Details
-
Bug
-
Resolution: Fixed
-
Major
-
Cheshire-Cat
-
Untriaged
-
Centos 64-bit
-
1
-
No
Description
Rebalancing out of assigned node of an index during schedule stage make index stuck on reassigned node.
Steps to reproduce:-
- Have 3 nodes cluster (index-kv-query, index, index) with 1B-1S-1C and 1M docs.
- Create multiple indexes in parallel (say 10) with 1 replica num and preassigned deploy node.
CREATE INDEX idx_7 ON default:test_bucket.test_scope_1.test_collection_1(suffix) USING GSI WITH {'nodes': ['10.112.205.102:8091', '10.112.205.103:8091'], 'defer_build': True}
- Rebalance out one of the assigned node before indexes moved out of schedule state. (Also observed similar behaviour if one of the assigned node is failed over)
- Observer that all the indexes get reassigned to other available index node in cluster and comes out of schedule state.