Unexpected Service dependencies

[gwakem]

This is in XI 5.2.5.
In an effort to avoid creating several thousand service dependencies, with each named after the specific host, I attempted to set up service dependencies in the following way:

Config Name: NRPE - linux - rhel7 - FirewallD
Manage Hosts: Added 21 hosts
Manage Hostgroups: Nothing added
Manage Services: Added "NRPE Process"
Manage Dependent Hosts: Nothing added
Manage Dependent Hostgroups: Nothing added
Manage Service Dependencies: Added "NRPE - FirewallD"

All 21 hosts have both the NRPE Process check and the NRPE - FirewallD check (via nrpe).

When I click the Relationship Info button, I see all 21 hosts I entered. I also see Server1-NRPE - FirewallD and Server2-NRPE Process.
Server1 and Server2 were not included in the group of 21 servers I added. When selecting the Service and Dependent Service, It doesnt show the host that either service is attached to, just the service name. It appears it just grabbed an arbitrary hosts that had the service and dependent service.

I'm betting I misunderstood how this was supposed to work, but I'm hoping someone can explain this to me. And then maybe give me an option outside of creating several thousand individual service dependencies.

[tmcdonald]

And then maybe give me an option outside of creating several thousand individual service dependencies.

Before I even address the rest of your post, I might have an answer for you. But that will depend on the end goal - are you trying to make all services dependent on their hosts? If so, I added a config directive to Core for just this purpose: host_down_disable_service_checks. Set that to 1 in nagios.cfg and restart, then all services will automatically be dependent on their hosts, and not be checked if the host is down.

[gwakem]

That is actually super helpful, and I was looking for that. I was trying to find the thread from 2015 earlier today that detailed the push for that and the efforts to get it pulled into core, so I will definitely be using that.

But this is more to prevent an avalanche of alerts if NRPE, NSClient++, or to a much lesser extent, SSH dies.

[tmcdonald]

I was trying to find the thread from 2015 earlier today that detailed the push for that and the efforts to get it pulled into core

Probably this one? https://support.nagios.com/forum/viewto ... 30#p140639

But this is more to prevent an avalanche of alerts if NRPE, NSClient++, or to a much lesser extent, SSH dies.

We'd need to look at the configs to get this sorted, most likely. Can you PM me a copy of your profile? In the XI web interface, go to Admin -> System Profile and click the blue "Download Profile" button. Then attach that profile.zip file in a PM to me.

Please don't post it in a reply here, it may contain sensitive information.

[gwakem]

Understood, I will get that over to you shortly. Thanks!

[gwakem]

I just attempted to do so, but I am getting an error stating I am not authorized to send PMs.

Mod Edit: File received and shared with support team. You had some weird forum permissions I had to reset.

[lmiltchev]

I suspect this is a bug. I can see some weirdness here. I created a test service dependency with 3 Windows hosts. All of them have "Ping" and "CPU Usage" service. Here's my service dependency definition:

Code: Select all

define servicedependency {
	#	config_name	TEST
		dependent_service_description 		CPU Usage
		host_name                     		Windows-SNMP,Windows7,Windows7-WMI
		service_description           		Ping
		inherits_parent               		1
		execution_failure_criteria    		w,o,
		notification_failure_criteria 		w,o,
		dependency_period             		24x7

}

When I click on the "Relationship Info" button, I see this:

example01.PNG

The highlighted section shows two services (CPU Usage & Ping) on completely different hosts. The 2 hosts in question are on the top of my list under "Services" (CCM), when I search for "Ping" or "CPU Usage" with no "Filter by Config Name" selected...

It seems like that the "same host dependencies" functionality is broken. We will have to do some more digging into this.

[gwakem]

That's exactly what I'm seeing. In at least one case, it wasn't the top host either, which is how I stumbled across this. In trying to remove a service prior to the removing the host, I discovered that about 1/4th of the systems complained about dependencies. I'm not sure if that was due to the service (attached the incorrect host) in question having a lower ID than the rest or something else, but it definitely wasn't the first alphabetically.

[lmiltchev]

I filed an internal bug report (TASK ID 8179). Thank you!

[lmiltchev]

@gwakem, our developers determined that this is a CCM bug (showing incorrect info). According to them, the service dependencies should work as intended. The only issue is that they are displayed incorectly in the CCM. Let us know if you have any more questions/issues. Thank you!