[
https://issues.apache.org/jira/browse/IGNITE-23920?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-23920:
---------------------------------------
Description: By default, RocksDB instances flush their changes to disk on
shutdown, and we don't disable this for the storage of the Metastorage. But
sometimes it can be seen that, on a start, a node replays Metastorage commands
that were already executed before, even if the node was stopped gracefully (and
not crashed). This means that either the storage was not flushed to disk on
shutdown, or that applied index (also stored in Metastorage's storage) does not
agree with the flushed data (this doesn't seem to be probable).
> Metastorage is sometimes not flushed to disk on graceful node stop
> ------------------------------------------------------------------
>
> Key: IGNITE-23920
> URL: https://issues.apache.org/jira/browse/IGNITE-23920
> Project: Ignite
> Issue Type: Bug
> Reporter: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
>
> By default, RocksDB instances flush their changes to disk on shutdown, and we
> don't disable this for the storage of the Metastorage. But sometimes it can
> be seen that, on a start, a node replays Metastorage commands that were
> already executed before, even if the node was stopped gracefully (and not
> crashed). This means that either the storage was not flushed to disk on
> shutdown, or that applied index (also stored in Metastorage's storage) does
> not agree with the flushed data (this doesn't seem to be probable).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
