Nagios Support Forum

This is a multi-part message in MIME format.

------=_NextPart_000_0490_01C6F445.7E044E50
Content-Type: text/plain;
charset="us-ascii"
Content-Transfer-Encoding: 7bit

Atcually my point about not setting the dependent host is that on same-host
services dependency is is useless. You want to set a dependent_host_name
only if the services are not on the same hosts.

So I think it does make sense to make it an optionnal variable. And in many
cases if you forget about this it should warn.

Ex:

define servicedependency {
host_name database
service_name mysql
dependent_service_name http
notification_failure_criteria c,u
}

You're likely not running an http service on a database, and I believe
nagios will return errors if it happens.

And at some point it's up to the user to make sure his config is correct.
There's hundread of things that could be wrong and Nagios will not page or
warn for them.

Thomas

> -----Original Message-----
> From: william(at)elan.net [mailto:[email protected]]
> Sent: October 20, 2006 12:03
> To: Thomas Guyot-Sionnest
> Cc: Nagios-Devel; Tobias Mucke
> Subject: RE: [Nagios-devel] Service dependencies
>
>
> It is not that our patches are entirely bad but I think that if there
> is enough demand (as may have been shown by this thread along) this is
> the kind of feature that should go into main nagios code and when it
> does its a lot better when its going with separate option
> like has been
> proposed - for the benefit of existing users who might have
> previously
> made a mistake and expected nagios to give warning about bad config,
> but the change would cause different behavior which is not
> appropriate
> for minor version updates.
>
> I'm starting to get back to some additional nagios-related development
> and will work on that (i.e. adding this with 'same_host'
> option both for
> service and servicegroup dependencies) together with updating my other
> patches for 2.5. Expect to hear from me next week about it unless you
> or someone else releases such patch before. After that the issue would
> be to convince Ethan that it is worth including in main code tree.
>
> On Fri, 20 Oct 2006, Thomas Guyot-Sionnest wrote:
>
> > Oops! There's almost certainely a bug (I didn't experienced
> it but I fixed
> > it) when you have multiple dependency definitions where
> some of them are not
> > same-host.
> >
> > After a good sleep I realised that it would make much more
> sense to have
> > host_name set and dependent_host_name unset.
> >
> > So now here's the new usage.
> >
> > How to use it:
> >
> > Just omit both "dependent_host_name" and
> "dependent_hostgroup_name" entries
> > in a servicedependency definition. Ex:.
> >
> > define servicedependency {
> > hostgroup_name pool1,pool2
> > service_description Master service
> > dependent_service_description Slave 1,Slave 2,Slave 3
> > notification_failure_criteria c,u
> > }
> >
> > Still EXPERIMENTAL (see comment in my previous post)
> >
> > Please send me any bugs/comments you have with this patch...
> >
> > Thomas Guyot-Sionnest
> > UNIX Systems Administrator
> > Zango
> > E: [email protected]
> > P: 514.787.4714 | F: 514.787.4707
> >
> > www.zango.com
> > Read our blog at http://blog.zango.com
> >
> >> -----Original Message-----
> >> From: [email protected]
> >> [mailto:[email protected]] On Behalf
> >> Of Thomas Guyot-Sionnest
> >> Sent: October 20, 2006 1:03
> >> To: Nagios-Devel
> >> Cc: william(at)elan.net
> >> Subject: Re: [Nagios-devel] Service dependencies
> >>
> >> Alright! I found out I'm not that bad in C and made my own
> patch.
> >>
> >> What it does:
> >>
> >> When Nagios duplicate servicedependency definitions, it
> checks for any
> >> missing required variables. When it checks for empty
> host_name entry,
> >> instead of returning an error I set a flag and fill-up the
> >> master_hostslist
> >> using the dependent_host_name entry instead.
> >>
> >> Later in the "duplication" loop, if the flag is present I
> >> don't duplic

...[email truncated]...


This post was automatically imported from historical nagios-devel mailing list archives
Original poster: [email protected]