Details
-
Bug
-
Resolution: Fixed
-
Critical
-
7.1.0
-
7.1.0-1480-enterprise
-
Untriaged
-
-
1
-
Unknown
Description
- Step 1: Create a 9 node cluster
- Step 1.1: Create a 5 node XDCR remote cluster
- Step 2: Create required buckets and 20 collections.
- Step 2.1: Create required buckets and 10 collections on XDCR remote.
- Step 3: Create 1000000 items sequentially
- Step 4: Update 1000000 RandonKey keys to create 50 percent fragmentation
- Step 5: Create 1000000 items sequentially
- Step 6: Update 1000000 RandonKey keys to create 50 percent fragmentation
- Step 7: Create full bucket to bucket replications for 10 collections
- Step 8: Start an async load: CRUD all 25% at a rate of total 20k ops.
- Step 9: Rebalance in with Loading of docs. Kill and resume rebalance at 20%, 40%
- Found the mentioned WARNING in memcahced logs
172.23.121.126
Found 'Stream request failed because the snap start seqno' logs - ['2021-10-14T04:23:20.860344-07:00 WARNING 6760: (GleamBookUsers0) DCP (Producer) eq_dcpq:xdcr:dcp_backfill_b67b789bcfd95e6b97c8af1e8fa5c7cb/GleamBookUsers0/GleamBookUsers0_172.23.121.126:11210_0:zp3EZ30z9hs-XalilQLKUQ== - (vb:271) Stream request failed because the snap start seqno (0) <= start seqno (31608) <= snap end seqno (0) is required\n']
Another observations is that on an idle state cluster the RAM usage is quite high:
On checking one of the node on Source cluster both goxdcr and memcahced seems to be using pretty high RAM:
NOTE: After the test finished/failed at the last step, total Items in source cluster are 37,326,445 while items in destination is 20,507,898. While there are constants ops seen on the destinations cluster the item counts is not changing in the destination cluster.
Attachments
Issue Links
- is caused by
-
MB-48938 XDCR - Backfill Pipeline non-Zero Starting Timestamp is set with snapStart+snapEnd of 0
- Closed