)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":13362,"name":"Tan Lin","email":"lin.tan@intel.com","username":"zero"},"change_message_id":"fcd445b6bdc2dec52b4e3fa4fb9787d8b82aa45b","unresolved":false,"context_lines":[{"line_number":5,"context_line":"CommitDate: 2015-12-02 13:28:01 +0800"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"Use cached copy of hash ring in api service"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Change-Id: I195d66487cea5ad9db9cb42a8e57957302d4bfc9"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"7a740942_e69a3414","line":8,"updated":"2015-12-04 10:50:34.000000000","message":"You should change the name of rst file as well:)","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"a45f1e8993d959ba634bfe088cfa83d2a9a202bd","unresolved":false,"context_lines":[{"line_number":5,"context_line":"CommitDate: 2015-12-02 13:28:01 +0800"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"Use cached copy of hash ring in api service"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Change-Id: I195d66487cea5ad9db9cb42a8e57957302d4bfc9"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"7a740942_56264f69","line":8,"in_reply_to":"7a740942_e69a3414","updated":"2015-12-04 14:48:04.000000000","message":"yeah, but filename must match with the bp name \u0027improve-api-performance\u0027.","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"}],"specs/approved/improve-api-performance.rst":[{"author":{"_account_id":2889,"name":"Aeva Black","email":"aeva.online@gmail.com","username":"tenbrae"},"change_message_id":"d8406ba0c86cb3d10068e2662eeee90fa3f141bb","unresolved":false,"context_lines":[{"line_number":5,"context_line":" http://creativecommons.org/licenses/by/3.0/legalcode"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":8,"context_line":"Improve Ironic API performance"},{"line_number":9,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":10,"context_line":""},{"line_number":11,"context_line":"https://blueprints.launchpad.net/ironic/+spec/improve-api-performance"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a8ffd7b_42e89b7c","line":8,"updated":"2015-12-02 01:47:49.000000000","message":"title should be less generic. instead of \"improve performance\", how about \"use cached copy of hash ring in api service\"","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"b4f5022f309b4257c16039bd03f47c777675dbb9","unresolved":false,"context_lines":[{"line_number":5,"context_line":" http://creativecommons.org/licenses/by/3.0/legalcode"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":8,"context_line":"Improve Ironic API performance"},{"line_number":9,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":10,"context_line":""},{"line_number":11,"context_line":"https://blueprints.launchpad.net/ironic/+spec/improve-api-performance"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a8ffd7b_05785dc3","line":8,"in_reply_to":"9a8ffd7b_42e89b7c","updated":"2015-12-02 02:47:31.000000000","message":"will update soon.","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":13362,"name":"Tan Lin","email":"lin.tan@intel.com","username":"zero"},"change_message_id":"7a426085c86f38f32187f5e142bcf9af276149b1","unresolved":false,"context_lines":[{"line_number":32,"context_line":"* Add a new parameter to ``get_topic_for()`` which defatul to False, If"},{"line_number":33,"context_line":"  set to True, then refreshing the cached hash ring."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Wrapped try-except to every rpc api, and catching the Timeout exception."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a8ffd7b_227317e9","line":35,"updated":"2015-12-02 01:41:05.000000000","message":"I have a concern about this. Because oslo messaging has a default timeout by 60s. If you want to catch the Timeout exception, then the ironic service will hang on 60s. This is very bad for API performance. I doubt this is a necessary case to catch Timeout exception.","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"5cbeb981a9ccc9967ada48d1c8d21dcf5e6c316b","unresolved":false,"context_lines":[{"line_number":32,"context_line":"* Add a new parameter to ``get_topic_for()`` which defatul to False, If"},{"line_number":33,"context_line":"  set to True, then refreshing the cached hash ring."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Wrapped try-except to every rpc api, and catching the Timeout exception."