errose28 commented on PR #9790:
URL: https://github.com/apache/ozone/pull/9790#issuecomment-4171157308
I've looked at this functionality in the past and there's a lot of cases
this doesn't cover. Off the top of my head:
- There is no handling for container + block delete on SCM side, so this
will result in verbose infinite deletion retry loops for empty containers,
over-replicated data, and deleted blocks.
- This will pin empty containers in the system on other nodes because the
replica on the read-only volume cannot have all blocks deleted and all replicas
must be empty before removing.
- No handling of volume selection within datanodes so this volume could
still be chosen for writes.
- If we are assuming that the used space tracking on the volume would
prevent this then there is no need for this feature in the first place. It only
exists as a fallback if used space tracking becomes inaccurate.
- With no way to delete data (since that involves appending to the DB WAL),
there is no way out of this state.
I would prefer closing this PR and raising a doc instead if we want to try
to design this, but it adds a fair amount of complexity.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
