[GitHub] [kafka] dengziming opened a new pull request #10255: KAFKA-12388: Share broker channel between alterIsrManager and lifecycleManager

[GitBox <gi...@apache.org>]

dengziming opened a new pull request #10255:
URL: https://github.com/apache/kafka/pull/10255


   *More detailed description of your change*
   There are some `BorkerToControllerChannerManager` in `BrokerServer` and `KafkaServer`,
   We are planning to consolidate into two channels eventually:
   1. broker to controller channel
   2. client to controller channel
   
   Auto topic creation and forwarding fall into the 2nd category, while AlterIsr, lifecycleManager, and logDirEventManager(see KAFKA-9837) would be the 1st category.
   KAFKA-10348 is consolidating the 2nd category, this pr is trying to consolidate the 1st category.
   
   *Summary of testing strategy (including rationale)
   for the feature or bug fix. Unit and/or integration
   tests are expected for any behaviour change and
   system tests should be considered for larger changes.*
   
   ### Committer Checklist (excluded from commit message)
   - [ ] Verify design and implementation 
   - [ ] Verify test coverage and CI build status
   - [ ] Verify documentation (including upgrade notes)
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [kafka] cmccabe commented on pull request #10255: KAFKA-12388: Share broker channel between alterIsrManager and lifecycleManager

[GitBox <gi...@apache.org>]

cmccabe commented on pull request #10255:
URL: https://github.com/apache/kafka/pull/10255#issuecomment-844497580


   Thanks for looking at this. The issue that I see with sharing channels (really sockets) in this way is that we don't want the broker heartbeat to get delayed by something else. If the heartbeat gets delayed too long, the consequences could be really bad.
   
   I think what we should do here is get rid of head-of-line blocking for Kafka RPCs, so that they can be done in parallel. Really, the main reason why we still have head-of-line blocking is because of the difficulty of moving the producer and consumer off of that model. So we could get rid of HOL blocking for all requests except produce and consume without too many changes to the code, I think.
   
   I think we need to fix the HOL blocking problem before we start consolidating channels, to avoid regressions in functionality. cc @ijuma @hachikuji @abbccdda 


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [kafka] dengziming commented on pull request #10255: KAFKA-12388: Share broker channel between alterIsrManager and lifecycleManager

[GitBox <gi...@apache.org>]

dengziming commented on pull request #10255:
URL: https://github.com/apache/kafka/pull/10255#issuecomment-801567386


   @abbccdda @mumrah @hachikuji to have a look.😉


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org