Alarms on virtual-machine level

[GreyhoundHH]

Hi!

We are facing a problem with the predefined alarms (in this case: memory usage) in VirtualCenter 2.5. They are fine for most of our virtual machines, but we need to have a different trigger-threshold on a single virtual machine.

We can't change the values on the VM's level since the alarm is defined in "hosts & clusters". So what is the best practice to set up a new alarm for memory usage for a single machine?

What I can do is defining a new alarm for that VM, but I guess the inherited alarm will still be triggered. Or will it be superseeded by the new alarm?

Any help appreciated!

Thanks

[bfent]

Yes, the 'inherited' alarm will trigger also. You will want to disable, or delete, the default alarm(s) and create new ones where they will best serve you (Resource Pool, VM, host, etc). Unfortunately, the alarm feature is not very versatile. Hopefully, VMware will improve the alarm functionality in future releases to be more beneficial.

[GreyhoundHH]

Hi!

Unfortunately thats exactly what I was expecting and what we'll have to do as a workaround then :/

Thanks for explaining that misery. I'm also hoping, that VMware will provide a more elegant solution for this in the (near) future.

Bye

[TheBadMan82]

Hello,

I have this same problem with CPU usage.  I'm hoping someone knows the real answer to this problem.

One VM on a host has a very irregular CPU usage pattern.  It shows consistent spikes of CPU up to 100% every 20min.  The vendor said this was expected.  They also said to use the default alarms and didn't have a good solution for this either (Interesting coincidence, don't you think).

From what I can gather, I would need to delete the default, vm, CPU alarm on the vCenter level.  Then re-create that alarm on the host level for every ESX server except for the one hosting the VM I need a "special" CPU alarm for.  Finally I would re-create the base alarm on all the VMs on that host except the "special" one.

At last I can create a good alarm for this one crazy VM's cpu usage.

That seems like a lot of work.  And stupid.

The only other viable solution would be to monitor each VM individually with a standard element monitoring solution.

--

Dave

[TheBadMan82]

Hi,

I am closer to a solution.  Each object under Hosts and Clusters, has the capability for you to define new alarms on that object.  This is separate from the inheritance of alarms from the, default, vCenter level.

I am creating folders for every type of resource I want monitored the same way.  It's still a lot of work to setup, but when I'm done I should be able to monitor objects by placing them in their appropriate folder.

This seems problematic just by looking at my Nagios deployment.  Under Nagios I have multiple groups of monitors and I have objects that belog to multiple groups.

What are other people doing to solve this problem?

--

Dave

[RParker]

TheBadMan82 wrote:

What are other people doing to solve this problem?

--

Dave

What you do, external / 3rd party monitoring.  I turn off all but a few alarms (including statistics internal performance meters) because they are unreliable anyway.

for one I have found that alams on one tab will "trigger" but alarms on other pages will not, and I can clear some alarms like local storage on a host (reset to green) but can't reset a shared datastore usage...

So I gave up, disable them, and use professional level (veeam / vFoglight) which are MUCH more effective and accurate.  Haven't tried Nagios, but working on Zabbix solution..