lhotari commented on PR #24672:
URL: https://github.com/apache/pulsar/pull/24672#issuecomment-3226831723

   ```
   # Whether to allow brokers to forcefully load topics by skipping ledger 
failures to avoid topic unavailability and 
   # perform auto repairs of the topics.
   ```
   
   btw. The original description of this property isn't that great. It's 
missing that detail that the tradeoff of using this setting is that data is 
lost permanently.
   
   When there are ledger failures, there's some other underlying issue that is 
causing them. One potential issue that is being addressed is 
https://github.com/apache/pulsar/pull/24665 . My assumption is that some ledger 
failures, could be resolved by restarting the BookKeeper cluster or restarting 
Pulsar cluster when the failures are caused by invalid in-memory state.
   For 24665, there's a workaround to configure Pulsar cluster's BookKeeper 
client to by-pass the metadata store layer and use ZooKeeper directly. An the 
BookKeeper side to by-pass the Pulsar metadata store layer and use ZooKeeper 
directly. There's isn't yet a workaround when Oxia is used.
   
   There are also other issues that might cause ledger failures. For example, 
BookKeeper 4.17.2 (included in Pulsar 4.0.6 release) contained a fix for a data 
loss issue, https://github.com/apache/bookkeeper/pull/4607.


-- 
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]

Reply via email to