[
https://issues.apache.org/jira/browse/IMPALA-11882?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17682242#comment-17682242
]
David Rorke commented on IMPALA-11882:
--------------------------------------
We have some per queue metrics already in the IO manager (visible via the Web
UI) for read and write sizes and latencies (including min/max and percentiles).
However these appear to be metrics aggregated for the lifetime of the impala
daemon so they're not very useful for monitoring changes in IO latency over
time. If we produced similar metrics reporting latencies over some recent time
interval (most recent N seconds where N is configurable) these would be more
useful for monitoring/charting and correlating against query activity.
> Add Storage Latency Metrics at Impala Server Level
> --------------------------------------------------
>
> Key: IMPALA-11882
> URL: https://issues.apache.org/jira/browse/IMPALA-11882
> Project: IMPALA
> Issue Type: Improvement
> Components: Backend, Perf Investigation
> Reporter: Manish Maheshwari
> Priority: Major
>
> Add Storage Latency Metrics at Impala Server Level to track if across queries
> if storage subsystem is seeing slow response especially from
> S3/ADLS/GCP/Ozone etc
> This would help determine performance issues and also recommend if we should
> add executors for better scan performance.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
