[Xymon] Combo status question

[J.C. Cleaver]

On Mon, November 9, 2015 7:12 pm, Colin Coe wrote:
> Hi JC
>
> First up, thanks for the terabithia RPMs  :)

No problem :)

> Yep, the devices are polled via devmon.  However, these are switched
> ports not routed ports so I can't go by IP.
>
> Also, is it combo.cfg or combostatus.cfg?  I have a combo.cfg but no
> combostatus.cfg...

Ack, my mistake. Yes, it's combo.cfg.


>
> There are only a few combo tests that I want to put in.  One is a test
> to see if a particular process is running on either serverA or
> serverB.  I can see how to do this for the 'procs' test in general but
> not a specific process.  The servers that I want to test are running a
> system made up of a couple of hundred processes and with the exception
> of one process, they both should be running the same processes.

Unfortunately, there's no great way to do this at this time. combostatus
only functions by evaluating overall test status and not individual
groupings. My suggestion would be to write an interval-based parser that
downloads current process status state (and content) en masse via
xymondboard, performs the correlation (I'd suggest perl, which is fine at
simple hash structures like this), and then re-submits new test results.



A blue sky feature I'd considered going forward came out of the epiphany
that the GROUP= concept (as used in status message generation and
alerting), the modify "sourcename" (for overriding an existing color and
test result), and the '&color + reason' lines from analysis.cfg or other
generators (which get nicely parsed out in xymongen availability reports)
are all conceptually the same thing: a manner for a sub-status, referred
to by tag or identifier, to automatically affect a larger status while
still being tracked and acted upon distinctly.

It would be nice to expand this to a fully-supported concept in a future
release. Being able to query and channelize changes in "sub-status" (for
lack of a better term) would make a lot of processes like this a lot
easier.


Regards,
-jc