Details
-
Bug
-
Resolution: Fixed
-
Critical
-
3.0
-
Security Level: Public
-
None
-
Untriaged
-
3
-
Unknown
-
June 30 - July 18
Description
I'm getting responses from stream request to rollback seqno 0 when failover log from stats indicates requested seqno should exist. Also verified that when I rollback to seqno 0, the next by_seqno is in fact the start_seqno I requested in original request and so I shouldn't have needed to resend the stream request.
This behavior is from Incremental rebalance in Test:
1) add node2, load 100 items, stream_req 100 mutations
2) add node3, load 100 items, stream_req 100 mutations
3) add node4, load 100 items, stream_req 100 mutations
5) remove node2, load 100 items, stream_req 100 mutations
6) AssertionError: Got unexpected response to rollback to: (0, 0), but start_seqno: 247
- during stream request I'm forced to roll back to seqno 0 even though failover log indicates last known seq for this vbucket is 247. expected to be able to stream mutations from step 5
repro:
https://github.com/couchbaselabs/pyupr/blob/master/unit.py#L1407
./pyupr -h 172.23.106.53:8091 10.6.2.118:8091 172.23.106.55:8091 172.23.106.56:8091 -b cb -s reb