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a740942_bb043e15","line":35,"in_reply_to":"7a740942_e3f30672","updated":"2015-12-04 13:56:45.000000000","message":"run get_topic on every request could be accurate but will cause low performance, please reference the problem description of this spec, and api will only hang 60s when conductor hash changed but the cached hash ring haven\u0027t refreshed yet.","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"b4f5022f309b4257c16039bd03f47c777675dbb9","unresolved":false,"context_lines":[{"line_number":32,"context_line":"* Add a new parameter to ``get_topic_for()`` which defatul to False, If"},{"line_number":33,"context_line":"  set to True, then refreshing the cached hash ring."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Wrapped try-except to every rpc api, and catching the Timeout exception."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a8ffd7b_45fe656c","line":35,"in_reply_to":"9a8ffd7b_227317e9","updated":"2015-12-02 02:47:31.000000000","message":"hi Tan, even if you don\u0027t catch the Timeout exception, the ironic service will hang on 60s, and as there is a logic to refresh the cached hash ring now, the api request will be mostly accurate, and together with this try/except block will ensure the api request succeed if we have other active conductors.","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":13362,"name":"Tan Lin","email":"lin.tan@intel.com","username":"zero"},"change_message_id":"b1e5b2d559e50758867e71cfe6629041380b44ce","unresolved":false,"context_lines":[{"line_number":32,"context_line":"* Add a new parameter to ``get_topic_for()`` which defatul to False, If"},{"line_number":33,"context_line":"  set to True, then refreshing the cached hash ring."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Wrapped try-except to every rpc api, and catching the Timeout exception."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7a740942_e3f30672","line":35,"in_reply_to":"9a8ffd7b_45fe656c","updated":"2015-12-04 10:43:31.000000000","message":"Yes, it will still hang on 60s, but if you run get_topic and find the conductor is lost before sent request, it could partially avoid this?","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":2889,"name":"Aeva Black","email":"aeva.online@gmail.com","username":"tenbrae"},"change_message_id":"d8406ba0c86cb3d10068e2662eeee90fa3f141bb","unresolved":false,"context_lines":[{"line_number":41,"context_line":""},{"line_number":42,"context_line":"Alternatives"},{"line_number":43,"context_line":"------------"},{"line_number":44,"context_line":"None"},{"line_number":45,"context_line":""},{"line_number":46,"context_line":"Data model impact"},{"line_number":47,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a8ffd7b_e518b977","line":44,"updated":"2015-12-02 01:47:49.000000000","message":"there are several alternatives, but mostly, I\u0027d like to call out that RPC timeout is a very poor cache validation test. We could, for example, use a DLM for that.","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":2889,"name":"Aeva Black","email":"aeva.online@gmail.com","username":"tenbrae"},"change_message_id":"d8406ba0c86cb3d10068e2662eeee90fa3f141bb","unresolved":false,"context_lines":[{"line_number":60,"context_line":"None"},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"RPC API impact"},{"line_number":63,"context_line":"--------------"},{"line_number":64,"context_line":"None"},{"line_number":65,"context_line":""},{"line_number":66,"context_line":"Driver API impact"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a8ffd7b_a5d3915e","line":63,"updated":"2015-12-02 01:47:49.000000000","message":"there\u0027s an increased risk of split-brain problems if the connection between API servers and conductors gets interrupted, since each API server may perform cache invalidation / cache regeneration at different times.\n\nNote that a conductor, upon receiving an RPC request, *does not* validate its hash ring to ensure that the Node is assigned to it -- it trusts the API servers to route correctly.\n\nWe should correct that as part of this work, and introduce a new exception, NodeNotMappedHere, which would be caught by rpcapi.py and trigger the API service to refresh its hash ring.","commit_id":"1e865d70e0629b12f1edcfba8d0b453a98ebda6a"},{"author":{"_account_id":6773,"name":"Lucas Alvares Gomes","email":"lucasagomes@gmail.com","username":"lucasagomes"},"change_message_id":"da90e773c80f0679a2847a7937e300cd4438d416","unresolved":false,"context_lines":[{"line_number":29,"context_line":"refreshing the hash ring, and fetching a new active topic node to issuing"},{"line_number":30,"context_line":"the request again."