On September 29, 2004 06:28 pm, cbbrowne at ca.afilias.info wrote:
> > Hi All,
> >
> > The overview is great!
> >
> > I am evaluating Slony as a potential replication solution for our
> > production environment (Linux / Postgres 7.4.2) and I've been creeping
> > around the past couple of weeks reading all the postings.  I have been
> > successful (I think) in configuring our environment for a simple master
> > --> single slave replication scenario.  We would ultimately want to be
> > able to add slave nodes and possibly cascade the replication for disaster
> > recovery purposes; most likely replicate to a backup machine also.
> >
> > Where I'm running into a little trouble is actually having a process for
> > starting and monitoring the replication.  Someone had mentioned in a
> > previous posting a request for Nagios monitoring, which (coincidentally)
> > is what we're using here :-).  If I understand correctly, it is really
> > just a matter of returning one of three values indicating service status:
> > ok, warning, or fatal results.
>
> I have got a trio of scripts that we are using to allow Nagios to monitor
> a bunch of Slony=I instances...
>
> 1.  There's a process that, given the identity of one Slony node, looks
> for all the active nodes in a set there, and then injects a test update to
> go to all the nodes.  That runs every few minutes, and generates a report
> consisting of a line for each node.
>
> 2.  There's a "controller" script that runs #1 several times, as we have
> several Slony-I clusters.
>
> 3.  Lastly, there's a Nagios-oriented script where you specify the cluster
> and the node, and it pulls back results for that node.  You'd set up one
> of these for each node you want Nagios to monitor.
>
> #1 and #3 would be of general interest; I'll see about adding this to CVS
> some time soon.  I have two challenges there, namely to
>
>  a) Clean up the code a bit;
>
>  b) Come up with some alternative to some local application dependancies.
>
>     Pointedly, since we're doing domain management, it seems a useful
>     thing to report back the latest domain created by our application,
>     as that boosts Nagios users' confidence in the output, and gives
>     Nagios something honestly useful to display.  For someone at a
>     library, it would be logical to report back something
>     about the latest lending transaction.  My query won't be your
>     query.

This would be a wonderfull addition, and a great starting point for the SNMP 
enabled version of slon that will be in -HEAD over the next weeks.

>
> I'm busy on 1.0.3 matters 'til Friday; hopefully I can touch on this
> then, and perhaps get it into 1.0.3.
>
> > Can anyone give me some guidance in configuring the environment from an
> > operations point-of-view?
>
> Possibly I can draft a new staff member to do some secretarial work in
> documenting some of what we're doing.
>
> I suspect that the right answer on this might be if someone can put
> together an outline for a document on this, and turn it into a Wiki so
> that questions can be added and so that people can add tidbits rather than
> feeling forced to write an 8 page report that they haven't time for right
> now.  Furthermore, that allows people to add questions that need answers.
>
Once the migration to pgfoundry is complete there should hopefully be this 
feature.. (I'll check with the foundry people on this)

> I've got a wiki running on my firewall at home that I could point people
> to, though that, being an old P200, is probably not the most robust
> location for it.  (And the portsentry's pretty paranoid, so you can easily
> lock yourself out :-).)
>
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at gborg.postgresql.org
> http://gborg.postgresql.org/mailman/listinfo/slony1-general

-- 
Darcy Buskermolen
Wavefire Technologies Corp.
ph: 250.717.0200
fx:  250.763.1759
http://www.wavefire.com