[Bucardo-general] Bucardo in HA environment

Adam Wendt thelsdj at gmail.com
Wed Feb 13 21:48:00 UTC 2008


Hi Stojan,

I think bucardo can run as any user that can connect as the bucardo
user to the bucardo database, and user you specify in the 'db' table.

I haven't run into the main bucardo control script dying, but if it
did in most cases just running bucardo_ctl start again should get
replication going, same if server reboots, (either the server running
bucardo or the other server).

The 'source' is the herd you specify as the 'source' in your sync
table, 'target' would be the db you specify as targetdb in your sync
table.

I think latest was added as an afterthought, it doesn't actually work
in the 3.0.7 release, but there is a fix in git you could apply to
3.0.7 if you didn't want entire bleeding edge.

There are some other issues with 3.0.7 and even git that I have come
across, with PostgreSQL 8.1 you are going to need to use 'pg_class'
method for disabling triggers/rules or your tables will be locked for
however long a replication batch takes. With 8.3 you can use 'replica'
and that doesn't lock the tables.

Also there are some issues if you are doing multi-master on a WAN with
the time replication takes, let me know if you think that might effect
you and we can talk about my attempts at fixing this.

- Adam
Adam Wendt Consulting

On Feb 13, 2008 1:17 PM, Stojan Rancic <stojan.rancic at iprom.si> wrote:
> Hello,
>
> I'm deploying bucardo on a two-server, master-master setup scenario,
> with Postgresql 8.1 (Debian Etch).
>
> The test deployment went by fine, but I have a few questions:
>
> As which user should the Bucardo processes be running, should this be
> root or can it be any unpriviledged user, with proper privileges in
> the Postgresql?
>
> As far as I see, I only need to be running bucardo_ctl one of the
> servers ? What happens if that server goes down, and then comes back
> up, after a while, will the data be automaticaly synchronised (upon
> starting bucardo_ctl) from the working server to the failed one ?
>
> What happens if, for some reason, bucardo_ctl dies, and I need to
> restart it by hand/by script, will the data be synched between the two
> servers ?
>
> As for the conflict handling, if I specify 'source' as a resolution
> method, which table is actually considered source:
>  a) The one on the server where bucardo_ctl is running
>  b) The one which was updated and needs syncing to the other (may be
>     on box A or box B)
>
> .. if a, then why is 'latest' not recommended as it sounds the logical
> way to go?
>
>
> Any other tips on a fully-redundant HA scenario appreciated :)
>
> Greetz, Stojan
>
> _______________________________________________
> Bucardo-general mailing list
> Bucardo-general at bucardo.org
> https://mail.endcrypt.com/mailman/listinfo/bucardo-general
>



More information about the Bucardo-general mailing list