},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"* Add a new parameter to ``get_topic_for()`` which defatul to False, If"},{"line_number":33,"context_line":"  set to True, then refreshing the cached hash ring."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Wrapped try-except to every rpc api, and catching the Timeout exception."}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_ba12ffb2","line":32,"updated":"2015-12-04 12:17:14.000000000","message":"default","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6773,"name":"Lucas Alvares Gomes","email":"lucasagomes@gmail.com","username":"lucasagomes"},"change_message_id":"da90e773c80f0679a2847a7937e300cd4438d416","unresolved":false,"context_lines":[{"line_number":32,"context_line":"* Add a new parameter to ``get_topic_for()`` which defatul to False, If"},{"line_number":33,"context_line":"  set to True, then refreshing the cached hash ring."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Wrapped try-except to every rpc api, and catching the Timeout exception."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_3ab12f90","line":35,"updated":"2015-12-04 12:17:14.000000000","message":"How long it takes to the timeout to be raised? Is it a reasonable time ?","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"a45f1e8993d959ba634bfe088cfa83d2a9a202bd","unresolved":false,"context_lines":[{"line_number":32,"context_line":"* Add a new parameter to ``get_topic_for()`` which defatul to False, If"},{"line_number":33,"context_line":"  set to True, then refreshing the cached hash ring."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Wrapped try-except to every rpc api, and catching the Timeout exception."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_31a14d95","line":35,"in_reply_to":"7a740942_3ab12f90","updated":"2015-12-04 14:48:04.000000000","message":"that could be configured by rpc_response_timeout which is default to 60s.","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6773,"name":"Lucas Alvares Gomes","email":"lucasagomes@gmail.com","username":"lucasagomes"},"change_message_id":"da90e773c80f0679a2847a7937e300cd4438d416","unresolved":false,"context_lines":[{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"},{"line_number":39,"context_line":"  issuing the request to the new conductor again."},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":42,"context_line":"between API servers and conductors gets interrupted, since each API server"},{"line_number":43,"context_line":"may perform cache invalidation/cache regeneration at different times. A"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_3a8a8fc0","line":40,"updated":"2015-12-04 12:17:14.000000000","message":"It would be useful if we could add a configuration option to keep the behavior we have today and force resetting the hash ring on all the requests","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"a45f1e8993d959ba634bfe088cfa83d2a9a202bd","unresolved":false,"context_lines":[{"line_number":37,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":38,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"},{"line_number":39,"context_line":"  issuing the request to the new conductor again."},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":42,"context_line":"between API servers and conductors gets interrupted, since each API server"},{"line_number":43,"context_line":"may perform cache invalidation/cache regeneration at different times. A"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_aca082f5","line":40,"in_reply_to":"7a740942_3a8a8fc0","updated":"2015-12-04 14:48:04.000000000","message":"hi Lucas, maybe we can use hash_ring_reset_interval for that, when it set to 0, we will reset the hash ring on every request, but the option also affect the cached hash ring in conductor.","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":7711,"name":"Yuriy Zveryanskyy","email":"yzveryanskyy@mirantis.com","username":"yuriyz"},"change_message_id":"dfa00e411455b4a8c6edb1e409e9aef70f1227ca","unresolved":false,"context_lines":[{"line_number":40,"context_line":""},{"line_number":41,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":42,"context_line":"between API servers and conductors gets interrupted, since each API server"},{"line_number":43,"context_line":"may perform cache invalidation/cache regeneration at different times. A"},{"line_number":44,"context_line":"conductor does not validate it\u0027s hash ring to ensure that the Node is assigned"},{"line_number":45,"context_line":"to it, so we should corret it."