swamirishi commented on PR #7418: URL: https://github.com/apache/ozone/pull/7418#issuecomment-2479280813
> > Technically we don't need to block reads. Writes is what can cause a corruption. Do you think we should also block read requests? > > I agree that, technically, only writes to the Ratis container can cause corruption. However, my understanding is that container commands sent via gRPC shouldn’t cause corruption due to the datanodeID either. > > Would it be clearer to validate only replicaIndex rather than datanodeID for EC container? Alternatively, are there any scenarios where the datanodeID could still potentially lead to corruption in EC containers? This is actually follow up task for https://github.com/apache/ozone/pull/7402 . The uber goal here is to validate the datanodeId on writes so that we don't end up writing partial containers when data volumes go down and a node comes back on the same host with all volumes being completely wiped off. In such a case we should prevent all inflight writes on the node since the datanodeId has changed(If we allow writes we could end up with containers having only a subset of the blocks it is supposed to have). As of date ratis doesn't perform this check, eventually we would love to have this peerId check on ratis as well. https://issues.apache.org/jira/browse/RATIS-2187 -- 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]
