[
https://issues.apache.org/jira/browse/HDDS-7516?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17637159#comment-17637159
]
Devesh Kumar Singh commented on HDDS-7516:
------------------------------------------
GetDB Updates at SCM may not solve the purpose unless until SCM DB is flushed
when Ratis snapshot is taken. So for now, below approach has been taken for
HDDS-3486 :
# Make paginated RPC calls to SCM for List container API depending on how
many containers we can bring in one call, decide this based on rpc call buffer
limit by checking container info object size.
# Mark the containers as "UNKNOWN" state which are not present on Recon . This
will be transient state for sometime.
3. Later if no FCRs or ICR will come from DN for those UNKNOWN
containers, they will be treated as MISSING as ContainerHealthTask will not
find any replica of those containers.
Since listContainer API is now paginated and called for CLOSED containers only,
this will not put load on SCM.
> Implement SCM getDBUpdates endpoint
> -----------------------------------
>
> Key: HDDS-7516
> URL: https://issues.apache.org/jira/browse/HDDS-7516
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Siyao Meng
> Priority: Major
>
> We need this endpoint on SCM as well to efficiently get delta DB updates
> which [OM is already capable
> of|https://github.com/apache/ozone/blob/39dfe7ad8638b4258f3759f077cb6fe985f93402/hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OzoneManager.java#L4138-L4159].
> This could make HDDS-3486 much, much more efficient.
> cc [~devmadhu]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
