[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [hobbit] resend: 2 questions



Hi,

I also have fairly large filesystems running nominally with just a few gigs left. I also run into the percentage problem. Looking through the description in hobbit-clients.cfg, there seem to be a solution:

 . . .
#    DISK filesystem IGNORE
#       If the utilization of "filesystem" is reported to exceed "warnlevel"
#       or "paniclevel", the "disk" status will go yellow or red, respectively.
#       "warnlevel" and "paniclevel" are either the percentage used, or the
#       space available as reported by the local "df" command on the host.
#       For the latter type of check, the "warnlevel" must be followed by the
#       letter "U", e.g. "1024U".
 . . .

This would give us the possibility to test on explicit disk space. I tried configuring it (windows clients) but never got it right. Anyone else tried or is this not implemented yet? Could there be a formatting discrepancy from the windows client? I do see a difference between the capacity columns reported. In my setup, it is reported with % from the windows client and with 1k-blocks from unix clients. Could this affect the test?

Regards,
--
Magnus Carlebjork
Stockholm, Sweden
+46 76 116 9008


On Fri, 18 Jul 2008, Gary Baluha wrote:

If that's the case, a fourth color would have the same limitation ;-)
(That's a lot of disk space if 100% full = gigs of free space)

With the lack of a finer granularity, the only option you have is to create
a custom script (client-side or server-side should work in this case) that
checks the _amount_ (as opposed to _percentage_) of free space, and set a
green/yellow/red threshold based on that.  You could then set up the Hobbit
alert rules like any other test, and it sounds like this would solve your
particular problem.

(a client-side script would probably be the easiest to set up, depending on
how many machines it would need to be propagated to)

On Fri, Jul 18, 2008 at 2:57 PM, michael nemeth <michael.nemeth (at) lmco.com>
wrote:

 Sorry, disagree!
I can have gigs of space left at 100%  not critical  at all !!!!  Its not
"beyond critical"  its  fatal if you hit zero free !
Either one needs finer granularity (isn't numerical limits in the work)  or
a new  fatal color.  I have licenses that run near    100 % all the time
too.


Gary Baluha wrote:

The philosophy Hobbit uses for alerting is that you're okay until you reach
a certain threshold.  At that point (yellow) you still have to respond to
the event and take care of it, before it becomes a bigger issue.  If it
continues, then you reach another threshold where stuff can (and usually
does) break.  At this point, you _need_ to respond to the event.

What you are proposing is a fourth level such that you are "beyond
critical".  This is a similar concept to being "fatally killed" (as opposed
to just being "killed").  The trick to running a successful monitoring
system is setting the thresholds in the first place (which is easier said
than done), such that you don't have any false-positives, but even more
importantly, no false-negatives (i.e. an alert you should have gotten, but
didn't).

Can you give a more specific example (in as far as I.P./security will
allow) of what you are trying to accomplish?

On Fri, Jul 18, 2008 at 11:52 AM, michael nemeth <michael.nemeth (at) lmco.com>
wrote:

 One case I can think of is for even 100% you've  lots of but if you hits
0 free you HAVE to do
some thing!

Gary Baluha wrote:

On Fri, Jul 18, 2008 at 10:59 AM, Jeff Newman <jeffnewman75 (at) gmail.com>
wrote:

Hi,

didn't see a reply, so thought i'd do a resend in case it got lost in
the shuffle

Hi All,

Two questions:

QUESTION #1: Is it possible to have a third color alert? Meaning:

One of my customers wants a setup like this:

Custom script runs on client server, reports:

foo : 80

for example.

They want less than 85 to be green, 85-90 yellow, 90-95 red, and above
95 any color, say orange.
So far as I can tell, I can only use green, yellow, and red for
alerts, and blue and purple are reserved.


Currently, no.  But it might help to understand why 4 alert levels are
desired.

 QUESTION #2:

lets say #1 above is possible, so my script sends hobbit the status
line based on the it sees, with the
status of green, yellow, red, and orange. The hobbit server recieves
it, and uses the NCV module to build the rrd etc..
In hobbit-alerts.cfg to say does the SERVICE keyword work for custom
NCV type columns?


The SERVICE tag in hobbit-alerts.cfg works for any column name, NCV or
otherwise.