Dependencies in large environments

[Mark T]

In our environment (6 main office locations with 10-60 switches each, a few hundred branch offices and 4 hosting locations), we started using Zabbix about 18 months ago, first as an experiment, but 8 months ago we started dedicating serious hours and hardware to it.

Ever since, we've got a basic setup of Zabbix (1.8.2) going (Simple checks of all infrastructure components, custom screens, fully mapped topology, etc). Zabbix has met our expectations, but now we are looking at the more advanced features, such as dependencies.

Unfortunately, the available documentation is a bit lacking. The Zabbix 1.8 network monitoring manual by Packt publishing is a bit more helpful, but only provides an example for a very simple, custom dependency for a single host, all manually setup

What we are looking at in our environment is:



Now, multiply the above by 10, plus hundred or so smaller offices, and you may see our problem - it would take way too much time to do this manually.

The question then - how do you guys implement dependencies in a large environments, with multiple redundant links and interfaces? Any way for automation at all (Other than scripting in a XML file)?

[Jason]

Could you use the API's and a script to define all your dependencies or how they're all related and then let it add the dependencies to your triggers?

[Mark T]

That would definately be an option, but the thing is, we would prefer a solution that does not involve scripting

[Yello]

That would definately be an option, but the thing is, we would prefer a solution that does not involve scripting

But when faced with a problem that involves significant labour isn't scripting, generally, the prescribed solution for that?

TBH, when triggering becomes non-trivial (dependencies are one example) then the api is the only sensible way to deal with it efficiently.


Regards,
David

[Jason]

If the dependencies are complicated and from the diagram above when scaled up they are likely to be then I think the scripted approach is the only one that makes sense. Also as long as your logic is sound and can be tested on small scale there is less chance of mistakes creeping in when using scripts as opposed to doing it all by hand.