Details
-
Bug
-
Resolution: Done
-
Critical
-
4.1.0
-
None
-
Untriaged
-
Unknown
Description
During incremental backup I noticed that the sequence numbers between backups appear to decrease at some point.
We run 3 backups with the cbbackup command. The UUID in each backup is summarized below.
MVSzYtxfLhjCKvaJ 2016-05-11T08:41:31.495097-04:00 Full
|
CYAuyVZskPRTGmKO 2016-05-12T09:00:22.090586-04:00 Incr
|
nPUOhIoTcuSGHkej 2016-05-12T09:09:16.736108-04:00 Full
|
The user then restores the first full and incremental backup to one cluster. The last full backup is restored to a different cluster. There is an item mismatch of 66 items.
If we look at VBucket 207 we can see two of the items missing. Below is how each stream is connected.
MVSzYtxfLhjCKvaJ (Requests 0 to 66)(Last Seqno Sent is 66)(62 items sent)
|
CYAuyVZskPRTGmKO 2016-05-12T09:00:22.090586-04:00 (Fails due to invalid seqno request of 66 to 64)
|
nPUOhIoTcuSGHkej (Requests 0 to 64)(Last Seqno Sent is 64)(64 items sent)
|
Two items were missed and the reason appears to be due to the fact that the max sequence number for some of the vbuckets decreased. What is interesting however is that there is no sign of a failover or restart of the server. All failover logs for all vbuckets on the server only have a single entry.
vb_207:0:id: 53541242870147
|
vb_207:0:seq: 0
|