[GitHub] [pulsar] zbentley commented on issue #12556: Permanent namespace corruption after attempting to delete a namespace shortly after it was created and after all contained topics were deleted; broker log of "MetadataNotFoundException: Managed ledger not found"

[GitBox <gi...@apache.org>]

zbentley commented on issue #12556:
URL: https://github.com/apache/pulsar/issues/12556#issuecomment-956856959


   > If that induces a permanent inconsistency state, it's a bug that absolutely needs to get fixed.
   
   This specific issue is a permanent inconsistency. The others usually go away after waiting a few minutes.
   
   > In reality, this is not an easy thing to do in any distributed system, barring of completely eliminating the caches.
   
   Is there some way instead to bypass cached metadata on a per-operation basis? Could that be exposed as a parameter for management API operations?
   
   Alternatively, would it be possible to add behavior like "when the backend encounters an error that could be due to an out-of-date cache, dump the cache and try once more against authoritative state"?
   
   I ask because the current workaround (add what amounts to large numbers of sleeps and retries around common management API and create-producer/subscribe interactions) is pretty costly both in runtime and complexity.


-- 
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: commits-unsubscribe@pulsar.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org