[
https://issues.apache.org/jira/browse/HDDS-8784?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18013192#comment-18013192
]
Ivan Andika commented on HDDS-8784:
-----------------------------------
Updated this to critical since we encountered issue where the compaction takes
longer than the
disk.check.timeout during very high load due to replication and causes all
volumes to be down and DN to shutdown.
> Move RocksDB compaction out of disk checker
> -------------------------------------------
>
> Key: HDDS-8784
> URL: https://issues.apache.org/jira/browse/HDDS-8784
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Ethan Rose
> Assignee: GuoHao
> Priority: Critical
> Labels: pull-request-available
> Fix For: 2.0.0
>
>
> Currently RocksDB compaction is triggered for schema v3 RocksDBs (one DB per
> volume) as part of the volume check. Periodic manual compaction is necessary
> because import of sst files as part of container replication can lead to many
> small sst files in RocksDB.
> This operation may slow down the volume check and can have unintended
> consequences if the minimum gap between volume checks is set very high or
> very low. Ideally compaction should be triggered on its own independently
> controlled thread.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
