)]}'
{"specs/liberty/rpc-object-compatibility.rst":[{"author":{"_account_id":12924,"name":"Patrick East","email":"east.patrick@gmail.com","username":"patrick.east"},"change_message_id":"a2cd75eaa9ce32000bc3c30bc9b9f93e0ce5b959","unresolved":false,"context_lines":[{"line_number":189,"context_line":"Testing"},{"line_number":190,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":191,"context_line":""},{"line_number":192,"context_line":"Ideally, there should be a rolling upgrade test within tempest to test basic"},{"line_number":193,"context_line":"RPC and object pinning between different releases."},{"line_number":194,"context_line":""},{"line_number":195,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa32b979_4ea9ad7a","line":192,"updated":"2015-06-16 03:45:39.000000000","message":"The wording here makes it sounds like it should be tested... but may not be. \n\nThere definitely should be testing for this feature, could probably it be integrated with the existing grenade testing framework? If the promise to deployers is that they can do a rolling upgrade we should make sure it actually works :D","commit_id":"3164fbc2ffdc3133daeba6d31e643f37bbc71c10"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"33fd1de0a0fb387d221d21f0a135c88347d30bda","unresolved":false,"context_lines":[{"line_number":189,"context_line":"Testing"},{"line_number":190,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":191,"context_line":""},{"line_number":192,"context_line":"Ideally, there should be a rolling upgrade test within tempest to test basic"},{"line_number":193,"context_line":"RPC and object pinning between different releases."},{"line_number":194,"context_line":""},{"line_number":195,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa32b979_763f91d6","line":192,"in_reply_to":"fa32b979_4ea9ad7a","updated":"2015-06-16 13:55:56.000000000","message":"Yes it should be tested through tempest, but it would only apply from M release onwards because the internals have to be in place to track the versions in the cinder database.  So far, I was only able to test it by creating a \"fake\" release (based off of liberty) and adding the internals there to try out certain RPC calls.  It is not ideal, but since most of the objects and RPC compatibility layers are not in previous releases, it is pretty hard to test.","commit_id":"3164fbc2ffdc3133daeba6d31e643f37bbc71c10"},{"author":{"_account_id":10263,"name":"Vilobh Meshram","email":"vilobhmm@yahoo-inc.com","username":"vilobhmm"},"change_message_id":"28e5d234a21a00b65bd0cbb9744b9bfc76d33837","unresolved":false,"context_lines":[{"line_number":13,"context_line":"The following proposal is to enable rolling upgrade in cinder.  More"},{"line_number":14,"context_line":"specifically, it allows cinder services (cinder-api, cinder-scheduler,"},{"line_number":15,"context_line":"cinder-backup, and cinder-volume) to be updated one at a time and yet still"},{"line_number":16,"context_line":"be operational (only one service is taken down).  It follows the rolling"},{"line_number":17,"context_line":"upgrade feature in nova, using oslo_versionedobjects, but is different in that"},{"line_number":18,"context_line":"there is no indirection API.  Instead, RPC versions and objects are going to"},{"line_number":19,"context_line":"be pinned to a backwards compatible version and later switched to the latest"}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_f3b16890","line":16,"updated":"2015-06-16 17:15:24.000000000","message":"Extra space.","commit_id":"7ee1f9fabcd6bdcf1e53713a7f48dbb1ddbd4df8"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"9f5c366ae7e2064c18111b9669bb7db797b0cdf1","unresolved":false,"context_lines":[{"line_number":13,"context_line":"The following proposal is to enable rolling upgrade in cinder.  More"},{"line_number":14,"context_line":"specifically, it allows cinder services (cinder-api, cinder-scheduler,"},{"line_number":15,"context_line":"cinder-backup, and cinder-volume) to be updated one at a time and yet still"},{"line_number":16,"context_line":"be operational (only one service is taken down).  It follows the rolling"},{"line_number":17,"context_line":"upgrade feature in nova, using oslo_versionedobjects, but is different in that"},{"line_number":18,"context_line":"there is no indirection API.  