[Bucardo-general] Very general/basic questions

[Goran Gugic]

On Mon, Jun 8, 2009 at 5:26 PM, Rodrigo Gonzalez
<rjgonzale at estrads.com.ar>wrote:

> On 06/08/2009 06:33 AM, Goran Gugic wrote:
>
>
>> In order of importance... :)
>>
>> 1. Can bucardo handle the following?
>>
>> MasterA <-> MasterB
>> MasterA <-> MasterC
>>
>
> Yes, I have exactly that working.


>
>> with conflict resolution that latest record wins and that the changes
>> are made to all three masters? Would this be normaly accomplished with 2
>> swap syncs with goats resolution set to latest record (what else is
>> important in this scenario? should makedelta be set to true in goat and
>> sync? should has_delta be set to true in goat?)
>>
> Not sure about this, doesnt matter in my situation, my information is not
> updated at same time at every server


Thanks for the reply, really appreciate it! :)

Update: I got it working. This was one of those cases where immediately
after making noise on the list things fall into place (my problems were
caused by incomplete entries in bucardo_delta_targets and therefore some
tables were not pushed in one direction, recreating syncs solved it; I also
run bucardo_purge_delta)


> 2. Is it possible to take a single sync and/or single database offline
>> and keep runing bucardo?
>>
>> For example: MasterC looses connection to MasterA. I would like to
>> continue replicating between A and B while fixing the connection. Can
>> this be done?
>>
> Yes, and when C connects again will be updated


Got this working, too (still have a few tests to run).

So far, I am under impression that I have to follow the status of the sync
and if there is a problem with db connection make the problematic sync
inactive by
1. settings sync status to inactive in bucardo.sync
2. running bucardo_ctl deactivate <sync_name>
(so that the MCP keeps running; otherwise MCP exits due to one broken db
connection and no replication is done even for live connections)

Consequtively, I am under impression that, when the connection problem is
resolved, I have to
3. sett sync status to active in bucardo.sync
4. run bucardo_ctl activate <sync_name>
(to bring back the sync in play)

First - am I right here about 2 steps or bucardo_ctl deactivate/activate
should do it in one step and update the table?
And secondly - does bucardo have capability to be configured to do the above
automatically or will I have to roll my own? Is it the proper way to do
these things in bucardo (running steps 1-4 while MCP is either active or
trying to respawn).

Forgot to specify versions and config earlier: bucardo 3.2.7, postgres 8.3.7
servers (MasterA and bucardo on ubuntu 8.04, secondary masters on windows
machines), connections are over ssh tunnels.


>> Pointers to documentation, other discussions, etc are all highly welcome
>> Best
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Bucardo-general mailing list
>> Bucardo-general at bucardo.org
>> https://mail.endcrypt.com/mailman/listinfo/bucardo-general
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mail.endcrypt.com/pipermail/bucardo-general/attachments/20090608/be1e619d/attachment.html