Best Practices for Host Monitor setup

G. Dieltiens · Post by **G. Dieltiens** » Mon Jul 20, 2009 7:26 am

Hi,

I'm part of a team of systemprogs and admins that manages a typical midsize serverfarm/mainframe/network. Together we manage about 250 servers.

We have been using Host monitor 7.1 for a while now. But the way it was implemented is not, well, "ideal". We would like to redesign our Host Monitor setup. Unfortunately, the guy that implemented HM in our shop has left the company.

So, I've been asked to have a look at it. Since I'm new to Host Monitor, it's a bit of challenge. I realize that HM can be setup in many ways depending on the serverfarm that has to monitored. And I would like to learn from the mistakes that others have made before me.

My question to you is: is there a "best practices" document or white paper available that can guide me in making the crucial choices when setting up HM?

I'm hoping it will answer questions like:
- Should we put every hardware component or functionality in a seperate folder?
- Should we create a separate Action Profile for every test?
- What are common pittfalls with regard to manageability of HM?
- What is the best naming convention for the monitoring tests?
- Should I be worried about the total number of tests and/or serverload?
- How do we define alarms and tests on redundant/double components or load-balancing servers?
- Where do we save our own definitions for profiles, mail templates, HM-options?
- etc...

Thanks,
Geert.

KS-Soft · Post by **KS-Soft** » Mon Jul 20, 2009 10:20 am

Should we put every hardware component or functionality in a seperate folder?

It depends on your needs. E.g.
- If you need separate report for each hardware component then separate folder for each component can be useful.
- If you plan to setup 10-20 tests for each server and you plan to use templates, then separate folder can be useful as well.
- If you plan to setup 1-2 tests for each server, then probably you don't need to create separate folder for each component

Should we create a separate Action Profile for every test?

I don't think so. Various macro variables allow you to use the same action profile for hundreds or thousands of monitoring tasks (test items). E.g. If you monitor 10 different services on each of 100 servers, you may use single action to restart any failed service on any of these systems.
http://www.ks-soft.net/hostmon.eng/mfra ... .htm#macro

What are common pittfalls with regard to manageability of HM?

H'm... we don't know

What is the best naming convention for the monitoring tests?

HostMonitor generates test name automatically depending on test method and test settings. However you can change the name or setup different Patterns
http://www.ks-soft.net/hostmon.eng/mfra ... tterns.htm
I don't think there is "the best naming convention" just like the is no the best color scheme. However we recommend to use unique test names, so test name should include at least name/address of the target server

Should I be worried about the total number of tests and/or serverload?

Lets HostMonitor worry about this. You may use menu View -> Auditing Tool to check for possible troubles. If you are using old version of HostMonitor, there is Estimated Load dialog instead, it does not offer all functionality of Auditing Tool however it can warn you when server load is too high.
http://www.ks-soft.net/hostmon.eng/mfra ... tm#LoadDlg
You may setup many thousand test items, just make sure you do not setup too short test interval for them.
E.g. if you setup 2000 test items to be performed every 5 min, its less than 7 tests per second. No problem at all. While 500 test items with 10 sec test interval will cause 50 test probes per second. Not a big problem as well but as you may see system load will be 7 times higher while you have 4 times less test items.

How do we define alarms and tests on redundant/double components or load-balancing servers?

Depends on your needs. Do you want to check each node and start alert when any node does not respond? Do you want to start alert when all nodes do not respond? Do you want to start different alerts when 1 node/all nodes does not respond?

Where do we save our own definitions for profiles, mail templates, HM-options?

HostMonitor saves the files when necessary. You should not worry about these files unless you want to create backup (good idea) or move HostMonitor configuration from one system to another
http://www.ks-soft.net/cgi-bin/phpBB/vi ... php?t=1099

Regards
Alex

G. Dieltiens · Post by **G. Dieltiens** » Tue Jul 21, 2009 1:04 pm

Alex, thank you for your reply.

I know Host Monitor is very flexible and can be setup in many ways, and that it all depends on what we want to monitor, and what we want to happen when things go wrong.

I guess what I was really asking was: has anyone developed a methodology for monitoring a typical (Windows) Server Farm, preferrably based on Host Monitor? Is so, would you be willing to share it?

Thanks,
Geert.