[
https://issues.apache.org/jira/browse/IGNITE-8090?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16610303#comment-16610303
]
Alexey Goncharuk commented on IGNITE-8090:
------------------------------------------
[~Artem Budnikov], I am not sure if users need this information, but here are
some details that may be added to the compaction section:
- when compaction is enabled, not only does Ignite compress WAL segments, but
it also gets rid of the physical records needed for node recovery after crash.
After compaction, WAL segments contain only logical records needed for
rebalancing.
- as for the WAL tuning, in some workloads it may be beneficial to increase
the WAL segment size from the default value (64MB) to larger size (up to 2GB)
because segment switch involves some synchronization and it may affect
throughput and latency percentiles. Increasing WAL segment size reduces the WAL
segment switch frequency. The drawback is that we preallocate a certain number
of segments in the work directory and increasing segment size will increase the
size of the storage
Otherwise looks good.
> Explain disk space compaction by WAL and compaction techniques
> --------------------------------------------------------------
>
> Key: IGNITE-8090
> URL: https://issues.apache.org/jira/browse/IGNITE-8090
> Project: Ignite
> Issue Type: New Feature
> Components: documentation
> Reporter: Denis Magda
> Assignee: Artem Budnikov
> Priority: Major
> Fix For: 2.7
>
>
> Explain how WAL occupies the disk space and how to enable the compaction:
> https://cwiki.apache.org/confluence/display/IGNITE/Ignite+Persistent+Store+-+under+the+hood#IgnitePersistentStore-underthehood-WALHistorySize
> There needs to be a special section in WAL documentation as well as in the
> capacity planning.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
