Dynamic Monitoring in Nagios XI

[logix88]

Hi,

Wondering if anyone has attempted to dynamically monitor services (for e.g. file systems, VMware VMs and other ever changing items) using Nagios XI? I realize you can create a blanket/global plugin, say for example for file systems to say if any mount point exceeds a threshold then fire, etc. but that's won't collect historical performance for each file system and also if more than one file system is over the CRITICAL threshold - After having alerted once, it may not be alerting us a second time when a new filesystem is over threshold, etc.

During the Nagios XI Wizards, you need to manually specify file systems rather than it being 'discovered'. I was wondering how people are dealing with this issue?

[jrdalrymple]

Hi,

Wondering if anyone has attempted to dynamically monitor services (for e.g. file systems, VMware VMs and other ever changing items) using Nagios XI? I realize you can create a blanket/global plugin, say for example for file systems to say if any mount point exceeds a threshold then fire, etc. but that's won't collect historical performance for each file system and also if more than one file system is over the CRITICAL threshold - After having alerted once, it may not be alerting us a second time when a new filesystem is over threshold, etc.

What specifically are you wanting to be dynamic? Are you wanting thresholds to change based on some metric or observed behavior? Like "If this VM is down for more than 4 hours, just forget about it?" It's really not entirely clear what you're trying to achieve. What about VMware VMs is

ever changing

?

During the Nagios XI Wizards, you need to manually specify file systems rather than it being 'discovered'. I was wondering how people are dealing with this issue?

That's not true for all wizards. One of the major improvements in XI5 was the expansive use of smart wizards. SNMP and WMI wizards will now gather filesystems for you and you can just clicky clicky to add them. Hopefully I'm not making that up, it has been awhile since I've run either, but I think that was included. I'm not sure what to expect out of the agent based wizards as far as smart discovery of filesystems go, but with say NSCP and NCPA I could definitely see the possibilities being there. I'd suggest a feature request, except I think the devs kind of have a standing goal to continue to make each and every wizard "smarter".

[tmcdonald]

I too am a little unclear on what is needed, but two things that might help are BPI and autodiscovery:

BPI: https://assets.nagios.com/downloads/nag ... _Addon.pdf

Autodiscovery is pretty straight-forward and is available under Configure -> Auto-Discovery. Essentially it uses nmap in the backend to scan remote hosts and determine open ports.

In regards to the dynamic thresholds that JR hinted at, I wrote some documentation on using bischeck for just this sort of thing:

https://assets.nagios.com/downloads/nag ... ios-XI.pdf

[WillemDH]

I think he means with dynamic a system which automatically detects new hosts / disk / services and add them to Nagios XI automatically once there are detected. I'm running a daily inventory script (in Powershell) which goes through all hosts of our Inventory db, checks if they exist in Nagios, updates the free variables, and add any services if necessary. This is possible thanks to Nagios Reactor, but requires some scripting abilities.

[tmcdonald]

Yea, nothing like that really built-in. It's been discussed in the past, and people (like Willem) have come up with ways to do it, but we don't have anything by default.

[WillemDH]

It would be really hard for Nagios or any Enterprise to make something like that work for every customer. I've worked for years with Microsoft SCOM where 'management packs' were able to discover new disks etc, but this brought a whole new range of problems such as excessive noise and complicated configuration layers. With the help of Nagios Reactor, the new REST API and some other XI features, such as the JSON query generator, and some scripting Bash / Powershell skills in my humble opinion Nagios XI / Reactor is far more dynamic then any other monitoring solution I worked with.

[tmcdonald]

I have to agree with Willem here, there are a lot of variables in play for this sort of setup and nothing we create would satisfy everyone's needs. If you have specific questions about curremt capabilities we would be more than happy to help, but as far as dynamic monitoring as you describe it, I don't think I have more to add.

[logix88]

That's not true for all wizards. One of the major improvements in XI5 was the expansive use of smart wizards. SNMP and WMI wizards will now gather filesystems for you and you can just clicky clicky to add them. Hopefully I'm not making that up, it has been awhile since I've run either, but I think that was included. I'm not sure what to expect out of the agent based wizards as far as smart discovery of filesystems go, but with say NSCP and NCPA I could definitely see the possibilities being there. I'd suggest a feature request, except I think the devs kind of have a standing goal to continue to make each and every wizard "smarter".

WMI and SNMP are not always options in a larger enterprise, it's actually frowned upon if you bring up making WMI/SSH/SNMP remote polling to every server every x minutes on the entire estate. Agent based connections are preferred but even with this SNMP/WMI wizards that discovers filesystems and "clicky clicky" add to a host, issue I am trying to highlight is when a new filesystem is added to the target host, somewhere down the line (after initial Nagios config) this is not 'discovered' by Nagios at that point (i.e. not dynamic). Now, I don't want to make comparisons of products here, but to explain what I am against: Zabbix Low Level Discovery (LLD). I personally don't like Zabbix, but having used both products - the LLD feature is something I can't figure out in Nagios without having some external scripting for each kind of application which is a tall order.

So if I somehow manage to write a script to 'scan' filesystems and update the Nagios config for each host (with the new APIs, this should be possible) - I then need to think of every application that would need this sort of capability. Say for VMware or XenServer, when you first add into Nagios - it dynamically picks up the VM level details, i.e. creates a service per VM. But when the VM drops out or new VMs are created... Nagios doesn't pick this up - so I basically need to them come up with yet another script to query the different sources, find with correct services/hosts I need to update. And doing this for each application we need monitored is just too much work and something I expect Nagios XI to be able to solve (at least for well known products like VMware and basic things like file system monitoring.).

Hope that explains a bit more on what I was after...

[logix88]

It would be really hard for Nagios or any Enterprise to make something like that work for every customer. I've worked for years with Microsoft SCOM where 'management packs' were able to discover new disks etc, but this brought a whole new range of problems such as excessive noise and complicated configuration layers. With the help of Nagios Reactor, the new REST API and some other XI features, such as the JSON query generator, and some scripting Bash / Powershell skills in my humble opinion Nagios XI / Reactor is far more dynamic then any other monitoring solution I worked with.

Using the Reactor is interesting... But are you then having this done for different applications, like for instance the VMware estate where datastores/VMs/NICs,etc. are all changing all the time - removing a VM when it doesn't exist in ESX or adding a new VM as it comes along is something we need to query the vCenter using it's APIs/SDK to know about. So you are now deep diving into third party applications to try and make it dynamic - and that adds quite a bit of overhead to delivering, what I think is pretty basic monitoring. Now, some environments aren't so bothered and are aware of the shortcomings and the need for external scripting to solve that issue - I just couldn't see this being done with various types applications. For the very popular ones (especially the ones with Wizards in XI), adding this sort of feature would really make it a lot easier for people to adopt with minimal 'development' investment.

[tmcdonald]

As it stands right now this is not something we have. Even competitors who "automagically" configure themselves need to be told what to do by a programmer at some point, and won't be able to handle every possible scenario that is presented. We can certainly make a feature request for some limited dynamic monitoring, but this isn't something we have right now.