},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"9a8ffd7b_92054d7b","line":43,"updated":"2015-12-02 15:09:03.000000000","message":"If I understand logic described here correctly in some\nsituations requests to the same node can be processed\non different conductors, example:\n 1) Conductor1 with works with node foo fails\n 2) API1 got timeout for node foo\n 3) API1 refresh hash ring, Conductor2 returned for foo\n 4) Conductor2 refresh our hash ring quickly, no exception is raised\n 5) Conductor1 quickly recovered\n 6) API2 does not get timeout from Conductor1 for foo\n\nNow for some time period requests for foo from API1 will\nbe served on Conductor2 and from API2 on Conductor1.","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":7711,"name":"Yuriy Zveryanskyy","email":"yzveryanskyy@mirantis.com","username":"yuriyz"},"change_message_id":"dfa00e411455b4a8c6edb1e409e9aef70f1227ca","unresolved":false,"context_lines":[{"line_number":41,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":42,"context_line":"between API servers and conductors gets interrupted, since each API server"},{"line_number":43,"context_line":"may perform cache invalidation/cache regeneration at different times. A"},{"line_number":44,"context_line":"conductor does not validate it\u0027s hash ring to ensure that the Node is assigned"},{"line_number":45,"context_line":"to it, so we should corret it."},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"* Introduce a new exception NodeNotMappedHere, which would be caught by"}],"source_content_type":"text/x-rst","patch_set":2,"id":"9a8ffd7b_81642122","line":44,"updated":"2015-12-02 15:09:03.000000000","message":"Please provide more info about changes on conductor side.","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"a45f1e8993d959ba634bfe088cfa83d2a9a202bd","unresolved":false,"context_lines":[{"line_number":41,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":42,"context_line":"between API servers and conductors gets interrupted, since each API server"},{"line_number":43,"context_line":"may perform cache invalidation/cache regeneration at different times. A"},{"line_number":44,"context_line":"conductor does not validate it\u0027s hash ring to ensure that the Node is assigned"},{"line_number":45,"context_line":"to it, so we should corret it."},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"* Introduce a new exception NodeNotMappedHere, which would be caught by"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_d1aa4919","line":44,"in_reply_to":"9a8ffd7b_81642122","updated":"2015-12-04 14:48:04.000000000","message":"will enrich it in the next patch set.","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":7711,"name":"Yuriy Zveryanskyy","email":"yzveryanskyy@mirantis.com","username":"yuriyz"},"change_message_id":"dfa00e411455b4a8c6edb1e409e9aef70f1227ca","unresolved":false,"context_lines":[{"line_number":42,"context_line":"between API servers and conductors gets interrupted, since each API server"},{"line_number":43,"context_line":"may perform cache invalidation/cache regeneration at different times. A"},{"line_number":44,"context_line":"conductor does not validate it\u0027s hash ring to ensure that the Node is assigned"},{"line_number":45,"context_line":"to it, so we should corret it."},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"* Introduce a new exception NodeNotMappedHere, which would be caught by"},{"line_number":48,"context_line":"  rpcapi.py and trigger the API service to refresh its hash ring, then"}],"source_content_type":"text/x-rst","patch_set":2,"id":"9a8ffd7b_c1e229a0","line":45,"updated":"2015-12-02 15:09:03.000000000","message":"s/corret/correct","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":13362,"name":"Tan Lin","email":"lin.tan@intel.com","username":"zero"},"change_message_id":"fcd445b6bdc2dec52b4e3fa4fb9787d8b82aa45b","unresolved":false,"context_lines":[{"line_number":44,"context_line":"conductor does not validate it\u0027s hash ring to ensure that the Node is assigned"},{"line_number":45,"context_line":"to it, so we should corret it."},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"* Introduce a new exception NodeNotMappedHere, which would be caught by"},{"line_number":48,"context_line":"  rpcapi.py and trigger the API service to refresh its hash ring, then"},{"line_number":49,"context_line":"  issuing the request to the right conductor again."