errose28 commented on PR #11751: URL: https://github.com/apache/ozone/pull/11751#issuecomment-6425133843
> The main question is when should we start guaranteeing the rolling upgrade compatibility? Since the released versions have no ZDU, Right, we don't actually need to enforce ZDU compatibility until 2.3.0 goes out. However, this issue can actually occur in non-rolling upgrades of SCM as well. It happens when a ratis request is written to the log in the old version, but applied with the new version. In the release before ZDU, OM still has prepare for upgrade to handle this. SCM never had a prepare operation which leads to this issue. > Hopefully [HDDS-17253](https://issues.apache.org/jira/browse/HDDS-17253) and [HDDS-17252](https://issues.apache.org/jira/browse/HDDS-17252) would be able to catch this fragile mechanism in the future or we can pursue the SCM leader side execution like OM. I'm pretty sure this class of issues can actually be caught with a unit test. I filed HDDS-17487 with the details. I also moved all the compatibility detection tasks to a dedicated epic: HDDS-17484. Long term I agree that leader execution is the real solution and we are just coping with existing fragility in the request processing path for OM and SCM. -- 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]
