Geoffrey lists at serioustechnology.com
Fri Dec 21 17:17:14 PST 2007
cbbrowne at ca.afilias.info wrote:
>> Vivek Khera wrote:
>>> On Dec 19, 2007, at 6:12 PM, Geoffrey wrote:
>>>
>>>> As I understand it, if I'm replicating from one machine to another, I
>>>> must have identical slon_tools.conf files on both machines.
>>> no, you need to change teh connection string so each slon connects to
>>> its own server it is responsible for.
>>>
>>> the slons can run on any box; they just connect as clients to postgres.
>>> most people tend to run the associated slon on the same box as the
>>> postgres server.
>> Thanks for the response.  Any ideas on the other part of my post?  That
>> is, how do I configure slony to handle multiple databases on one machine
>> to be replicated to multiple instances on a second machine?
>>
>> We have eleven postmasters running on one machine in order to be able to
>> better control the configuration of each and limit downtime.  What I
>> need to do is replicate all eleven to another machine.
>>
>> The example slon_tools.conf file does not appear to lend itself to this.
>>   There is one MASTERNODE, whereas I need eleven.
> 
> On the one hand, the "altperl" Slony-I tools were only intended to deal
> with relatively simple situations.  What you're trying to do falls pretty
> clearly outside that "simplicity."
> 
> On another, I am not sure what sort of configuration to prefer, for this. 
> I see two ways:
> 
> 1.  To have 11 clusters, one for each "master database."  In that case,
> each cluster is a simple one; you could use slon-tools.conf, but you'd
> need to have 11 configurations.

As I expected.

> 2.  To have 1 cluster, where there are 12 nodes and 11 subscriptions. 
> This feels like it has fewer "moving parts" than the "11 cluster"
> approach.

I don't see how this would work. I guess I'm stuck on the single 
MASTERNODE value in the slon_tools.conf file.  Wouldn't all 11 
subscriptions be subscribed to a single database? I'll have to put some 
thought to this approach.

> If you're doing things that are this esoteric, don't expect the
> 'simplified tools' to work for you.  You'll need to write slonik scripts
> "by hand."  Fortunately, it's not a severely complex language - it's
> pretty simple to write such scripts.  (At least, the language is
> simple...)

This is what I'm working on right now.  The nice thing is that the 
scripts send the output to standard out, so I can take that output 
produced, save it, tweek it for the next database and so on.

-- 
Until later, Geoffrey

Those who would give up essential Liberty, to purchase a little
temporary Safety, deserve neither Liberty nor Safety.
  - Benjamin Franklin


More information about the Slony1-general mailing list