Instead, RPC versions and objects are going to"},{"line_number":19,"context_line":"be pinned to a backwards compatible version and later switched to the latest"}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_39f491f0","line":16,"in_reply_to":"fa32b979_f3b16890","updated":"2015-06-16 17:36:06.000000000","message":"It is standard to put 2 spaces after a period.","commit_id":"7ee1f9fabcd6bdcf1e53713a7f48dbb1ddbd4df8"},{"author":{"_account_id":10263,"name":"Vilobh Meshram","email":"vilobhmm@yahoo-inc.com","username":"vilobhmm"},"change_message_id":"28e5d234a21a00b65bd0cbb9744b9bfc76d33837","unresolved":false,"context_lines":[{"line_number":50,"context_line":""},{"line_number":51,"context_line":"The implementation is as follows:"},{"line_number":52,"context_line":"1. When a service starts, it registers the RPC and object versions it knows"},{"line_number":53,"context_line":"about, i.e. minimum and latest versions."},{"line_number":54,"context_line":"2. On upgrade, the admin upgrades each individual cinder service.  When the"},{"line_number":55,"context_line":"service starts, it knows that there is a newer version, but continue to run"},{"line_number":56,"context_line":"at the current/pinned version indicated in the database."}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_fe3b6fcf","line":53,"updated":"2015-06-16 17:15:24.000000000","message":"Is this true for cinder-api as well ?","commit_id":"7ee1f9fabcd6bdcf1e53713a7f48dbb1ddbd4df8"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"9f5c366ae7e2064c18111b9669bb7db797b0cdf1","unresolved":false,"context_lines":[{"line_number":50,"context_line":""},{"line_number":51,"context_line":"The implementation is as follows:"},{"line_number":52,"context_line":"1. When a service starts, it registers the RPC and object versions it knows"},{"line_number":53,"context_line":"about, i.e. minimum and latest versions."},{"line_number":54,"context_line":"2. On upgrade, the admin upgrades each individual cinder service.  When the"},{"line_number":55,"context_line":"service starts, it knows that there is a newer version, but continue to run"},{"line_number":56,"context_line":"at the current/pinned version indicated in the database."}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_39cb31a7","line":53,"in_reply_to":"fa32b979_fe3b6fcf","updated":"2015-06-16 17:36:06.000000000","message":"Yes, I mentioned this in L14.","commit_id":"7ee1f9fabcd6bdcf1e53713a7f48dbb1ddbd4df8"},{"author":{"_account_id":1207,"name":"Duncan Thomas","email":"duncan.thomas@gmail.com","username":"duncan-thomas"},"change_message_id":"442686fbcabfcf869e8228322250ddc3756eaa26","unresolved":false,"context_lines":[{"line_number":62,"context_line":"should be using.  The service will continue to use the specific versions"},{"line_number":63,"context_line":"until an all clear is given by the admin (see below)."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"3. While other cinder services are being upgraded, the service has to"},{"line_number":66,"context_line":"constantly check the database for the current and available RPC and object"},{"line_number":67,"context_line":"versions.  A flag is used internally within the service to track if it should"},{"line_number":68,"context_line":"run in backwards compatibility mode (e.g. BACKWARD_COMPAT_MODE), i.e. RPC and"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_f264464a","line":65,"updated":"2015-06-21 12:24:27.000000000","message":"Maybe consider a broadcast RPC to flag the upgrade?  Or at least cache the db table for a few seconds... hitting the db for every RPC (and every main DB access) is going to hurt, a lot","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"1a2626ff1e626f26018b5a9bbfa21aa5712a0227","unresolved":false,"context_lines":[{"line_number":62,"context_line":"should be using.  The service will continue to use the specific versions"},{"line_number":63,"context_line":"until an all clear is given by the admin (see below)."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"3. While other cinder services are being upgraded, the service has to"},{"line_number":66,"context_line":"constantly check the database for the current and available RPC and object"},{"line_number":67,"context_line":"versions.  A flag is used internally within the service to track if it should"},{"line_number":68,"context_line":"run in backwards compatibility mode (e.g. BACKWARD_COMPAT_MODE), i.e. RPC and"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_8f83ef55","line":65,"in_reply_to":"fa32b979_ebb513d0","updated":"2015-06-23 19:37:16.000000000","message":"So I am trying to understand how to send a broadcast out to the services, to turn on (or off) the flag that tells the service to run in backwards compatibility mode.