[
https://issues.apache.org/jira/browse/HBASE-29171?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eungsop Yoo updated HBASE-29171:
--------------------------------
Description:
filteredReadRequestsCount region metric counts the read requests of records
which is deleted, TTL expired and filtered all together. But we need the new
count metric for the read requests of records which is deleted and TTL expired
only. It can be used to distinguish the regions to be major compacted. Major
compaction on these regions can mitigate the performance degradation of reading
the deleted or TTL expired records.
Additionally, I applied a bug fix that do not increase readRequstsCount when
the result is empty.
was:filteredReadRequestsCount region metric counts the read requests of
records which is deleted, TTL expired and filtered all together. But we need
the new count metric for the read requests of records which is deleted and TTL
expired only. It can be used to distinguish the regions to be major compacted.
Major compaction on these regions can mitigate the performance degradation of
reading the deleted or TTL expired records.
> Add a new region metric of read requests count on deleted or TTL expired
> records
> --------------------------------------------------------------------------------
>
> Key: HBASE-29171
> URL: https://issues.apache.org/jira/browse/HBASE-29171
> Project: HBase
> Issue Type: Improvement
> Components: metrics
> Reporter: Eungsop Yoo
> Assignee: Eungsop Yoo
> Priority: Major
> Labels: pull-request-available
>
> filteredReadRequestsCount region metric counts the read requests of records
> which is deleted, TTL expired and filtered all together. But we need the new
> count metric for the read requests of records which is deleted and TTL
> expired only. It can be used to distinguish the regions to be major
> compacted. Major compaction on these regions can mitigate the performance
> degradation of reading the deleted or TTL expired records.
> Additionally, I applied a bug fix that do not increase readRequstsCount when
> the result is empty.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
