Re: RH tools and conf changes

["Chester R. Hosey" <Chester Hosey gianteagle com>]

On Thu, 2005-07-21 at 22:47 +0800, John Summerfied wrote:
> Chester R. Hosey wrote:
>  
> > Can you set up dependency trees? I do like the appearance of mon's
> > configuration format over that used by Nagios; however if you're sending
> > notifications to summer office lan if the network link between the mon
> > server and the www servers drops, that's definitely a problem in a
> > larger environment.
> 
> I'm not sure I understand.
> terad.net is another site: it could monitor this site and notify me at 
> another address.
> 
> If the routers are down at the office. I can still dial in. If the 
> power's off, we don't care any more:-)
> 
> If it's important enough, there is the paging capability, or presumably 
> one can do SMS so if the routers are down there's a pager and/or phone.

I'm sorry for having been unclear in my query.

You're in center-1, administering the web servers. I'm in center-2,
administering the email and monitoring systems. We're connected through
router-1 on your end and router-2 on mine. In center-1 is switch-1,
center-2 has switch-2.

The monitoring system is set to dial up directly to an SMS service and
send notifications if something fails. It will page the network guy if a
router fails, me if email fails, you if a web server fails.

My switch fails. The monitoring software can't access my switch, my
email servers, my router, your router, your switch, or your web servers.

I can tell Nagios that my access to web-group is dependent on switch-1,
which depends on router-1, which depends on router-2, which depends on
switch-2. If Nagios can't hit your web servers, it will start checking
the upwards dependencies until it finds the first thing it can't hit.

Without the dependency checking, when my switch fails the monitoring
software sends pages to me about my email servers, you about your web
servers, and the network guy about all of the switches and routers. With
dependency checking, Nagios knows that it won't be able to access web-
group without switch-1, switch-1 without router-1, router-1 without
router-2, router-2 without switch-2. If one fails, it can disable
notification for anything which depends on something which isn't
available either. If switch-2 fails, it won't page anyone about email-
group, router-2, router-1, switch-1, or web-group. It will, however,
page the network guy ONLY about switch-2, which is the only service
whose dependencies are all functioning.

When you have many people responsible for many things, it's definitely
nice to page only the responsible party, only for the device or service
which is actually failing.

Does that make more sense?

Chet