Is there a way to customize VM monitoring?

[J8kel]

My understanding of VM monitoring is that vCenter receives regular heartbeats from the VMware Tools process running inside the VM.  I have also read that the heartbeat itself includes some information about the state of the OS and that vCenter can trigger vMotion based on the contents of the heartbeat.

I would like to be able to trip a vMotion event based on the health of my process running inside the VM.  Is this possible?  Is there some kind of event I could generate, or some way to induce VMware Tools to return some information in the hearbeat that would trigger the vMotion?

Has anyone found a different way to address this?

Does anyone know exactly what information is contained in the heartbeat?  I tried doing a packet capture, but it is using ssl so I couldn't read the contents.

[rickardnobel]

I would like to be able to trip a vMotion event based on the health of my process running inside the VM.  Is this possible?  Is there some kind of event I could generate, or some way to induce VMware Tools to return some information in the hearbeat that would trigger the vMotion?

If you would like to do a vMotion when the VM Heartbeats stop coming (even if it seems a bit strange?) I think you could create an alarm on the VM and choose VM Heartbeat as a condition and as action select Migrate VM.

Does anyone know exactly what information is contained in the heartbeat?  I tried doing a packet capture, but it is using ssl so I couldn't read the contents.

I belive this heartbeat from the VM to the HA Agent inside the ESXi is done in some "secret" way and not through ordinary guest network traffic?

[J8kel]

Thanks for the pointer to the custom alarms.  You're probably right, the choices of conditions within the vm alarm triggers is probably based on the contents of the heartbeat.

My end goal is to be able to trigger vMotion in the case that the main process on my server dies and can not be restarted properly.  Perhaps I can cause a false reporting of one of the vm alarm conditions when my process dies and trigger it that way.

Does anyone know if there is a way to monitor the status of a specific process on the VM?  It could work with just checking that the process is running, or sending http poll requests to make sure there is a response.

[rickardnobel · Accepted solution]

J8kel wrote:

My end goal is to be able to trigger vMotion in the case that the main process on my server dies and can not be restarted properly.

I am actually quite unsure of one thing, what would you gain by initiating a vMotion if one process dies in the VM? This would actually not help you at all. Besides that the VM will be running on different hardware, everything inside the VM will be exactly the same - including the crashed process.

Perhaps it is not vMotion you are after, but a restart of the VM?

[J8kel]

Good point.  Thank you.