yandrey321 commented on PR #11727: URL: https://github.com/apache/ozone/pull/11727#issuecomment-6418053711
> Could we also extend the integration test through the remaining recovery steps? Currently, waiting for the container to leave `OPEN` proves that allocation stops, and reading the existing key verifies availability with surviving fragments. It does not establish that closure completes or reconstruction occurs. > > Waiting for SCM's container state to reach `CLOSED`, then verifying restoration of all required EC replica indexes, would cover those claims. The test may need a spare DataNode to satisfy reconstruction placement constraints. Done. The integration test now waits for SCM to report the container CLOSED, and then until every RS(3,2) index (1–5) has a CLOSED replica again, before reading the key a second time. The cluster has 6 datanodes, so the reconstruction target doesn't have to be the datanode that lost the replica. The close-container wait and the replication manager and report intervals are shortened to keep the test at about 20 s. In the test log the replication manager sends a reconstructECContainersCommand and the datanode completes it about 5 s after the volume failure. -- 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]
