[jira] [Commented] (KAFKA-5503) Idempotent producer ignores shutdown while fetching ProducerId

["Evgeny Veretennikov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KAFKA-5503?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16074714#comment-16074714 ] 

Evgeny Veretennikov commented on KAFKA-5503:
--------------------------------------------

[~hachikuji], part of call stack for blocking idempotent producer is:

{noformat}
Selector.poll(long): L321
NetworkClient.poll(long, long): L433
NetworkClientUtils.sendAndReceive(KafkaClient, ClientRequest, Time): L89
Sender.sendAndAwaitInitProducerIdRequest(): L405
Sender.maybeWaitForProducerId(): L419
Sender.run(long): L204
{noformat}

So, selector blocks thread. While we invoke {{initiateClose()}} method, part of call stack is:

{noformat}
Selector.wakeUp(): 240
NetworkClient.wakeUp(): L498
Sender.wakeup(): L675
Sender.initiateClose(): L390
{noformat}

So, that seems, that blocking selector will actually be waken up and thus sender thread won't be blocked right after concurrent {{initiateClose()}} call.

So, is this issue still relevant? Do I miss something?

> Idempotent producer ignores shutdown while fetching ProducerId
> --------------------------------------------------------------
>
>                 Key: KAFKA-5503
>                 URL: https://issues.apache.org/jira/browse/KAFKA-5503
>             Project: Kafka
>          Issue Type: Sub-task
>          Components: clients, core, producer 
>    Affects Versions: 0.11.0.0
>            Reporter: Jason Gustafson
>            Assignee: Evgeny Veretennikov
>             Fix For: 0.11.0.1
>
>
> When using the idempotent producer, we initially block the sender thread while we attempt to get the ProducerId. During this time, a concurrent call to close() will be ignored.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)