[Pkg-nagios-devel] default configuration audit

[sean finney]

hey marc,

On Sat, Feb 04, 2006 at 02:02:11PM +0100, Marc Haber wrote:
> On Sat, Feb 04, 2006 at 07:38:54AM -0500, sean finney wrote:
> > - reorder/recomment the various cfg_dir/cfg_file comments/directives in
> >   nagios.conf to make more sense to the "new reader"
> 
> What is the recommended order?

well, my thoughts were that the current order (starting at OBJECT
CONFIGURATION FILE(S) line 17ish) is not exactly intuitive to
someone installing nagios not extensively familiar with
how things are set up.  i don't have an order to recommend
off the top of my head, but my thought was to have it
be more sensible for someone reading it in top-to-bottom
order.

> > - oh, look: p1_file=/usr/sbin/p1.pl  <- so we can change that afterall.
> 
> Yes, but we still don't know whether it makes sense to call it from
> the command line or not.

well, since we don't know, i'd opt for defaulting to not having it
in /usr/sbin.  if it breaks something, i'm sure someone will complain :)

> > - should we enable regex matching by default?
> 
> I don't think so.

okay.  i suppose it could have unintended consequences...

> > - conf.d file numbering isn't necessary, the order is still arbitrarily
> >   random (based on the results of getdirent).
> 
> Yes, you're right, and nagios doesn't care about order. I'd still
> vouch for keeping the package name in the file name.

yes, i'd say keeping the package name in the filename would
be good just to drive the point home where the file came from.

> > - contacts and contactgroups are combined in the same file, which is
> >   nice.. maybe we should do the same for hosts/services.
> 
> I think that depends very much on the opinion of the local admin. On
> my main nagios installation, I have services connected with host
> groups and control which services are checked for a host by putting
> the host in the respective host group:

yes, this in fact how i do it as well.  i don't think there's a bettter
method that doesn't involve a complicated makefile or some wonky
perl/ruby script to autogenerate them.  i think what i actually meant in
my head was "for hosts and hostgroups", and somehow that was
mistranslated by my fingers.

> > - our config files could have a comment at the top that says "this is the
> >   default config supplied by debian, if you'd like to change it feel
> >   free to do so, but if you want to *add* to it we suggest you make life
> >   easier for yourself by using additional files in the conf.d directory
> >   or elsewhere".   at least for the template config files.
> 
> I advise not doing so. People are bound to change the files and not
> change the comment.

you're right that "this is the default config" woudl be misleading if they
did modify it... perhaps a better choice of wording could be used.
my point is that they have the right to edit it as much as they want,
but they will get less conffile conflicts in future upgrades
if they simply use seperate files for additional config.

> > - notification_interval for hosts and services should *definitely* be 0
> >   by default.
> 
> I don't care either way, you have more nagios experience.

basically, notification_interval set to anything but zero will
flood the admin's mailbox with emails.  set to zero will only send
notifications on change in status.  i've found from practice
that if you don't get the first email, you probably won't get the
next 1500 either :)  worse, if the notifaction is something that
actually *costs* you to send (like an sms page)....

> > - example hostextinfo 
> 
> I still don't understand hostextinfo. It looks a lot less flexible,
> and clumsy to configure. See my message on nagios-users dated
> yesterday.

yes... personally i think the hostextinfo should have been integrated into
the host objects themselves a long time ago.  i can't see why it
shouldn't, anyway.

but in any case, you should be able to assign the hostextinfo
to hostgroups using hostgroup_name instead of host_name.  unfortunately
of the two servers i had that made use of this, one is in a place i
no longer work, and the other is powered off under a table thousands
of km from here :(

> > and escalations configuration?
> > - hostgroups?
> 
> Do you think that we should deviate that far from upstream's
> configuration? Or do we have a chance to get our configuration adopted
> by upstream?

i think for escalations we'd probably leave a few sample configs
commented out for reference, as it's one of those things that
is often fairly specific to the site...  

as for hostgroups, i don't think it's really deviating from upstream
to have a couple hostgroups defined.  i was thinking something like an
"all" wildcard group, maybe a single service-focused group (like
imap-servers or something) as an example, and/or then maybe a few commented
out example groups.


	sean
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.alioth.debian.org/pipermail/pkg-nagios-devel/attachments/20060204/656cbb00/attachment.pgp