Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
2.0.0
-
None
Description
We have integration tests failing that create a blobstore and then fail to submit a topology. Digging into the logs, it looks like there is a nimbus thread that runs every 10 seconds and deletes blobstores if there is no associated topology. We're hitting a race between the two calls periodically.
A possible fix could be to save the timestamp when the blobstore AtomicOutputStream closes and give some grace period before cleaning up based on this timestamp.
Attachments
Issue Links
- links to