\n\nSince we want to make the objects and RPC interface backwards compatible before it gets sent over RPC, I am currently doing it in the rpcapi.py, which is essentially an RPC client.  The RPC broadcast sends the message to any host that subscribes to the topic.  But we need the RPC client to get that message and update the flag.  Are there examples of how this is done elsewhere?\n\nOtherwise, caching the table in memory is simpler to implement, and I can implement it so that the cache refreshes every few seconds or minutes if the backward_compat_mode \u003d True.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a7ac3b64dde83523b32906461c920b5b073bb8c9","unresolved":false,"context_lines":[{"line_number":62,"context_line":"should be using.  The service will continue to use the specific versions"},{"line_number":63,"context_line":"until an all clear is given by the admin (see below)."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"3. While other cinder services are being upgraded, the service has to"},{"line_number":66,"context_line":"constantly check the database for the current and available RPC and object"},{"line_number":67,"context_line":"versions.  A flag is used internally within the service to track if it should"},{"line_number":68,"context_line":"run in backwards compatibility mode (e.g. BACKWARD_COMPAT_MODE), i.e. RPC and"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_ebb513d0","line":65,"in_reply_to":"fa32b979_f264464a","updated":"2015-06-22 15:48:07.000000000","message":"Fair point. Probably cinder.conf can be an option too.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"3710b92dca376d450fdfb757b160874b5fee2eab","unresolved":false,"context_lines":[{"line_number":62,"context_line":"should be using.  The service will continue to use the specific versions"},{"line_number":63,"context_line":"until an all clear is given by the admin (see below)."},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"3. While other cinder services are being upgraded, the service has to"},{"line_number":66,"context_line":"constantly check the database for the current and available RPC and object"},{"line_number":67,"context_line":"versions.  A flag is used internally within the service to track if it should"},{"line_number":68,"context_line":"run in backwards compatibility mode (e.g. BACKWARD_COMPAT_MODE), i.e. RPC and"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_8481df91","line":65,"in_reply_to":"fa32b979_f264464a","updated":"2015-06-22 13:15:12.000000000","message":"Let me try out the RPC broadcast in my prototype to see if it works.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":14250,"name":"Grzegorz Grasza","email":"xek@redhat.com","username":"xek"},"change_message_id":"2bf29a0777843b871a797b3b663467d32b9b70c1","unresolved":false,"context_lines":[{"line_number":122,"context_line":"    Column(\u0027object_current_version\u0027, String(length\u003d36)),"},{"line_number":123,"context_line":"    Column(\u0027object_available_version\u0027, String(length\u003d36)),"},{"line_number":124,"context_line":"    mysql_engine\u003d\u0027InnoDB\u0027"},{"line_number":125,"context_line":")"},{"line_number":126,"context_line":""},{"line_number":127,"context_line":"The service_id is service name + host.  The *_current_version is the version"},{"line_number":128,"context_line":"a service is currently running and pinned at.  The *_available_version is the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_9e85835a","line":125,"updated":"2015-06-23 09:32:24.000000000","message":"Maybe you can store this in a config variable?\n\nAs I read in the original object proposal, versionedobjects are \"objects that behave the same whether they\u0027re able to access the database directly or required to go over RPC\" and this is what the indirection API is all about.\nhttps://wiki.openstack.org/wiki/ObjectProposal\n\nSo questions to consider are:\n\nIs access to the DB necessary for all Cinder services?\n\nDo we want to make this an additional dependency?\n\nWill there be a versioned object model for this table?\n\nAs I understand, versionedobjects were created to mitigate the problem of DB access, so even if Cinder doesn\u0027t have these requirements right now, it may want to implement the indirection API in the future. It would  also be beneficial for all projects to have a singular implementation of this upgrade mechanism.