[
https://issues.apache.org/jira/browse/IGNITE-23411?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-23411:
---------------------------------------
Description:
When finishing recovering the Metastorage (that is, getting current revision
from the leader and applying all of them to the local state machine), we should
update the Metastorage Safe Time to the timestamp of the last applied revision.
Without this, Metastorage revision update listeners that are invoked on node
recovery (this happens after Metastorage recovery, but before the Metastorage
Safe Time starts being updated by new revisions and idle Safe Time propagation
mechanism) will see a zero Safe Time and not the one that corresponds to
already applied revisions. This might break the logic of those listeners.
> Recover Metastorage SafeTime on Metastorage recovery
> ----------------------------------------------------
>
> Key: IGNITE-23411
> URL: https://issues.apache.org/jira/browse/IGNITE-23411
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Assignee: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
> Time Spent: 10m
> Remaining Estimate: 0h
>
> When finishing recovering the Metastorage (that is, getting current revision
> from the leader and applying all of them to the local state machine), we
> should update the Metastorage Safe Time to the timestamp of the last applied
> revision.
> Without this, Metastorage revision update listeners that are invoked on node
> recovery (this happens after Metastorage recovery, but before the Metastorage
> Safe Time starts being updated by new revisions and idle Safe Time
> propagation mechanism) will see a zero Safe Time and not the one that
> corresponds to already applied revisions. This might break the logic of those
> listeners.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
