Details
-
Bug
-
Resolution: Fixed
-
Critical
-
7.0.0, 7.0.1, 7.0.2, 7.0.3, 7.0.4, 7.1.4, 7.0.5, 7.1.0, 7.1.1, 7.1.2, 7.2.0, 7.1.3, 7.2.1, 7.1.5, 7.2.2, 7.2.3
-
Untriaged
-
0
-
Unknown
-
KV 2023-4
Description
The following sequence of events will trigger DCP errors
- seqno:1 mutate(k1) -> replicate over DCP
- seqno:2 prepare(k1) -> replicate over DCP
- ephemeral auto delete triggers
The first two events will leave a committed k1 and prepare k1 in the active hash-table. When auto delete triggers it will use the hash-table visitor. The hash-table visitor will see all of the items, ignoring the prepare but will see the committed item at seqno 1, this is candidate for deletion.
DCP effectively sends
- seqno:1 mutate(k1)
- seqno:2 prepare(k1)
- seqno:3 delete(k1) ... a delete of seqno:1
At this point seqno:3 triggers an error and disconnects DCP.
Attachments
Issue Links
- relates to
-
MB-60105 A more efficient implementation of the fix for MB-60046
- Open