[
https://issues.apache.org/jira/browse/HDDS-9452?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Duong updated HDDS-9452:
------------------------
Description:
After locks stripping optimization, it's noticeable that the way we capture
"lockHeldTime" metric using ThreadLocal is not performant. This is serious
considering locks are acquired and released many times per any operation.
!lockMetrics.png|width=854,height=181!
We should consider another efficient way to capture the metrics, for example,
look at how ReentrantReadWriteLock maintains the refCount.
cc. [~tanvipenumudy] [~ritesh]
was:
After locks stripping optimization, it's noticeable that the way we capture
"lockHeldTime" metric using ThreadLocal is not performant.
!lockMetrics.png|width=854,height=181!
We should consider another efficient way to capture the metrics, for example,
look at how ReentrantReadWriteLock maintains the refCount.
> OzoneManagerLock: lock metrics capture is too costly
> ----------------------------------------------------
>
> Key: HDDS-9452
> URL: https://issues.apache.org/jira/browse/HDDS-9452
> Project: Apache Ozone
> Issue Type: Improvement
> Reporter: Duong
> Priority: Major
> Attachments: lockMetrics.png
>
>
> After locks stripping optimization, it's noticeable that the way we capture
> "lockHeldTime" metric using ThreadLocal is not performant. This is serious
> considering locks are acquired and released many times per any operation.
> !lockMetrics.png|width=854,height=181!
> We should consider another efficient way to capture the metrics, for example,
> look at how ReentrantReadWriteLock maintains the refCount.
> cc. [~tanvipenumudy] [~ritesh]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
