management of multiple perlbal instances?

[dormando]

We do something a little ghetto.

We have two syncs. One for pushing out new configs, and one for making 
live changes to everything but node members.

Since perlbal automatically updates a node list when a file changes, 
that gets synced out along with the template files. Then perlbal picks 
up the changes, blah blah.

When we make a config change, we edit the config file, then hand-write a 
"diff" (since it'd have `use service` instead of `create services`, 
etc). We test the diff on one machine, then have a script blast it out 
to the management interface on all of the perlbals.

The rollout process is pretty quick, most of the time is spent writing 
the new config and testing it.

-Dormando

> How do people handle propagating configuration changes to multiple 
> instances of perlbal?   I can see extending/modifying things to check 
> for new config files, or by putting together a management proxy that 
> could connect to each instance.  Any suggestions?  Anyone already doing 
> something similar?
> 
> Thanks much.
> 
> -- jeff gelb