[Xymon] xymon feature I would like to add too

[cleaver at terabithia.org]

> I vote for the feature discussed. It can be very useful.
>
> Another thing I would like to offer is to add new dimension: target object
> and some king of screen display with grouping by target objects of user
> defined type or just per target level.
>
> For example, on some hosts we monitor databases (there may be several
> per host), on some application server instances.
> Now I have implemented loop at client side to test and show status for
> each database instance per host and calculate overall status for metric
> (dbup, dbpga, etc.)
> Problems with this are:
> a) when we have both databases and application server on same host -
> screen width is naturally to short to display all metrics (I use Opera
> browser and it just compact to display all the screen, so this a bit
> resolve the problem for me, but formatting is not so good).
> b) we can't see statistics and history per target (database/application
> server instance) , so if I need to check problems of the exact instance
> I have to check all events for the metric, check details for each event
> and filter out only required in my memory.
>
> So my offer is to add onether one dimension target object (that can be
> of user defined type, by default all is grouped by target type=host)
> I know it is not very simple to implement, but may be you will find this
> may be useful.
>
>

Another option, depending on how complex your page generation already is,
s to create separate host entries for each service. I'm using pseudo DNS
names to track "instances" of apps or dbs as distinct servers. I can then
report status results (and store client data) for each service
independently.

To wit - our config generator creates a separate page for each DB/App
server, with groupings below it for each instance on that box. For
snapshoting purposes, I have a script listening on the stachg channel for
anything relating to app or db hosts that sends a spurious status flap for
a "tied" test for each related instance/node on that box to ensure that
clientdata gets recorded for each of them.

With a small patch to allow for generic/unknown clientOS types, you can
even got logfetch and client-local.cfg working properly for per-instance
error logs and such.

HTH,
-jc