Details
-
Bug
-
Resolution: Fixed
-
Critical
-
4.0.0
-
Security Level: Public
-
None
-
centOS 6.x
-
Untriaged
-
-
No
Description
Build
-------
3.5.0-1270
Testcase
-------------
test#37 in http://qa.sc.couchbase.com/view/BY%20COMPONENT/job/cen006-p0-xxdcr-vsetST-02-goxdcr-rebalance/18/consoleFull
./testrunner -i INI_FILE.ini get-cbcollect-info=True,get-logs=False,stop-on-failure=False,replication_type=xmem,enable_goxdcr=True,checkpoint_interval=3600,GROUP=ALL -t xdcr.rebalanceXDCR.Rebalance.swap_rebalance_replication_with_view_queries_and_ops,items=10000,update=C1-C2,delete=C1-C2,rdirection=bidirection,rebalance=C1,poll_timeout=300,GROUP=P0;xmem
Steps:
1. C1 [.51,.47] <--> C2 [.48, .49]
2. Load 10K keys on C1, C2.
3. Perform updates and deletes on C1 and C2.
4. Swap rebalance .47.on C1 with .50
5. Verify keys on both clusters
C1:
[2015-02-18 00:19:38,136] - [task:487] WARNING - Not Ready: vb_replica_curr_items 14003 == 14000 expected on '172.23.105.51:8091''172.23.105.50:8091', default bucket
[2015-02-18 00:19:42,173] - [task:487] WARNING - Not Ready: vb_active_curr_items 8746 == 14000 expected on '172.23.105.51:8091''172.23.105.50:8091', default bucket
C2:
[2015-02-18 00:19:47,474] - [task:491] INFO - Saw vb_active_curr_items 14000 == 14000 expected on '172.23.105.48:8091''172.23.105.49:8091',default bucket
[2015-02-18 00:19:48,689] - [task:491] INFO - Saw vb_replica_curr_items 14000 == 14000 expected on '172.23.105.48:8091''172.23.105.49:8091',default bucket
To investigate
-------------------
What explains the drop in active_curr_items count(and not replica) after swap rebalance?
Even if there is a drop, why is C2 unable to replicate missing keys to C1?
Logs attached.