\n\nThe configuration may be reloaded (https://github.com/openstack/oslo.service/blob/master/oslo_service/service.py#L490)\nwith a RPC broadcast, like Duncan proposed.\n\nConfiguration reloading without restarts was discussed yesterday:\nhttp://lists.openstack.org/pipermail/openstack-dev/2015-June/thread.html#67623\nbut it\u0027s really not an issue, since the version information should be cached in memory regardless of where the persistent configuration is placed (configuration or DB), so it can be set there (in memory), and the persistence layer can be changed later, when the upgrade is deemed successful.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":14250,"name":"Grzegorz Grasza","email":"xek@redhat.com","username":"xek"},"change_message_id":"0ed9df51ff23316ff00c3104c6a4486dead857d4","unresolved":false,"context_lines":[{"line_number":122,"context_line":"    Column(\u0027object_current_version\u0027, String(length\u003d36)),"},{"line_number":123,"context_line":"    Column(\u0027object_available_version\u0027, String(length\u003d36)),"},{"line_number":124,"context_line":"    mysql_engine\u003d\u0027InnoDB\u0027"},{"line_number":125,"context_line":")"},{"line_number":126,"context_line":""},{"line_number":127,"context_line":"The service_id is service name + host.  The *_current_version is the version"},{"line_number":128,"context_line":"a service is currently running and pinned at.  The *_available_version is the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_6fb796d8","line":125,"in_reply_to":"fa32b979_215528c1","updated":"2015-06-23 15:24:13.000000000","message":"As I stated above, the config reloading problem is not an issue. The config doesn\u0027t have to be reloaded. The version information should be cached in memory regardless of the storage method. I think storing this information in the DB sets a requirement, which can be troublesome in the future. (The requirement that each Cinder service has to have a direct connection to the DB.)","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":14250,"name":"Grzegorz Grasza","email":"xek@redhat.com","username":"xek"},"change_message_id":"6517811af5262700fce64bf861b731f235c3fe93","unresolved":false,"context_lines":[{"line_number":122,"context_line":"    Column(\u0027object_current_version\u0027, String(length\u003d36)),"},{"line_number":123,"context_line":"    Column(\u0027object_available_version\u0027, String(length\u003d36)),"},{"line_number":124,"context_line":"    mysql_engine\u003d\u0027InnoDB\u0027"},{"line_number":125,"context_line":")"},{"line_number":126,"context_line":""},{"line_number":127,"context_line":"The service_id is service name + host.  The *_current_version is the version"},{"line_number":128,"context_line":"a service is currently running and pinned at.  The *_available_version is the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_a3837255","line":125,"in_reply_to":"fa32b979_2f6aeeed","updated":"2015-06-23 17:29:28.000000000","message":"I may be wrong here, but I think it\u0027s safe to assume that the configuration is already synced to all hosts and we wouldn\u0027t be creating new issues this way. I\u0027m assuming this because the configuration options can change from version to version, so the upgrade process should take care of these changes.\n\nMaybe we could work on a common mechanism to cashe/store this information in memory? It could be read on start from the config or DB, and changed via a RPC broadcast.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"a93926d08565c008717863428d9398f5b61ca54a","unresolved":false,"context_lines":[{"line_number":122,"context_line":"    Column(\u0027object_current_version\u0027, String(length\u003d36)),"},{"line_number":123,"context_line":"    Column(\u0027object_available_version\u0027, String(length\u003d36)),"},{"line_number":124,"context_line":"    mysql_engine\u003d\u0027InnoDB\u0027"},{"line_number":125,"context_line":")"},{"line_number":126,"context_line":""},{"line_number":127,"context_line":"The service_id is service name + host.  The *_current_version is the version"},{"line_number":128,"context_line":"a service is currently running and pinned at.  The *_available_version is the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_2f6aeeed","line":125,"in_reply_to":"fa32b979_6fb796d8","updated":"2015-06-23 15:29:21.000000000","message":"If you cache it in memory, all info would be lost if the service/host crashes or restarts.  