},{"line_number":50,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_0682e094","line":47,"updated":"2015-12-04 10:50:34.000000000","message":"nit:s/NodeNotMappedHere/``NodeNotMappedHere``","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":13362,"name":"Tan Lin","email":"lin.tan@intel.com","username":"zero"},"change_message_id":"fcd445b6bdc2dec52b4e3fa4fb9787d8b82aa45b","unresolved":false,"context_lines":[{"line_number":52,"context_line":"Alternatives"},{"line_number":53,"context_line":"------------"},{"line_number":54,"context_line":"None"},{"line_number":55,"context_line":""},{"line_number":56,"context_line":"Data model impact"},{"line_number":57,"context_line":"-----------------"},{"line_number":58,"context_line":"None"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_03abb21e","line":55,"updated":"2015-12-04 10:50:34.000000000","message":"As Devananda said, DLM should be a better alternative to mention..","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6773,"name":"Lucas Alvares Gomes","email":"lucasagomes@gmail.com","username":"lucasagomes"},"change_message_id":"da90e773c80f0679a2847a7937e300cd4438d416","unresolved":false,"context_lines":[{"line_number":52,"context_line":"Alternatives"},{"line_number":53,"context_line":"------------"},{"line_number":54,"context_line":"None"},{"line_number":55,"context_line":""},{"line_number":56,"context_line":"Data model impact"},{"line_number":57,"context_line":"-----------------"},{"line_number":58,"context_line":"None"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_5ad99bf2","line":55,"in_reply_to":"7a740942_03abb21e","updated":"2015-12-04 12:17:14.000000000","message":"++\n\nI think the DLM is the way to go","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"a45f1e8993d959ba634bfe088cfa83d2a9a202bd","unresolved":false,"context_lines":[{"line_number":52,"context_line":"Alternatives"},{"line_number":53,"context_line":"------------"},{"line_number":54,"context_line":"None"},{"line_number":55,"context_line":""},{"line_number":56,"context_line":"Data model impact"},{"line_number":57,"context_line":"-----------------"},{"line_number":58,"context_line":"None"}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_16276713","line":55,"in_reply_to":"7a740942_5ad99bf2","updated":"2015-12-04 14:48:04.000000000","message":"thanks, will add in the next patch set.","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":7711,"name":"Yuriy Zveryanskyy","email":"yzveryanskyy@mirantis.com","username":"yuriyz"},"change_message_id":"dfa00e411455b4a8c6edb1e409e9aef70f1227ca","unresolved":false,"context_lines":[{"line_number":121,"context_line":"  hash ring refresh logic under the refresh condition."},{"line_number":122,"context_line":""},{"line_number":123,"context_line":"* Introduce a new exception ``NodeNotMappedHere``, and add a conductor"},{"line_number":124,"context_line":"  hash ring valiadtion upon receiving an RPC request."},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"* Catching rpcapi request Timout and NodeNotMappedHere exceptions, if"},{"line_number":127,"context_line":"  it is, then fetching the active corresponding conductor from db, and"}],"source_content_type":"text/x-rst","patch_set":2,"id":"9a8ffd7b_e1ec0dbd","line":124,"updated":"2015-12-02 15:09:03.000000000","message":"s/valiadtion/validation","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":6773,"name":"Lucas Alvares Gomes","email":"lucasagomes@gmail.com","username":"lucasagomes"},"change_message_id":"da90e773c80f0679a2847a7937e300cd4438d416","unresolved":false,"context_lines":[{"line_number":123,"context_line":"* Introduce a new exception ``NodeNotMappedHere``, and add a conductor"},{"line_number":124,"context_line":"  hash ring valiadtion upon receiving an RPC request."},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"* Catching rpcapi request Timout and NodeNotMappedHere exceptions, if"},{"line_number":127,"context_line":"  it is, then fetching the active corresponding conductor from db, and"},{"line_number":128,"context_line":"  issuing the request again."},{"line_number":129,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"7a740942_dac10be6","line":126,"updated":"2015-12-04 12:17:14.000000000","message":"Timeout","commit_id":"0b5840a092202ff808265ad01176bcdcdda34f09"},{"author":{"_account_id":10343,"name":"Jim Rollenhagen","email":"jim@jimrollenhagen.com","username":"jimrollenhagen"},"change_message_id":"dc21fa94aceff775db93fa76988aefd05dc2c56b","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"Currently, the API service resets hash ring on every request, which would"},{"line_number":20,"context_line":"be accurate but cause low performance in production environment, it is"},{"line_number":21,"context_line":"unacceptable."