c5bab68a0669eb4d880","unresolved":false,"context_lines":[{"line_number":17,"context_line":"In a large-scale cloud, there are thousands of storage backends with a lot of\r"},{"line_number":18,"context_line":"types, such as Ceph RBD, EMC, IBM and so on.  As we usually create volume with\r"},{"line_number":19,"context_line":"a specific volume type, one volume type for one storage backend may be\r"},{"line_number":20,"context_line":"inconvenience, so one volume type for multi-backends is recommended. In the\r"},{"line_number":21,"context_line":"case, storage aggregate is recommended to schedule storage backends.\r"},{"line_number":22,"context_line":"\r"},{"line_number":23,"context_line":"Use Cases\r"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5f201791_2393ad51","line":21,"range":{"start_line":20,"start_character":69,"end_line":21,"end_character":67},"updated":"2017-06-23 10:56:10.000000000","message":"Recommended by whom? \n\nVolume types are in no way tied to a backend name; they certainly *can* be, but the can just as easily be tied to a backend type, a list of backend types, an intercetional list of names and types, or literally any other feature the backend expresses, including operator defined backend labels.","commit_id":"48234038781891cf5b34b0f4513ee6a30a61fc5c"},{"author":{"_account_id":18529,"name":"Dongdong Pei","email":"peidongdong121@163.com","username":"dongdongpei"},"change_message_id":"9304e5f442911fb319b852814834f7c399e52816","unresolved":false,"context_lines":[{"line_number":17,"context_line":"In a large-scale cloud, there are thousands of storage backends with a lot of\r"},{"line_number":18,"context_line":"types, such as Ceph RBD, EMC, IBM and so on.  As we usually create volume with\r"},{"line_number":19,"context_line":"a specific volume type, one volume type for one storage backend may be\r"},{"line_number":20,"context_line":"inconvenience, so one volume type for multi-backends is recommended. In the\r"},{"line_number":21,"context_line":"case, storage aggregate is recommended to schedule storage backends.\r"},{"line_number":22,"context_line":"\r"},{"line_number":23,"context_line":"Use Cases\r"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5f201791_1de23c30","line":21,"range":{"start_line":20,"start_character":69,"end_line":21,"end_character":67},"in_reply_to":"5f201791_2393ad51","updated":"2017-06-24 08:06:06.000000000","message":"Thanks for you review. I understand what you say. Volume type in Cinder is equal to host aggregate in Nova in a way. In that case, the feature or something else must be contained in cinder.conf\nahead of time. If you want to change features or something else, you must restart cinder-volume service. If we use storage-aggregate, we can easily aggregate backends without restarting cinder-volume service when different features or something else are needed.","commit_id":"48234038781891cf5b34b0f4513ee6a30a61fc5c"},{"author":{"_account_id":13945,"name":"ZhangYang","email":"wosunoozzy@gmail.com","username":"WosunOoO"},"change_message_id":"8fa966cee4173006767f1e47e03f511b92ebd3d9","unresolved":false,"context_lines":[{"line_number":126,"context_line":"---------------\r"},{"line_number":127,"context_line":"\r"},{"line_number":128,"context_line":"This will introduce at least the 8 new calls mentioned in the proposal section,\r"},{"line_number":129,"context_line":"these will be part of a new API object named aggregate. We\u0027ll treat the\r"},{"line_number":130,"context_line":"aggregate as a first class object in Cinder. The new API will be released under\r"},{"line_number":131,"context_line":"a new microversion.