Description
What is the issue?
When working on MB-50012 we noticed that a document that is backed up, deleted and then restored with the --force-updates flag is rejected by XDCR and does not appear on the replication cluster (as expected since its CAS value is lower than that of the tombstone because of how restore with the --force-updates flag works) but that rejection is not recorded in the XDCR stats UI (the 'XDCR Mutations Skipped' stat still shows 0).
Steps to reproduce:
- Setup two one node clusters, one with the backup service (Cluster A)
- Setup XDCR between the two clusters - single directional (Cluster A -> Cluster B)
- Add a bucket to the cluster A (the one with the backup service)
- Create a doc on cluster A, ensure it appears on cluster B
- Take a backup of cluster A
- Delete the document, again ensure it's also deleted on cluster B
- Do a restore using --force-update against cluster A, example command:
cbbackupmgr restore -a /home/janu/backups -r 6a833b37-a185-45c4-9f2a-eda7c4677fa7 -c http://127.0.0.1:9000 -u Administrator -p asdasd --force-updates
- Observe that the Cluster A will have the document and the Cluster B will not (this is the issue described in
MB-50012) - Navigate to the XDCR stats UI and see that the 'XDCR Mutations Skipped' stat still shows 0, even though one mutation should had been clearly rejected
The screenshot of the stats UI after following the above steps on my end:
The screenshot of the Documents page on the replication target cluster after the above steps:
Logs from both clusters are attached.
Attachments
Issue Links
- relates to
-
MB-50152 Need to add docs_failed_cr_target to XDCR Mutations Skipped stats info
- Closed