Details
-
Bug
-
Resolution: Fixed
-
Critical
-
4.0.0
-
Security Level: Public
-
Untriaged
-
No
-
KV: Dec 7 - Dec 25
Description
Build: 4.0.0-4047 (also seen on 3.1.0)
OS: CentOS 6 but not believed to be unique to it
Test case:
./testrunner -i /Users/eric/servers/tunable-vms.ini -t rebalance.rebalancein.RebalanceInTests.rebalance_in_with_ops,nodes_in=1,replicas=1,items=500,doc_ops=create,value_size=1024,active_resident_threshold=90,dgm_run=true,eviction_policy=fullEviction,max_verify=1000
After the rebalance the cbstats for curr_items and vb_replica_curr_items add up to about 2 more than they should:
2015-08-24 14:56:28 | WARNING | MainProcess | Cluster_Thread | [task.check] Not Ready saw: vb_replica_curr_items 1000751 == 1000749 and expected on '172.23.106.24:8091''172.23.106.25:8091', default bucket
2015-08-24 14:56:28 | WARNING | MainProcess | Cluster_Thread | [task.check] Not Ready saw: curr_items_tot 2001500 == 2001498 and expected on '172.23.106.24:8091''172.23.106.25:8091', default bucket
Note that if I do value eviction the problem does not occur.
Server logs and test logs are attached. The test logs contain debug output that I added for my local environment.
Please let me know if there is more information I can provide.