[
https://issues.apache.org/jira/browse/HDDS-7728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17765351#comment-17765351
]
Ethan Rose commented on HDDS-7728:
----------------------------------
I agree with Stephen. The replication manager already has a lot of selection
criteria when deciding which containers to copy or remove. Adding even one more
criteria increases the complexity by an order of magnitude because it must be
weighed against every other existing selection criteria. For example, a
container goes unhealthy, but it has the highest delete transaction ID. The
other healthy replicas have a lower delete transaction ID and the block delete
was already acked much earlier and is not saved in SCM anymore. Both options
are bad.
I think orphan container cleanup is much easier than orphan block cleanup and
solves most of the same problems in this case. If we are correctly executing
most of the block deletes in most cases, then a few orphan blocks may prevent
the container from being deleted when there is no namespace data mapping to it.
I think some work in Recon has already been done in this area as part of the
OM/SCM DB insights. If Recon can inform SCM of a container that has no
namespace data, then SCM can force delete it, even if a few orphan blocks
remain. This would delay orphan block cleanup until a container is empty, but I
think this is an acceptable trade-off since there is no feasible proposal for
orphan block detection right now. This also handles the case where all replicas
are lost or missing but they will still show up as alerts in Recon even if
there is no key data.
> Block should be safely deleted from the containers if they are instructed
> from OM and containers are in missing state.
> ----------------------------------------------------------------------------------------------------------------------
>
> Key: HDDS-7728
> URL: https://issues.apache.org/jira/browse/HDDS-7728
> Project: Apache Ozone
> Issue Type: Improvement
> Components: SCM
> Affects Versions: 1.3.0
> Reporter: Uma Maheswara Rao G
> Assignee: Ashish Kumar
> Priority: Major
>
> Currently when OM instructs to delete the blocks and if containers are in
> missing state, deletion may not be processed properly. This Jira to track
> this requirement and implement to safe deletion os blocks what ever state
> they are on. Otherwise containers would never get cleaned up even though all
> blocks in that files deleted.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
