siddhantsangwan opened a new pull request, #4756: URL: https://github.com/apache/ozone/pull/4756
## What changes were proposed in this pull request? With EC containers, if there is a small cluster of say 6 nodes with EC-3-2, a container will require 5 nodes. If 2 containers become unhealthy, reconstruction will be required to recover the 2 containers, but there is only 1 spare node. This means one will get recovered, and we will have 4 "good" containers and 2 `UNHEALTHY` and the container will remain stuck like this because `UNHEALTHY` containers are only removed once the container has no over or under replication. A similar problem was resolved previously where an EC container with both over and under replication can meet the same problem, where under replication cannot proceed due to insufficient spare nodes. In that case, the solution was to check for this case, and call the over-replication handler to clear up the excess replicas. This PR is still in draft state for some early reviews while I write tests and think about edge cases. Here, we try to delete an `UNHEALTHY` replica in the same handler to free up a DN. Then we throw the exception so that this container gets queued again in the under replication queue. Perhaps it's better to throw first if over replication handling is invoked, so we don't delete multiple replicas in one go. ## What is the link to the Apache JIRA https://issues.apache.org/jira/browse/HDDS-8535 ## How was this patch tested? Wrote one UT. -- 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]
