[Bucardo-general] Sync from one table to many, but conditioned by a column with database destination

[Guido Brugnara]

----- Il 28-mag-15, alle 4:25, Greg Sabino Mullane greg at endpoint.com ha scritto:
> On Mon, May 25, 2015 at 05:36:49PM +0200, Guido Brugnara wrote:
>> I need to synchronize a table that contains a column with the name of the target
>> database.
>> So in the table there are lines that go replicated in different databases.
> 
> That is going to be tricky. Yes, it may be possible with a well-written
> customcode. At the very least, you would need to have all the databases
> pre-defined as targets. Then the customcode would limit the rows before
> they were written to each target.
> 
Well!

I did some tests and I see that there are 4 connections each target.
With 100 connected servers have in total 400 connections to the central server; It could be a problem of overload of the resources.
What are uses of the four separate connections?

Trying version 5.3.1 and the Head in GitHub I verified that syncs "stalled" (due to temporary disconnection) do not return to active.
It is a bug or a absent feature?
In the mail-list people suggest to restart the daemon Bucardo, but in my case with ~ 100 server reboot is slow.
Unfortunately outages are frequent (radio link), so I can not use this work-around

Also, when a target can not be reached synchronization stops for the other target.
Will I forced to configure sync different for each target?

I thank you for your time.
Guido Brugnara


>> I must also solve the inverse problem; to synchronize master tables
>> that have different databases towards a slave only table that has a
>> column with the name of the database of origin.
> 
> Should be possible as well.
> 
> --
> Greg Sabino Mullane greg at endpoint.com
> End Point Corporation
> PGP Key: 0x14964AC8