[PATCH] dbwrap_ctdb: Treat empty records as non-existing

[Christian Ambach]

>> Can you show me the callstack where it happens? We should
>> prevent it within dbwrap_ctdb.c.
>
> Attached is the callstack when opening/closing/re-opening a file within
> a brief period of time (so vacuuming did not delete the record in the
> local TDB).
>
> However, this does not give a clue which path was taken through dbwrap.
> I think the main point is that this record needs to be local, so access
> goes through the ltdb shortcut. A fetch from a remote node would not
> have delivered the record.

Spend some more time debugging this and it seems that I was mistaken and
the issue that I described does not exist. Sorry for the noise.

Cheers,
Christian