sodonnel opened a new pull request, #3850: URL: https://github.com/apache/ozone/pull/3850
## What changes were proposed in this pull request? A datanode is stopped and before the stale node handler is triggered, SCM is restarted. When SCM restarts its loads all the only pipelines and nodes from RocksDB, and then all the nodes will register again. In the case of EC pipelines, there is nothing to trigger the close of a pipeline (and the containers on it) except: 1. The Container getting full and the DN triggering the close 2. The stale / dead node handlers noticing a node on it has gone dead. In the case above, the EC pipeline will sit forever in an Open state, but any attempt to write to it will likely result in errors on the client due to one of the nodes not being available. These errors still will not trigger it to close. A solution to this problem, is to add logic to the pipeline scrubber to close any pipelines that have unregistered nodes. Stale / Dead nodes should be handled by the existing stale / dead node handlers. ## What is the link to the Apache JIRA https://issues.apache.org/jira/browse/HDDS-7341 ## How was this patch tested? New unit test and validated on a docker compose cluster. -- 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]
