Details
-
Bug
-
Resolution: User Error
-
Major
-
5.5.0
-
Untriaged
-
-
No
Description
1. Create a 9 nodes vulcan cluster with following configuration
- Group 1: 2 KV nodes and 1 node with index, FTS and query services
- Group 2: 2 KV nodes and 1 node with index, FTS and query services
- Group 3: 2 KV nodes and 1 node with index, FTS and query services
2. Enable autofailover with server groups on the cluster
3. Introduce failures in group 3 by stopping the couchbase servers
4. Wait for failover of nodes
The KV nodes in the group are failed over as expected. But the non-kv node is not failed over along with the KV nodes. The logs does say trying to failover server group 3 but only with the KV nodes. According to design docs, the non-kv node must be failed over automatically too since the cluster is well balanced.
The above steps can be reproduced by running the following system test.
1. Start a job on http://qa.sc.couchbase.com/job/component_systest_launcher/ with following parameters
- Version number: Latest build or latest weekly build
- test: -test tests/rebalance/test_servergroup_failover.yml -scope tests/rebalance/scope_servergroup_failover.yml
- Provider: file:centos_component.yml
- duration: 3600
Keep the rest of the parameters as default. Please contact me if you need to start a run on the system test boxes to reproduce the issue again for live debugging.