[Bucardo-general] One node down scenario

Błażej Cegiełka blazej.cegielka at gmail.com
Wed Feb 6 09:44:18 UTC 2013

Hi all

Could you please follow with me with such scenario:

Our bucardo multi-master replication consist of 3 servers.

bucardo add db region1 dbhost=host1 dbuser=rep_user dbname=dbname
bucardo add db region2 dbhost=host2 dbuser=rep_user dbname=dbname
bucardo add db region3 dbhost=host3 dbuser=rep_user dbname=dbname

bucardo add relgroup rep_tables table1 table2

bucardo add dbgroup rep_regions region1:source region2:source region3:source

bucardo add sync rep_sync dbs=rep_regions relgroup=rep_tables

We check the behavior when one of the node will go down and we observed 
that whole sync stops replicating (even between working nodes).
We tried to find some workaround for that - i.e. create fallback sync on 
only working nodes, but it drove us to another problem.
Assuming that one of the nodes is just not accessible by bucardo MCP, 
and still get some deltas that later should be populated, newly created 
syncs will only populated changes made after sync was created.

Question is:
* what is best solution to handle multi-master replication but in 
non-blocking way
* is there a way for the sync to populate all deltas since a specific 
txntime to a specific target?



More information about the Bucardo-general mailing list