Can I set a dependency in this way with my services

Hi all,

I see the dependency section in the director but I am not sure how to use it and have read the monitoring basics - dependency section as well. I’m looking for advice please.

I have a clustered situation where I have 2 printing nodes. One of the two is the active node and let’s say there are 2 service checks running on both nodes. The way we have to do this now is to go to the inactive node where the 2 service checks will fail and put them into Acknowledge. At some point there will be a failover to the other node and then the service checks will fail on the other node.

Is there any way I can fix this so that it is ‘cluster aware’ and I don’t have to ack one set of services in the inactive node and is this type of scenario what the dependency section in the Director is actually for?

Thanks

The dependendy configuration is for :drum: dependencies :wink:

With this configuration you can set parent-child dependencies for hosts/services.
Example:
You have a central switch. If that fails you can’t reach anything that is behind it.
So you configure a host dependency where the central switch is the parent and all the hosts behind it are the children.

apply Dependency "Central-Switch" to Host {
    disable_checks = false
    disable_notifications = true
    ignore_soft_states = false

    parent_host_name = "central_switch01"
    assign where match("behind_central_switch*", host.name)
    states = [ Up ]
}

This means, as long as the central switch is UP, all hosts behind it are reachable.
If the switch goes down, the hosts behind it are unreachable. In this case the dependency will fail, the host will be set to an “UNREACHABLE” state and the notifications are deactivated.

Long story short: Dependencies are generally for disabling mass notifications caused by the failure of one central system.

Regarding you problem:
I don’t think a dependency is the right way for this, because it is not dynamic.

I would recommend you take a look the business process module:

With that you can define a process for your printing service that is “green” as long as one of the printing hosts and it’s own service is running.

Exmaple BP for our Citrix Terminalserver env:

Thanks again log1c, always very helpful!

I was already testing out the business process for Printing actually last year but hadn’t had time to play around with this any more. But from what I understood, this is more of a graphical thing, ie one’s view for say - printing - could be a BP view. But then the services in tactical overview and such would stay red even if the Business Process stays green as there is availablility. This is fine, but in an OPS environment, I guess one would still have to ACK the ‘down’ service that is part of a cluster for example. And this is what I meant with the original post. I should have added that I was talking from an operations point of view and especially with the Tactical Overview, which I know you can customise and turn into a new dashboard/dashlet if needed.

I understand what you mean.
And yes, the single checks will still show up as a problem in the monitoring, as they should, bc they will need to be fixed.
Possible solution:
Have a “service dashboard” that displays the provided services (like printing). You could also disabled the notifications for the normal checks and only get notified on a full service failure.


Well they don’t really need to be fixed as it’s not a problem, it’s just a cluster :slight_smile: but they still show up as red of course. But yes notifications only on full service failure sounds like a good idea. The general tactical overview is the screen we use the most, so of course reds receive critical glances from upper management.

Thanks for the screenshots, I think having the different services on BP would look good for some of the dashboards I create, will investigate that much more.