[
https://issues.apache.org/jira/browse/CASSANDRA-20980?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dmitry Konstantinov updated CASSANDRA-20980:
--------------------------------------------
Reviewers: Dmitry Konstantinov
> GCInspector should use different thresholds on GC events for warning messages
> -----------------------------------------------------------------------------
>
> Key: CASSANDRA-20980
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20980
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Observability/Logging
> Reporter: Yuqi Yan
> Assignee: Yuqi Yan
> Priority: Normal
> Fix For: 5.x
>
>
> Running Cassandra 4.1 and recently testing with ShenandoahGC.
> Noticed that the log message is flooded with the warning message coming from
> GCInspector.
> Taking a closer look, that seems that we're warning on every GC event that
> takes longer than 1 second (default). And from document that the intention
> was to monitor GC pause time.
> But for ShenandoahGC and ZGC, there are concurrent phases (no STW) and phases
> that actually STW.
> For those concurrent phases, it's expected to take longer time (if we're on
> the non-generational version), and I feel that these phases:
> 1. require STW
> 2. doesn't require STW
> should use different thresholds in warning message.
> At least to me, phases require STW actually pause the thread, and should use
> lower threshold compared to those concurrent phases.
> Will share a patch soon
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
