[
https://issues.apache.org/jira/browse/HDFS-17909?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18078866#comment-18078866
]
ASF GitHub Bot commented on HDFS-17909:
---------------------------------------
kokonguyen191 commented on PR #8448:
URL: https://github.com/apache/hadoop/pull/8448#issuecomment-4395500039
Hi @hfutatzhanghb @KeeProMise, I also thought about using
`DFS_ROUTER_ASYNC_RPC_MAX_ASYNCCALL_PERMIT_KEY` as an alternative control
mechanic but there were 2 things that didn't really sit well with me:
- Unless DFS_ROUTER_FAIRNESS_ACQUIRE_TIMEOUT = 0 (which kinda removes a big
chunk of the purpose of the fairness controller), there is a wait time for
permit acquisition, and the async handler queue can still get flooded. Though,
even with zero wait time, it's still possible to flood the queue, just more
difficult.
- I think the fairness controller shouldn't be the OOM preventive solution
since that's not its intended usage.
> [ARR] AsyncRouterHandlerExecutors should use bounded queue
> ----------------------------------------------------------
>
> Key: HDFS-17909
> URL: https://issues.apache.org/jira/browse/HDFS-17909
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: rbf
> Reporter: Felix N
> Assignee: Felix N
> Priority: Major
> Labels: pull-request-available
>
> AsyncRouterHandlerExecutors should use bounded queues to prevent queues from
> growing without a limit.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
