[
https://issues.apache.org/jira/browse/IGNITE-23289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17892401#comment-17892401
]
Kirill Tkalenko commented on IGNITE-23289:
------------------------------------------
Having discussed it with the author of
[IEP-128|https://cwiki.apache.org/confluence/display/IGNITE/IEP-128%3A+CMG+and+Metastorage+Disaster+Recovery]
[~rpuch], we decided that there is no need to stop the compaction during the
process of repairing the metastorage, it is expected that this will be a very
rare situation. If we send a command, it will either be saved in the raft log,
or not sent at all and should not break the logic of the compaction. Since in
the case of recording in the raft log, we will sooner or later apply the
command and notify about the completion of the local compaction and schedule a
new compaction. If there is an error during sending, we will simply re-schedule
the compaction.
> Stop the Metastorage compaction leader duty on disaster recovery
> ----------------------------------------------------------------
>
> Key: IGNITE-23289
> URL: https://issues.apache.org/jira/browse/IGNITE-23289
> Project: Ignite
> Issue Type: Improvement
> Reporter: Kirill Tkalenko
> Assignee: Kirill Tkalenko
> Priority: Major
> Labels: ignite-3
>
> Since the Metastorage leader acts as a compaction coordinator. This is
> another 'secondary duty' for a Metastorage leader. During Metastorage Group
> disaster recovery, we should pause secondary duties for some time (and then
> unpause them). So coordinating MS compaction should also be paused/unpaused.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
