F64116045 commented on PR #11199:
URL: https://github.com/apache/ozone/pull/11199#issuecomment-6040587684
Thinking about this some more, if we want checksum mismatch to remain
visible while other issues are being repaired, the main problem is that
representing every overlap with combined states does not scale well.
My thought for the longer term is to separate the primary state RM acts on
from the other issues found during the scan. RM would still act on one state,
while the report could keep all detected issues.
But I think this would require a larger change to the health model, so I
would leave it outside this PR.
For now, I would like to propose:
- Treat checksum mismatch like the other container health states.
- Keep one primary health state per container.
- Give repairable replication problems priority. Because RM only acts on the
primary state, so letting checksum mismatch take priority could prevent an
under-replicated or over-replicated container from being repaired.
- Report `DATA_CHECKSUM_MISMATCH` in a later scan after the replication
problem is resolved.
For scope:
- Keep this PR focused on checksum mismatch detection and reporting in RM.
- Move the RPC and `ozone admin container info` changes to a follow-up.
- Move the Recon changes to another follow-up.
@sodonnel @devmadhuu if you have time, I’d appreciate any thoughts on this
direction. Thanks again for the earlier discussion.
--
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]
