HP VM resource monitoring

[Trojan36]

I am looking for a good solution to effectively monitor the resource usage (memory, CPU) on each Integrity VM. Let’s say I allocated 4 vCPUs and 16 GB memory to a VM, I want to be able to see it’s usage over a period of time and also pinpoint on a particular time slot. And if we can collect the data to a single place and run graph for each VM, that would be a plus.

Thx in advance

[Olivier Masse]

If you have the Virtual Server OE, it comes with a licence for Openview Performance Agent. You can configure it to monitor each VM as an OVPA "application" by defining hooks in /var/opt/perf/parm and you'll be able to see it in Glance, but as a bonus it will gather data on its CPU and disk usage, etc. See the best practices whitepaper that documents this:
http://docs.hp.com/en/9983/BestPractices2.2.pdf

You can then use "extract" to get data from OVPA and send it to textfiles but it's not easy to use. If you're lucky enough to have an Openview Performance Manager licence, you will be able to plot from one central location advanced graphs covering each VM over a long time period. It is way better than the built-in performance charts that are available with ESX. While it's a very useful tool, on the downside, OVPM is quite pricey. If you go this way, I'll monitor this thread, just ask questions and I'll do my best to answer them.

I think there is a SIM-based product that offers similar functionality, but I don't know much about it as I already have OVPM (and don't like SIM that much).

As a free, quick and dirty alternative, you can script something around hpvmsar that will log limited performance data in text files. It's then easy to feed it in Excel as CSVs to make simple graphs.

Good luck

[Emil Velez]

In addition

The measureware agent on the VM host can display and collect metrics on the virtual machines it has.

Glance has a logical system report but measureware (the performance agent) does not by default collect the VM metrics but you can let it. Use the following cookbook on your VM host.



#vi /var/opt/perf/parm

add logicalsystem to the log line

log logicalsystem global application process device=disk,cpu,filesystem, transaction

# utility -xp (to check the syntax)
# ovpa restart

After about 15 minutes you should have data especially if you do this before you eatcpu program

# cp /var/opt/perf/reptall /tmp/ls.rpt
# vi /tmp/ls.rpt
uncomment out all of the LS class metrics or else you only get 4 of them.

# extract -xp -iI -r /tmp/ls.rpt -f /tmp/systems.txt

# /usr/dt/bin/dtpad /tmp/systems.txt