ayushtkn opened a new pull request, #11976: URL: https://github.com/apache/ozone/pull/11976
## What changes were proposed in this pull request? The open key cleanup scan reads `openKeyTable` from RocksDB only, so an open key whose first hsync is still in the table cache is submitted for `DeleteOpenKeys` as a plain expired key. `OMOpenKeysDeleteRequest` then checks only existence and `updateID` — both of which pass, since the hsync's `updateID` is lower than the cleanup transaction's — so it deletes the open key and queues the blocks the committed key references for deletion, and the next overwrite commit dereferences the missing open entry while marking it `OVERWRITTEN_HSYNC_KEY`, throwing `NullPointerException` in apply and terminating the OM. The delete request now skips an open key whose committed key still carries that writer's `HSYNC_CLIENT_ID`, and both commit requests tolerate a missing open entry so a cluster already holding such a key recovers instead of terminating. ## What is the link to the Apache JIRA https://issues.apache.org/jira/browse/HDDS-16764 ## How was this patch tested? Unit tests, `TestOMOpenKeysDeleteRequest#testDeleteHsyncedKeys` (both bucket layouts) and `TestOMKeyCommitRequest#testOverwriteHsyncKeyWithoutOpenKey` (inherited by the FSO variant). -- 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]
