[
https://issues.apache.org/jira/browse/HDDS-8529?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Siyao Meng updated HDDS-8529:
-----------------------------
Description:
This is a follow-up to HDDS-8166.
This is required for OM consistency if we don't want OM to fail startup in the
scenario where:
1. OM previous had snapshot feature enabled
2. users created some snapshots
3. OM crashed and had some Ratis logs to be replayed (mainly
{{OMKeyPurgeRequest}} and {{OMDirectoriesPurgeRequestWithFSO}})
4. admin disabled snapshot feature on OM in ozone-site.xml () before starting
it back up
5. admin started OM -- OM would fail to start because snapshot feature is
disabled and omSnapshotManager is not initialized
Above is just one of the examples. Similar startup failure can also happen in
the case of OM HA where a follower has disabled snapshot feature but the leader
has snapshot feature enabled. Though in this HA case it can be argued that
there is a configuration mismatch between the leader and the follower(s). Q: Is
such conf mismatch being detected right now?
cc [~aswinshakil] [~hemantk]
was:
This is a follow-up to HDDS-8166.
TBA
> [Snapshot] Refactor OmSnapshotManager constructor to allow already committed
> Ratis Txs to apply even when snapshot feature is disabled
> --------------------------------------------------------------------------------------------------------------------------------------
>
> Key: HDDS-8529
> URL: https://issues.apache.org/jira/browse/HDDS-8529
> Project: Apache Ozone
> Issue Type: Sub-task
> Components: Snapshot
> Reporter: Siyao Meng
> Priority: Major
>
> This is a follow-up to HDDS-8166.
> This is required for OM consistency if we don't want OM to fail startup in
> the scenario where:
> 1. OM previous had snapshot feature enabled
> 2. users created some snapshots
> 3. OM crashed and had some Ratis logs to be replayed (mainly
> {{OMKeyPurgeRequest}} and {{OMDirectoriesPurgeRequestWithFSO}})
> 4. admin disabled snapshot feature on OM in ozone-site.xml () before starting
> it back up
> 5. admin started OM -- OM would fail to start because snapshot feature is
> disabled and omSnapshotManager is not initialized
> Above is just one of the examples. Similar startup failure can also happen in
> the case of OM HA where a follower has disabled snapshot feature but the
> leader has snapshot feature enabled. Though in this HA case it can be argued
> that there is a configuration mismatch between the leader and the
> follower(s). Q: Is such conf mismatch being detected right now?
> cc [~aswinshakil] [~hemantk]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
