Details
-
Bug
-
Resolution: Fixed
-
Major
-
Cheshire-Cat
-
Untriaged
-
-
1
-
Unknown
Description
Build: 7.0.0-a 5127
- Create the following cluster:
sg1: {kv}, {fts}, {fts}, {fts}
|
sg2: {kv}, {fts}, {fts}, {fts}
|
sg3: {kv}, {fts}, {fts}, {fts}
|
- Create bucket `default`, load 1000 docs having the same `type` field value
- Create fts index, everything is default, replicas: 2, partitions: 12
Expected index partitions distribution through server groups is: 12 partitions per server group.
Found distribution is:
sg1: 13 partitions
sg2: 10 partitions
sg3: 13 partitions.
This distribution can lead to incorrect search results:
- shutdown couchbase service on all fts nodes from sg1 and sg3 to simulate 2 racks shutdown
- failover all unreachable fts nodes
- run fts query that should return 1000 docs normally against fts nodes from sg2
- you'll find less than 1000 results
Current cluster layout is here:
