Description
1. As of now, when starting stream requests, xdcr always sets snapStartSeqno to 0. As a result, whenever a rollback happens, it always rollback to 0.
The right thing to do is to track the latest snapStartSeqno received from DCP, save it to checkpoint doc, and use it to set snapStartSeqno the next time replication resumes.
2. There is an issue with getFailoverUUIDForSeqno(), which could cause a 0 vbuuid to be recorded in checkpoint and included in start stream request. This also needs to be fixed.
Attachments
Issue Links
- relates to
-
MB-22224 XDCR does not set snapStartSeqno and vb uuid correctly when starting stream request
- Closed