[
https://issues.apache.org/jira/browse/IMPALA-14267?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18054265#comment-18054265
]
ASF subversion and git services commented on IMPALA-14267:
----------------------------------------------------------
Commit 96a3b09964ed945409cc7ec3d423c969166d57b7 in impala's branch
refs/heads/master from Yida Wu
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=96a3b0996 ]
IMPALA-14267: Reduce log frequency when dequeuing fails
When the admission controller fails to dequeue queries, like when at
capacity, it logs a message on every retry loop. This can cause
excessive logs, like ten per second until queuing timeout.
This patch reduces the log frequency to once every 5 seconds using
KLOG_EVERY_N_SECS. This prevents too many logs while keeping the
message visible when admission control is at capacity.
This also changes the dequeuing fails log level from VLOG(1) to INFO
because the macro uses a standard log level, but this is safe because
the log rate is under control now.
Testing:
Verified logs are throttled to 5s intervals when the queue is full.
Change-Id: I682b4416a2255f5be6480a7db8835f6aa54477bb
Reviewed-on: http://gerrit.cloudera.org:8080/23881
Reviewed-by: Michael Smith <[email protected]>
Tested-by: Impala Public Jenkins <[email protected]>
> Excessive logging when failing to dequeue queries
> -------------------------------------------------
>
> Key: IMPALA-14267
> URL: https://issues.apache.org/jira/browse/IMPALA-14267
> Project: IMPALA
> Issue Type: Improvement
> Components: Backend
> Reporter: Yida Wu
> Priority: Major
> Labels: admission-control
>
> When the admissiond is at capacity, often due to long-running queries or
> queries that cannot be canceled properly, newly queued queries may not be
> dequeued. In such cases, excessive log messages are repeatedly printed each
> time dequeuing fails. This can result in a large volume of redundant logs
> that provide limited diagnostic value.
> It would be better to improve this by reducing the frequency of these logs,
> especially when the condition could last a long time.
> {code:java}
> I20250723 10:01:56.808945 3705 admission-controller.cc:2679] Could not
> dequeue query id=7c4b6e233a47c50a:6fad07ef00000000 reason: number of running
> queries 1 is at or over limit 1.
> I20250723 10:01:56.908238 3705 admission-controller.cc:2679] Could not
> dequeue query id=7c4b6e233a47c50a:6fad07ef00000000 reason: number of running
> queries 1 is at or over limit 1.
> ...
> I20250723 10:02:00.016685 3705 admission-controller.cc:2679] Could not
> dequeue query id=7c4b6e233a47c50a:6fad07ef00000000 reason: number of running
> queries 1 is at or over limit 1.
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
