Details
-
Bug
-
Resolution: Fixed
-
Blocker
-
2.6.4
-
9 - Cataclysm, 10 - Path of Vengeance
-
4
Description
Currently the Operator uses the pools/default/tasks REST API to detect the outcome of graceful failover. In particular, if after a successful kick-off of graceful failover the tasks REST API reports that a rebalance with subtype gracefulFailover is "notRunning", the operation is deemed to have finished successfully.
Unfortunately, this API is not completely reliable. In particular, if graceful-failover abruptly terminates (e.g. the orchestrator crashes) the tasks API will show that the graceful failover is not running and the Operator will incorrectly conclude that it finished successfully.
Attachments
Issue Links
- relates to
-
MB-61888 The Failover and Graceful Failover REST APIs should provide a way to assert the state of the cluster
- Open
-
K8S-3446 Cloud Native 2.6.4 - Release Notes
- Closed
-
K8S-3493 [Operator 2.6.4-109] No progress during delta recovery upgrade (7.2.5 -> 7.6.1) because of "Reconciliation failed" in operator
- Closed
- mentioned in
-
Page Loading...
For Gerrit Dashboard: K8S-3472 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
210209,2 | K8S-3472: Improve graceful failover detection | 2.6.x | couchbase-operator | Status: MERGED | +2 | +1 |