[Bucardo-general] Replication problems? data overwritten?

Michelle Sullivan michelle at sorbs.net
Mon Aug 26 15:22:25 UTC 2013

eric machine wrote:
> Hi Michelle,
> Wow, I liked your proposed replication strategy, that's look good :)
> Thanks so so much :)
> > ALTER SEQUENCE $schema start with 1 INCREMENT BY 10;
> > ALTER SEQUENCE $schema start with 2 INCREMENT BY 10;
> ...
> By the way, few more questions to clarify further:-
> a)
> Because the status is only set after the first successful sync...
> >> sorry, I am not sure what do you mean by first successful sync?
> Since autokick is set to yes, shouldn't it be automatically "status =
> good" when it detects new changes (when insert new record?)?

When you start bucardo it doesn't always perform a sync (there might not
be any changes) so 'No records found' is the appropriate status.  With
Auto-Kick set when it kicks the sync (because of a detected change) the
sync will be performed.  The status will be updated at the *end* of the
sync, so it will show 'No records found' until the end of the first sync.

> b)
> I'm not running 5 yet - but from my understanding your config would
> indicate multimaster between dbbranch1, dbbranch2 and dbbranch3 with the
> dbhq being a slave.
> >> Yes that's correct. But for HQ, I can always create new records but
> it will never get replicated to the branches. Just I still can't
> figure out why this will happen?
> dbbranch2 -> HQ (correct)
> dbbranch1 -> HQ (correct)
> dbbranch2 -> dbbranch1 (not supposed to be happen, but it happens, any
> idea?)
> Any help? Thanks.

MultiMaster is not 3 masters to one target, it's 3 masters between
themselves..  Consider this (my understanding of Bucardo5 is your
configuration results in the following):

dbbranch1 <-> dbbranch2 <-> dbbranch3
      \            |          /
                     \-> dbhq

 (if my ASCII art works)


Michelle Sullivan

More information about the Bucardo-general mailing list