Planned new Nagios XI systems and gearman

[gormank]

Hi,
We have a planned new system. I'm beginning to understand that the network, or the firewall setup will be blocking NRPE for example, and I expect most everything they can. I think their plan is to have redundant gearman VMs on the other side from Nagios, and allow the gearman boxes data through to act as concentrators for Nagios messages. Unfortunately the only description of what they want to do or how was the word "gearman."

I've read the .pdf from exchange and various other docs but it still seems a bit thin. Got any additional info?

thanks

[dwhitfield]

I think https://labs.consol.de/nagios/mod-gearman/ is the canonical resource.

There are also some presentations from the Nagios World Conferences on the matter: https://www.youtube.com/user/nagiosvide ... ry=gearman

If you have specific questions we can point you in a more specific direction.

[Box293]

Probably the only thing not 100% clear is the inbound network port that needs to be open on the XI server.

It's the remote workers that initiate communication with the Nagios XI server. Therefore TCP and UDP port 4730 should be allowed inbound on the XI server and any firewalls in the path.

[gormank]

I've read the doc on consul.de, and it provides a number of scenarios, and things are still thin. I read it before posting initially.
For example, the list of hosts indicates that there will be redundant gearman hosts. How does the 2nd gearman host know that it shouldn't send a duplicate request for the same check to run?
How are SNMP traps handled? It seems pretty obvious that they never hit the gearman hosts. Snmpget/walk handled by NRPE would go through gearman it seems.
The port info is interesting. I'm not sure how the scheduling system (Nagios) is no longer involved in scheduling, unless gearman boxes poll Nagios, which is somewhat senseless.

This may be clear if you've dealt with it before, but its obviously not as clear to someone new to gearman.

[dwhitfield]

I know you said you were using gearman, but you might also want to look at DNX: https://assets.nagios.com/downloads/nag ... ios-XI.pdf

[gormank]

Actually I said I was given the one word description of gearman. What I want seems irrelevant. I'm being told they plan something, but have few details.

Interestingly, their name for the extra boxes is Nagios Executor, so it sounds familiar at least in name.

The instructions in that doc say log onto one of my Nagios servers acting as a slave... I will have 2 Nagios XI servers, an active and a passive, both in the same location in the network. Thus, this solution doesn't help, unless I start installing Nagios Core on parts of the network behind firewalls and on other subnets.
This doesn't seem like a solution that's viable.

[dwhitfield]

For the sake of clarity, is https://assets.nagios.com/downloads/nag ... ios_XI.pdf the PDF you read from exchange?

There's also https://support.nagios.com/kb/article.php?id=484

[gormank]

I don't see any attempt to respond to these questions from above:

There will be redundant gearman hosts. How does the 2nd gearman host know that it shouldn't send a duplicate request for the same check to run? Our systems are redundant, so a gearman solution also needs to be redundant.
How are SNMP traps handled? It seems pretty obvious that they never hit the gearman hosts. Snmpget/walk handled by NRPE would go through gearman it seems.

Yes, that's one of the few, if only gearman pdf on exchange and it too is less than satisfying.

[Box293]

Because it sounds like you are new to Mod-Gearman let me explain how it works.

On the Nagios XI server the Mod Gearman engine takes the check requests from Nagios and places them in queues.
The workers connect to the Mod Gearman engine and execute any jobs in the queues. So it's the worker's responsibility to take checks from the queue, execute them and then send the result back.
If there are no workers connected then no checks will be executed
If you have multiple workers connected they will execute checks in the queue, if one of the workers disappears then the other workers will take up the slack.
You can configure workers to only execute specific checks using hostgroups and servicegroups. This is handy when you have a remote site the Nagios server cannot access.

This Mod-Gearman guide explains how it all works in more detail:
https://support.nagios.com/kb/article.php?id=484

There will be redundant gearman hosts. How does the 2nd gearman host know that it shouldn't send a duplicate request for the same check to run? Our systems are redundant, so a gearman solution also needs to be redundant.

The official documentation is a bit thin, on this page https://labs.consol.de/nagios/mod-gearman/ the word redundant is only featured once so it makes it hard to answer as it's not clearly defined.

How are SNMP traps handled? It seems pretty obvious that they never hit the gearman hosts.

SNMP Traps are not handled my Mod Gearman however there does seem to be a send_gearman command in the documentation which could do this, so if you had the snmptrapd and snmptt engine on the worker then you could perhaps using the send_gearman command instead of the snmptraphandling.py command.

Snmpget/walk handled by NRPE would go through gearman it seems.

Correct, you must make sure that your devices accept connections from the worker IP addresses as they are the ones doing the talking.