Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
Description
accepted epoch is first written to temporary file acceptedEpoch.tmp then this file is renamed to acceptedEpoch.
Failure, either because of exception or power-off, in renaming the acceptedEpoch.tmp file will cause server startup error with message "The current epoch, x, is older than the last zxid y"
To handle this scenario we should read accepted epoch from this temp file as well.
For more context, refer https://github.com/apache/zookeeper/pull/1109
Attachments
Issue Links
- relates to
-
ZOOKEEPER-1653 zookeeper fails to start because of inconsistent epoch
- Closed
- links to