Details
-
Improvement
-
Resolution: Fixed
-
Critical
-
3.1.0, 3.0.5
-
Security Level: Public
-
None
-
CBG Sprint 126
-
1
Description
This is broken in 3.0.5. This was found by Yuvraj Kanwar where a customer has bad data in their dcp checkpoint and it'll repeatedlyshow:
2023-03-19T09:56:50.470495-04:00 WARNING 14083: (clinicalsolutions) DCP (Producer) eq_dcpq:0da283d0-ff9a-4d66-814b-026d789124b1_mark - (vb:0) Stream request requires rollback to seqno:0 becau |
se purge seqno (69087) is greater than start seqno - could miss purged deletions. Client requested seqnos:{9845,93792} snapshot:{0,90747} uuid:57812914539535 |
2023-03-19T09:56:50.470928-04:00 INFO 14083: (No Engine) DCP (Producer) eq_dcpq:0da283d0-ff9a-4d66-814b-026d789124b1_mark - Removing connection [ {"ip":"10.160.217.41","port":46656} - {"ip":" |
10.160.217.21","port":11210} (<ud>sguser</ud>) ] |
2023-03-19T09:56:50.801914-04:00 INFO 14083: (clinicalsolutions) DCP (Producer) eq_dcpq:0da283d0-ff9a-4d66-814b-026d789124b1_mark - Destroying connection. Created 0 s ago. Paused 16 times, for 48 ms total. Details: {"Initializing": {"count":1, "duration":"29 us"},"ReadyListEmpty": {"count":15, "duration":"48 ms"}} |
https://github.com/couchbase/sync_gateway/blob/master/db/attachment_compaction.go#L559 is a bug and this is present in 3.0.5
The recommended fix given to Yuvraj Kanwar was to delete the dcp checkpoint docs found from:
Attachment Compaction: Starting new compaction run with compact ID: %q |
where {{%q will become uuid}}
sync:dcp_ck:{uuid}_mark
|
sync:dcp_ck:{uuid}_sweep
|