Dummy Host / Virtual Host

[anrstone]

This may seem a bit bazaar but is there any way of creating a dummy host (i.e. a host that doesn't in fact receive any events but has triggers). My reason for wanting to do this is so that I can in effect split common servers in terms of event escalation.

In particular we have DB servers that are common to a lot of applications and I need to be able to raise an event so that only one group is targeted with the message (i.e. I don't really want DBX has event Y even though that's actually what's happening - I would rather have Client X has event Y)

Thanks

[richlv]

maybe you can provide a more precise example, as the one here was a bit obscure.

anyway, it seems to me that you might get it all to work for you with decent user, media and action setup - but some more exact examples might help to verify this suspicion

[anrstone]

Yes fair point - my explanation was very obscure...

We host a number of clients on a mix of shared and dedicated machines. For the shared hosts typically we will have one DB server with a number of databases and 2 or more load balanced web servers to provide the site instances. In the case of the DB servers we perform a number of zabbix tessts some of which are generic (e.g. server load, disc space etc.) and some of which are very client specific (e.g. Have you received an order in X seconds)

As it stands at the moment any trigger raised on the common DB servers effectively alerts at the Database server level (which is partly correct but may in fact be because something upstream has failed). What I was trying to do is get the specific triggers to be raised on a kind of Virtual instance of the underlying DB server to indicate that the event was raised by the DB server but is not necessarily because it is faulty. This can then be used on Maps and in the Dashboard system status to specifically highlight the customer that is having the problem thus improving response times and reducing diagnosis time.

I suppose what I'm suggesting is a kind of crude event correlator which takes events from specific hosts and abstracts them up the chain a bit to more clearly indicate where the problems have started.

I hop this makes more sense?

Nick

[alixen]

Hi,

You can create an host that don't have any item but defines triggers that references items from other hosts.

For example:
On DBhost:
Item1
Item2
Item3

On Vhost1:
Trigger references {DBhost:Item1.last(0)}

On Vhost2:
Trigger references {DBhost:Item2.last(0)}

Hope this helps
Alixen

[richlv]

creating triggers on other hosts won't work here - they are associated with hosts they have items from.

as for the original problem, maybe you can solve this with different triggers - say having trigger reference more items to check for conditions you are interested in

[alixen]

creating triggers on other hosts won't work here - they are associated with hosts they have items from.

as for the original problem, maybe you can solve this with different triggers - say having trigger reference more items to check for conditions you are interested in

You are wright.
It can still be done using a dummy item in virtual host, but it is a little bit clumsy

On DBhost:
Item1

On Vhost1:
DummyItem (ex: Type Zabbix Trapper so that it will never be polled by Zabbix)
Trigger references {DBhost:Item1.last(0)=0} | {Vhost1: DummyItem.last(0)=1}

In this case Trigger will be associated to both DBhost and Vhost1.

Regards,
Alixen

[anrstone]

Thanks for the feedback - I'm going to have a play around with a few things and see what works for us. I'll leave some feedback once I've got something working.

Thanks

[anrstone]

Final soultion:

Thanks for all the feedback in the end I approached this using the Zabbix Trapper method you've suggested. I set up a new "dummy" server and then sent Zabbix Traps to that server specifically - the only slight issue, of course, is that the host will not show in the system status on the dashboard as it isn't actively monitored...

[richlv]

...the only slight issue, of course, is that the host will not show in the system status on the dashboard as it isn't actively monitored...

if i understood you correctly, you just have to add proper triggers and host will show up.

[anrstone]

Cool - I hadn't got that far I was just trying to capture some sensible data before working out what the right trigger levels would be - thanks

[untergeek]

http://simple-evcorr.sourceforge.net/

I use SEC with zabbix_sender. It's exceptionally useful.

[anrstone]

Looks very interesting - thanks for that

[anrstone]

By way of update

The way I got this to work was as per the suggestions in this thread. I've set-up new hosts to use as dummies (e.g. if the sever was web5 then I've set-up a new host web5.customername).

I set-up a simple template with the specific web data I want to capture and then add that to each virtual host and as reported here as soon as a trigger is defined everything shows up reports as it should - so thanks for your help

[anrstone]

By way of an update we now use "dummy" hosts a fair bit to handle a number of things though in nearly all cases we use trappers rather. I have tried recently to use a dummy host to set active checks but the results are inconclusive. Just in case anyone is interested I configured the host as follows:

<actual_host_name>.<client_name>

We set up a DNS record to point to the IP of the server and then using the DNS set a number of active items. Initially this worked well but then stopped working and I can't seem to get it working again.

Either way using dummy host has allowed us to split client functionality out where we have shared servers so belatedly thanks for helping all those months back.