HA VM monitoring, how to test?

[dpinedo]

Hello,

We have 2 ESXi 5.1 in a cluster with HA enabled.

We have a software suite distributed in several VM's and we'd like to test how the whole behaviour is affected when some of the VM (node) crashes.

The HA is working properly: When I shut down or reboot one of the ESXi, the VM are Migrated to the other ESX.

But we'd like not to have to shut down the entire ESX to test a node (a node of our software suite) crashing. Most of the nodes are linux systems (CentOS). We have tried:

     1)Powering off the VM from the vSphere client and from the VM itself --> But it's not restarted (which I suppose is correct, as it is not a crash, is a voluntary shutdown).

     2)With a fork bomb, but nothing happens: The VM gets frozen, but it isn't restarted.

Any suggestion will be wellcomed.

[a_p_]

Welcome to the Community,

only a guess. Since VM monitors VMware Tools as well as the VM's I/O, you may need to somehow force the guest OS to panic in order for HA to consider the VM as dead.

André

[maishsk]

Have you tried this?

Cause a Linux Kernel Panic or a Windows BSOD

echo c > /proc/sysrq-trigger

[dpinedo]

I have tried with:

cat /dev/zero > /dev/mem

and it worked for me in CentOS: The VM is immediatelly rebooted

In this post there is more information about it:

http://archive.geekworld.co.za/node/277

Thank you all

[dpinedo · Accepted solution]

cat /dev/zero > /dev/mem

Generates a CPU fault, which causes the system to reboot (at least in vSphere; in VMWare Workstation ask to reboot or to shootdown), so it's not a valid option.

Finally I forced a "kernel panic", the guest gets frozen and after a few minutes the HA system reboots the guest.

Forcing a Linux kernel panic | Borgström