Each cinder service already accesses the DB in one way or another (indirectly).  Having the info distributed across possibly hundreds/thousands of hosts/services creates a syncing issue.  Having one source of \"truth\" is more ideal in my opinion.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"3e28d8c40feb0492dbb36700a6d2c7e7e6c5946d","unresolved":false,"context_lines":[{"line_number":122,"context_line":"    Column(\u0027object_current_version\u0027, String(length\u003d36)),"},{"line_number":123,"context_line":"    Column(\u0027object_available_version\u0027, String(length\u003d36)),"},{"line_number":124,"context_line":"    mysql_engine\u003d\u0027InnoDB\u0027"},{"line_number":125,"context_line":")"},{"line_number":126,"context_line":""},{"line_number":127,"context_line":"The service_id is service name + host.  The *_current_version is the version"},{"line_number":128,"context_line":"a service is currently running and pinned at.  The *_available_version is the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_215528c1","line":125,"in_reply_to":"fa32b979_9e85835a","updated":"2015-06-23 14:00:17.000000000","message":"Please see Duncan\u0027s comments to Michal because Michal brought up the same issue.\n\n\"The amount of changes we\u0027d need to make config reloading actually work are very significant, and are not, I think, going to happen anytime soon. I\u0027d be highly resistance to a \u0027reload config\u0027 option that only actually reloaded some of the config - that kind of thing is just begging for all sorts of breakage. Making config reload work is far more than just oslo.\"\n\nEvery project in OpenStack has different requirements.  It is difficult to justify the upgrade path for one project should apply to another.  Having an indirection API may be in the future as OpenStack evolves, but it would take a lot of refactor within cinder.  Using version pinning is a less intensive approach and solves the same issue.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":11600,"name":"Michał Dulko","email":"michal.dulko@gmail.com","username":"dulek"},"change_message_id":"a7ac3b64dde83523b32906461c920b5b073bb8c9","unresolved":false,"context_lines":[{"line_number":205,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":206,"context_line":""},{"line_number":207,"context_line":"* oslo_versionobjects for volumes, backups, service, consistency_group,"},{"line_number":208,"context_line":"quota need to be merged so that objects can be made backwards compatible."},{"line_number":209,"context_line":""},{"line_number":210,"context_line":""},{"line_number":211,"context_line":"Testing"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_8ec5753e","line":208,"updated":"2015-06-22 15:48:07.000000000","message":"Is that a complete list? We\u0027ll definitely need also cgsnapshots. What about request_spec and filter_properties used in scheduler rpcapi? We\u0027ll rely on using dictionaries?\n\nDo we need Quota object? I only see reservations in extend_volume (volume.rpcapi).","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"},{"author":{"_account_id":8247,"name":"Thang Pham","email":"thang.g.pham@gmail.com","username":"thang.pham"},"change_message_id":"0d6ebe97c586333dfc846632c77cc59ccae7a188","unresolved":false,"context_lines":[{"line_number":205,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":206,"context_line":""},{"line_number":207,"context_line":"* oslo_versionobjects for volumes, backups, service, consistency_group,"},{"line_number":208,"context_line":"quota need to be merged so that objects can be made backwards compatible."},{"line_number":209,"context_line":""},{"line_number":210,"context_line":""},{"line_number":211,"context_line":"Testing"}],"source_content_type":"text/x-rst","patch_set":4,"id":"fa32b979_3744bcb5","line":208,"in_reply_to":"fa32b979_8ec5753e","updated":"2015-06-22 16:29:16.000000000","message":"I consider consistency_group to include cgsnapshot.  quota object is listed above.\n\nAs far as the request_spec object, it depends on the state of taskflow going forward.  I am not sure if we really need to make it into an object, since it is more of a taskflow internal.  The rpc compat layer will be there in any case, so that the request_spec can be modified before going over RPC.","commit_id":"cb8ba74578f0c9c5947b3819bfd1f0c576d638bf"}]}
