)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":7715,"name":"Hirofumi Ichihara","email":"ichihara.hirofumi@gmail.com","username":"Hirofumi-Ichihara"},"change_message_id":"1bebf093496f5e34f8316a9f3893ef9a26e2a697","unresolved":false,"context_lines":[{"line_number":14,"context_line":"This will be further optimized by running dedicated RPC workers"},{"line_number":15,"context_line":"for state reports processing."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"Closes-Bug: #1496410"},{"line_number":18,"context_line":"Change-Id: Id86a1f962aaa4f64011d57ae55d240f890cca4f7"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":4,"id":"ba15a1d1_fd132a63","line":17,"updated":"2015-09-24 05:04:28.000000000","message":"I guess that this change affects operation. Don\u0027t we need RFE tag in the bug or APIImpact in message?\n\nFuthermore, could you elaborate how do we get the benefit? If my understanding is right, it may be useless  when rpc_workers \u003d1, right?","commit_id":"b14a6ad03b10759ac98963f93def04bac9d889d8"},{"author":{"_account_id":7715,"name":"Hirofumi Ichihara","email":"ichihara.hirofumi@gmail.com","username":"Hirofumi-Ichihara"},"change_message_id":"38eec9f0846197c3d1ea152e506fa0c84d5c3db0","unresolved":false,"context_lines":[{"line_number":14,"context_line":"This will be further optimized by running dedicated RPC workers"},{"line_number":15,"context_line":"for state reports processing."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"Closes-Bug: #1496410"},{"line_number":18,"context_line":"Change-Id: Id86a1f962aaa4f64011d57ae55d240f890cca4f7"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":4,"id":"9a1a9d01_855a6082","line":17,"in_reply_to":"9a1a9d01_32887de7","updated":"2015-09-25 14:37:12.000000000","message":"I got it. Thanks.","commit_id":"b14a6ad03b10759ac98963f93def04bac9d889d8"},{"author":{"_account_id":6072,"name":"Eugene Nikanorov","email":"enikanorov@mirantis.com","username":"enikanorov"},"change_message_id":"8e58a526f9ea89ea9dbaaaa5e99fac3c77ab88c9","unresolved":false,"context_lines":[{"line_number":14,"context_line":"This will be further optimized by running dedicated RPC workers"},{"line_number":15,"context_line":"for state reports processing."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"Closes-Bug: #1496410"},{"line_number":18,"context_line":"Change-Id: Id86a1f962aaa4f64011d57ae55d240f890cca4f7"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":4,"id":"9a1a9d01_32887de7","line":17,"in_reply_to":"ba15a1d1_110605ce","updated":"2015-09-25 12:07:21.000000000","message":"When you have a bunch of greenthreads consuming from  2 queues, you\u0027ll have more even distribution of message consumption between the queues.\nAnd that\u0027s the point.","commit_id":"b14a6ad03b10759ac98963f93def04bac9d889d8"},{"author":{"_account_id":7715,"name":"Hirofumi Ichihara","email":"ichihara.hirofumi@gmail.com","username":"Hirofumi-Ichihara"},"change_message_id":"9be658c0ce40b8e2a4dea378731069833ec301c2","unresolved":false,"context_lines":[{"line_number":14,"context_line":"This will be further optimized by running dedicated RPC workers"},{"line_number":15,"context_line":"for state reports processing."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"Closes-Bug: #1496410"},{"line_number":18,"context_line":"Change-Id: Id86a1f962aaa4f64011d57ae55d240f890cca4f7"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":4,"id":"ba15a1d1_110605ce","line":17,"in_reply_to":"ba15a1d1_762e578d","updated":"2015-09-24 07:31:29.000000000","message":"I have heard that an operator uses tool to gather RPC info so that the tool warns RPC overflow and gives information to debug someting about error. However, that may be corner case so you can ignore my comment.\n\n\u003e even single process could process state reports sooner.\n\nWhy? I\u0027m not sure how is RPC queue accepted. Does RPC worker process alternately q-pugin and q-reports-plugin message not in order of arrival?\n\nI can understand this change is very effective if the dedicated RPC workers is implemented in future.","commit_id":"b14a6ad03b10759ac98963f93def04bac9d889d8"},{"author":{"_account_id":6072,"name":"Eugene Nikanorov","email":"enikanorov@mirantis.com","username":"enikanorov"},"change_message_id":"4c6b4c241328d70183ee335ae6033ac70a076336","unresolved":false,"context_lines":[{"line_number":14,"context_line":"This will be further optimized by running dedicated RPC workers"},{"line_number":15,"context_line":"for state reports processing."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"Closes-Bug: #1496410"},{"line_number":18,"context_line":"Change-Id: Id86a1f962aaa4f64011d57ae55d240f890cca4f7"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":4,"id":"ba15a1d1_762e578d","line":17,"in_reply_to":"ba15a1d1_fd132a63","updated":"2015-09-24 06:31:00.000000000","message":"How do you think it affects operation?\nAlso, I\u0027ve left state report consumer on q-plugin too for compatibility with any agents that could now be out of neutron code base.\n\nFor rpc_workers \u003d1 the benefit could still be noticeable, because even single process could process state reports sooner.\nWe\u0027ve been testing DVR recently and found that starting from certain scale/number of VMs the cloud begins to have problems with agents that start appearing/disappearing just because rpc_workers were not able to process RPC requests from agents fast enough, that in turn triggered rescheduling increasing amount of those RPC requests.\nSo in order to mitigate this, we launch dedicated RPC workers for state reports queue. That approach will be a continuation of this work.","commit_id":"b14a6ad03b10759ac98963f93def04bac9d889d8"},{"author":{"_account_id":11682,"name":"Ryan Moats","email":"rmoats@us.ibm.com","username":"regXboi"},"change_message_id":"07a118a85909b9b28f0a9f20112cfa60b7ddd860","unresolved":false,"context_lines":[{"line_number":8,"context_line":""},{"line_number":9,"context_line":"This optimization is needed for big clusters with hundreds"},{"line_number":10,"context_line":"of agents where the spike of activity may trigger a burst"},{"line_number":11,"context_line":"of RPC requests that would prevent neutorn-server from processing"},{"line_number":12,"context_line":"agent heart beats in time, triggering resource rescheduling."},{"line_number":13,"context_line":""},{"line_number":14,"context_line":"This will be further optimized by running dedicated RPC workers"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":5,"id":"9a1a9d01_bd2f3ef5","line":11,"updated":"2015-09-28 13:39:38.000000000","message":"nit: neutron","commit_id":"2989f1a3a3893640d7793220648869f6f954d894"},{"author":{"_account_id":6072,"name":"Eugene Nikanorov","email":"enikanorov@mirantis.com","username":"enikanorov"},"change_message_id":"f32dcb7714dc50fa45cd20372681b4aa514d226a","unresolved":false,"context_lines":[{"line_number":8,"context_line":""},{"line_number":9,"context_line":"This optimization is needed for big clusters with hundreds"},{"line_number":10,"context_line":"of agents where the spike of activity may trigger a burst"},{"line_number":11,"context_line":"of RPC requests that would prevent neutorn-server from processing"},{"line_number":12,"context_line":"agent heart beats in time, triggering resource rescheduling."},{"line_number":13,"context_line":""},{"line_number":14,"context_line":"This will be further optimized by running dedicated RPC workers"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":5,"id":"9a1a9d01_b8bd0c6a","line":11,"in_reply_to":"9a1a9d01_bd2f3ef5","updated":"2015-09-28 14:03:21.000000000","message":"oh. sure","commit_id":"2989f1a3a3893640d7793220648869f6f954d894"}]}