Details
-
Bug
-
Resolution: Fixed
-
Critical
-
5.1.1
-
None
-
Untriaged
-
No
Description
Planner fails to place repaired replica on resource constrained nodes. With the similar set of resource constraints, the rebalance out works correctly and is able to place the index on same nodes.
Steps of reproduce:
1. COUCHBASE_NUM_VBUCKETS=64 ./cluster_run -n4
2. ./cluster_connect -n 4 -T n0:kv+n1ql,n1:index,n2:index,n3:index
3. ./cbworkloadgen -n localhost:9000 -b bucket1 -j -u Administrator -p asdasd -i 300000
4. Create indexes with 1 replica till Ram quota is exhausted.
5. Failover 1 index node and rebalance out. The lost replicas don't get repaired as planner is unable to find the right placement.
2018-08-27T18:05:14.511-07:00 [Info] SatisfyClusterConstraint [5434818501388593063:17542529750046802955 6897107842609752352:12936802422695152065]
|
2018-08-27T18:05:14.511-07:00 [Info] SatisfyIndexHAConstraintAt 5434818501388593063:17542529750046802955 5434818501388593063:11757409155596068012
|
2018-08-27T18:05:14.511-07:00 [Info] SatisfyClusterConstraint Failed 127.0.0.1:9001
|
|
MemoryQuota: 268435456
|
CpuQuota: 8
|
--- Violations for index <b1i2 (replica 1), bucket1> (mem 96.1592M, cpu 0) at node 127.0.0.1:9001
|
Can move to 127.0.0.1:9002: NoViolation (free mem 1.67772e+07T, free cpu 7.925)
|
--- Violations for index <b1i4, bucket1> (mem 98.7384M, cpu 0) at node 127.0.0.1:9002
|
Can move to 127.0.0.1:9001: NoViolation (free mem 1.67772e+07T, free cpu 7.927927927927928)
|
If in Step5, node is rebalanced out, the indexes on the node get correctly placed on the remaining nodes in the cluster.
Attachments
Issue Links
- relates to
-
-
- Closed
-