\r"},{"line_number":132,"context_line":"\r"},{"line_number":133,"context_line":"Security impact\r"},{"line_number":134,"context_line":"---------------\r"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5f201791_e8a21bae","line":131,"range":{"start_line":129,"start_character":0,"end_line":131,"end_character":19},"updated":"2017-06-24 08:04:53.000000000","message":"Do we need to introduce the api details here or not.","commit_id":"48234038781891cf5b34b0f4513ee6a30a61fc5c"},{"author":{"_account_id":18529,"name":"Dongdong Pei","email":"peidongdong121@163.com","username":"dongdongpei"},"change_message_id":"d57929c85a92c2b7ce496e0af83915dc9026646b","unresolved":false,"context_lines":[{"line_number":126,"context_line":"---------------\r"},{"line_number":127,"context_line":"\r"},{"line_number":128,"context_line":"This will introduce at least the 8 new calls mentioned in the proposal section,\r"},{"line_number":129,"context_line":"these will be part of a new API object named aggregate. We\u0027ll treat the\r"},{"line_number":130,"context_line":"aggregate as a first class object in Cinder. The new API will be released under\r"},{"line_number":131,"context_line":"a new microversion.\r"},{"line_number":132,"context_line":"\r"},{"line_number":133,"context_line":"Security impact\r"},{"line_number":134,"context_line":"---------------\r"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5f201791_a881c3fd","line":131,"range":{"start_line":129,"start_character":0,"end_line":131,"end_character":19},"in_reply_to":"5f201791_e8a21bae","updated":"2017-06-24 08:08:10.000000000","message":"Will introduce in implementation.","commit_id":"48234038781891cf5b34b0f4513ee6a30a61fc5c"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"b8e11e270f88b1707a8c02ab1bde7840a19e43b7","unresolved":false,"context_lines":[{"line_number":22,"context_line":"Nova in a way, which are used to aggregate storage backends with different\r"},{"line_number":23,"context_line":"features or something else. But some customized features are usually\r"},{"line_number":24,"context_line":"configured in cinder.conf, so we have to restart cinder-volume service\r"},{"line_number":25,"context_line":"for customized features changes. If we use storage aggregates, we can\r"},{"line_number":26,"context_line":"use command line without restarting cinder-volume service.\r"},{"line_number":27,"context_line":"\r"},{"line_number":28,"context_line":"Use Cases\r"},{"line_number":29,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"}],"source_content_type":"text/x-rst","patch_set":5,"id":"5f201791_a480c747","line":26,"range":{"start_line":25,"start_character":33,"end_line":26,"end_character":58},"updated":"2017-06-25 22:32:50.000000000","message":"Changes to volume types are done via the API without restarting anything. Can you give a concrete example of something that you can\u0027t do with volume types that you can do with your new API, please?","commit_id":"11c4d6c20aa727c3ca6e403f91c58cb7c297ae00"},{"author":{"_account_id":18529,"name":"Dongdong Pei","email":"peidongdong121@163.com","username":"dongdongpei"},"change_message_id":"d1abd7290911166e2dabb386222d28d489fd2a00","unresolved":false,"context_lines":[{"line_number":22,"context_line":"Nova in a way, which are used to aggregate storage backends with different\r"},{"line_number":23,"context_line":"features or something else. But some customized features are usually\r"},{"line_number":24,"context_line":"configured in cinder.conf, so we have to restart cinder-volume service\r"},{"line_number":25,"context_line":"for customized features changes. If we use storage aggregates, we can\r"},{"line_number":26,"context_line":"use command line without restarting cinder-volume service.\r"},{"line_number":27,"context_line":"\r"},{"line_number":28,"context_line":"Use Cases\r"},{"line_number":29,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"}],"source_content_type":"text/x-rst","patch_set":5,"id":"5f201791_dc100103","line":26,"range":{"start_line":25,"start_character":33,"end_line":26,"end_character":58},"in_reply_to":"5f201791_7c68fa85","updated":"2017-06-27 13:26:01.000000000","message":"We divide storage backends into different logical groups by logical features, such as perfomance, HA and so on.\n\nFor example, storage backend 1 has performance1 and HA1, storage backend 2 has performance1 and HA2, storage backend 3 has performance2 and HA2.\n\nperformance1 - storage backend 1 \u0026 storage backend 2\n\nperformance2 - storage backend 3\n\nHA1 - storage backend 1\n\nHA2 - storage backend 2 \u0026 storage backend 3\n\nWe can easily define four volume types reference to performance1, performance2, HA1 and HA2.\n\nMeanwhile, We should define performance1, performance2, HA1 adn HA2 for storage backends in cinder.conf.\n\nWe may have changes to those local features in cinder.conf, then we should restart cinder-volume service.","commit_id":"11c4d6c20aa727c3ca6e403f91c58cb7c297ae00"},{"author":{"_account_id":18529,"name":"Dongdong Pei","email":"peidongdong121@163.com","username":"dongdongpei"},"change_message_id":"b08ca3f1cfaa9e2755bc71c508663475198108fd","unresolved":false,"context_lines":[{"line_number":22,"context_line":"Nova in a way, which are used to aggregate storage backends with different\r"},{"line_number":23,"context_line":"features or something else. But some customized features are usually\r"},{"line_number":24,"context_line":"configured in cinder.conf, so we have to restart cinder-volume service\r"},{"line_number":25,"context_line":"for customized features changes. If we use storage aggregates, we can\r"},{"line_number":26,"context_line":"use command line without restarting cinder-volume service.\r"},{"line_number":27,"context_line":"\r"},{"line_number":28,"context_line":"Use Cases\r"},{"line_number":29,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"}],"source_content_type":"text/x-rst","patch_set":5,"id":"5f201791_e77fe1dd","line":26,"range":{"start_line":25,"start_character":33,"end_line":26,"end_character":58},"in_reply_to":"5f201791_a480c747","updated":"2017-06-26 01:53:09.000000000","message":"We compare features configured in volume types with features read from storage backends and cinder.conf to schedule the best storage backends. Really, changes to volume types are done via the API without restarting anything. But changes to customized features configured in cinder.conf need to restart cinder-volume service.","commit_id":"11c4d6c20aa727c3ca6e403f91c58cb7c297ae00"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"0f21aeea42a840e595383c4479ff2dcf85eea6bb","unresolved":false,"context_lines":[{"line_number":22,"context_line":"Nova in a way, which are used to aggregate storage backends with different\r"},{"line_number":23,"context_line":"features or something else. But some customized features are usually\r"},{"line_number":24,"context_line":"configured in cinder.conf, so we have to restart cinder-volume service\r"},{"line_number":25,"context_line":"for customized features changes. If we use storage aggregates, we can\r"},{"line_number":26,"context_line":"use command line without restarting cinder-volume service.\r"},{"line_number":27,"context_line":"\r"},{"line_number":28,"context_line":"Use Cases\r"},{"line_number":29,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\r"}],"source_content_type":"text/x-rst","patch_set":5,"id":"5f201791_7c68fa85","line":26,"range":{"start_line":25,"start_character":33,"end_line":26,"end_character":58},"in_reply_to":"5f201791_e77fe1dd","updated":"2017-06-26 11:15:56.000000000","message":"I\u0027m not trying to be awkward here, I\u0027m just not understanding your issue, so I\u0027m unable to tell whether what you\u0027re proposing makes sense.\n\nCan you please provide a concrete example of something you can do with aggregates that you can\u0027t do with volume types? This will aid my understanding far more than anything in the spec so far. I don\u0027t mind if it\u0027s added to the spec or just put in a comment.","commit_id":"11c4d6c20aa727c3ca6e403f91c58cb7c297ae00"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"b8e11e270f88b1707a8c02ab1bde7840a19e43b7","unresolved":false,"context_lines":[{"line_number":143,"context_line":"Notifications impact\r"},{"line_number":144,"context_line":"--------------------\r"},{"line_number":145,"context_line":"\r"},{"line_number":146,"context_line":"None.\r"},{"line_number":147,"context_line":"\r"},{"line_number":148,"context_line":"Other end user impact\r"},{"line_number":149,"context_line":"---------------------\r"}],"source_content_type":"text/x-rst","patch_set":5,"id":"5f201791_64962f96","line":146,"range":{"start_line":146,"start_character":0,"end_line":146,"end_character":4},"updated":"2017-06-25 22:32:50.000000000","message":"There are definitely going to need to be notifications for CRUD events related to aggregates","commit_id":"11c4d6c20aa727c3ca6e403f91c58cb7c297ae00"}]}