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Proposed change"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa69d971_d2124d58","line":21,"updated":"2016-01-08 02:03:26.000000000","message":"I\u0027d like to see some benchmarks here. We know it isn\u0027t optimal, but I wouldn\u0027t call it \"unacceptable\".","commit_id":"0473df66676aa7ac35b4067cc86d496bb146aa25"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"37870ba92afbe17f0d2ed45d2f2957d8556c74ba","unresolved":false,"context_lines":[{"line_number":18,"context_line":""},{"line_number":19,"context_line":"Currently, the API service resets hash ring on every request, which would"},{"line_number":20,"context_line":"be accurate but cause low performance in production environment, it is"},{"line_number":21,"context_line":"unacceptable."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Proposed change"},{"line_number":24,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa69d971_ba5d2fee","line":21,"in_reply_to":"fa69d971_d2124d58","updated":"2016-01-11 13:09:56.000000000","message":"yes Jim, maybe \"unacceptable\" is not suitable here.","commit_id":"0473df66676aa7ac35b4067cc86d496bb146aa25"},{"author":{"_account_id":10343,"name":"Jim Rollenhagen","email":"jim@jimrollenhagen.com","username":"jimrollenhagen"},"change_message_id":"dc21fa94aceff775db93fa76988aefd05dc2c56b","unresolved":false,"context_lines":[{"line_number":38,"context_line":""},{"line_number":39,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":40,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"},{"line_number":41,"context_line":"  issuing the request to the new conductor again."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":44,"context_line":"between API servers and conductors gets interrupted, since each API server"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa69d971_b79b0b72","line":41,"updated":"2016-01-08 02:03:26.000000000","message":"So if a conductor goes down, all API calls that should route to that conductor will hang until the first call times out and things are re-balanced? Seems like this could get backed up quickly.","commit_id":"0473df66676aa7ac35b4067cc86d496bb146aa25"},{"author":{"_account_id":10342,"name":"Jay Faulkner","display_name":"JayF","email":"jay@jvf.cc","username":"JayF","status":"youtube.com/@oss-gr / podcast.gr-oss.io"},"change_message_id":"7180c1a9cdc9801e5b8be9d60801ecfe8218f8f5","unresolved":false,"context_lines":[{"line_number":38,"context_line":""},{"line_number":39,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":40,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"},{"line_number":41,"context_line":"  issuing the request to the new conductor again."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":44,"context_line":"between API servers and conductors gets interrupted, since each API server"}],"source_content_type":"text/x-rst","patch_set":5,"id":"dab17558_fb8f1673","line":41,"in_reply_to":"fa69d971_7071a40b","updated":"2016-05-12 21:40:12.000000000","message":"I\u0027m also very concerned about the failure semantics around this.","commit_id":"0473df66676aa7ac35b4067cc86d496bb146aa25"},{"author":{"_account_id":6610,"name":"Zhenguo Niu","email":"Niu.ZGlinux@gmail.com","username":"niu-zglinux"},"change_message_id":"37870ba92afbe17f0d2ed45d2f2957d8556c74ba","unresolved":false,"context_lines":[{"line_number":38,"context_line":""},{"line_number":39,"context_line":"* If got Timeout exception first time, we will call ``get_topic_for()`` with"},{"line_number":40,"context_line":"  the refresh parameter to force fetching active conductors from the db, then"},{"line_number":41,"context_line":"  issuing the request to the new conductor again."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"But there\u0027s an increased risk of split-brain problems if the connection"},{"line_number":44,"context_line":"between API servers and conductors gets interrupted, since each API server"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa69d971_7071a40b","line":41,"in_reply_to":"fa69d971_b79b0b72","updated":"2016-01-11 13:09:56.000000000","message":"yes, and this will only happen when the conductor down and the cached hash ring hasn\u0027t reach the refresh time.","commit_id":"0473df66676aa7ac35b4067cc86d496bb146aa25"}]}
