How do you monitor servers?

[JamesC]

I was wondering what software people use to monitor servers. Namely- services, disk space, virus definitions, network card bandwidth, event log errors. I have seen a product by a company- Nimsoft called NimBUS which monitors servers for all of these aspects. Ideally I want a lightweight agent that runs on the servers and reports to a central server and would notify me of any problems.

 

I'm not really looking for an app that I would have to run manually to generate a report for each server, I already have a utility for this that generates a report about absolutely everything on that workstation/server. I want something that monitors realtime

[localzuk]

I use nagios and cacti. Nagios monitors services, disk space, cpu usage etc... Cacti keeps track of network usage on each server and on each switch port.

 

Sophos keeps track of its own definitions.

[JamesC]

Thanks for the quick reply! Im looking at Nagios now. Can I ask what methods it uses to alert you to a problem? Email, txt message perhaps?

[localzuk]

Well, at the moment, I rely on having it open on a screen - until I get email sorted out (i've yet to figure out how to get it to send via a secure smtp server). I used to get it to message me via jabber, but it was kinda pointless, as a webpage is as good as that.

 

I may link it to our SMS modem at some point, to send out errors via.

 

But really, you can get it to send via anything you want, as it is modular. You can create your own script/program to do the notifications with little effort.

[ahuxham]

We use Ilients Sysaid for monitoring.

 

Notifications on service stops, network drops, anything specified really. We also monitor CPU,Mem,HD usage with Sysaid, checking strain on servers, and timeframes for spikes.

 

Quite dandy really.

[Ric_]

Nagios for me too - there is a guide on the wiki for setup, including instructions for monitoring Windows servers.

 

@localzuk: I found that the best way to do the email was to simply use sendmail on my Nagios box. If you set things up correctly you can even send externally from the box using Exchange as a smarthost.

[localzuk]

@localzuk: I found that the best way to do the email was to simply use sendmail on my Nagios box. If you set things up correctly you can even send externally from the box using Exchange as a smarthost.

 

Fancy writing an addendum to the wiki article about it?

[dhicks]

I was wondering what software people use to monitor servers.

 

A Python script that periodically connects to each server via SSH and runs various commands to get information. I probably should switch to Nagios / Cacti too, although the servers I want monitored are Xen virtual machine servers so I want to be able to monitor VM state on each server.

 

--

David Hicks

[JamesC]

@dhicks- without throwing this topic off track I wasnt sure if you were aware of the getstate command for vmware?

 

You can check the VM state by typing:

 

“vmware-cmd //server.vmx getstate”

 

should return:

 

getstate() = on

 

This could be handy for a script to check virtual machines. Thinking about it now ICMP could also be enabled within the os of the virtual machine and a ping could be used to verify the server is 'up'.

[bizzel]

Once a week we send a student into the server room to anoint the cabinet with holy water and offer a sacrifice of freshly demagnetised floppies. Other than that, no monitoring here.

[dhicks]

@dhicks- without throwing this topic off track I wasnt sure if you were aware of the getstate command for vmware?

 

Nope - our server are running Xen, so I don't get much practice with vmware :-) I'm sure there's an equivalent, easy-to-use command for Xen, though I still rather like my own script as I can lay out the information exactly how I like it!

 

--

David Hicks

[localzuk]

Nope - our server are running Xen, so I don't get much practice with vmware :-) I'm sure there's an equivalent, easy-to-use command for Xen, though I still rather like my own script as I can lay out the information exactly how I like it!

 

--

David Hicks

 

You can integrate your own scripts with nagios - just get nagios to call a script and parse the result. Easy enough to do.

[JamesC]

Nope - our server are running Xen, so I don't get much practice with vmware :-) I'm sure there's an equivalent, easy-to-use command for Xen, though I still rather like my own script as I can lay out the information exactly how I like it!

 

--

David Hicks

 

LOL Sorry mate! I didnt read the word Xen in your post

[Geoff]

Nagios + Cacti here too.

 

For virtual machines, why do you have to treat them any differently than real physical boxes? The same software and configuration will work surely?

[JamesC]

Nagios + Cacti here too.

 

For virtual machines, why do you have to treat them any differently than real physical boxes? The same software and configuration will work surely?

 

Yes Geoff you are dead right. You could install the nagios windows client into the virtual machine running windows and have it monitor the virtual machine as though it is a physical server. I think dhicks was talking about monitoring a virtual machine state i.e. is it online, are there snapshots available etc.

 

For what its worth I would probably use your method and install the client into the virtual machine and have it monitoring services/processes/disk space etc

[dhicks]

I think dhicks was talking about monitoring a virtual machine state i.e. is it online, are there snapshots available etc.

 

Yes, that's it - my current script logs in to each server via SSH and runs "xm list" to get a list of running virtual machines, something like this:

 

Name ID Mem VCPUs State Time(s)

ACSFILES001 8 256 1 -b---- 10485.6

ACSLDAP001 4 256 1 -b---- 11841.5

Domain-0 0 235 1 r----- 1045673.1

 

My script then parses that output and shoves it in to an HTML page. I'll defiantly have to look at Nagios, see if I can get a plugin going.

 

--

David Hicks

[Geoff]

Yes, that should work. You might have to slightly alter your script so it enumerates though each xen domain and reports it as a separate 'service' (that's the only way nagios will understand).

[JamesC]

localzuk, do you use the NSClient++ for windows client reporting?