Vmotion and potential problems that can occur

[millardus]

Hi all,

As part of my rapid introduction to VMWare 3.5 Patch level 4 I am building up a bank of knowledge and documents alongside studying to get myself in the bets postion to support it.

This question is one I would like to firm up. I am currently updating all my ESX hosts and in the process migrating VM's around constantly. I have no test environment so everything is "seat of the pants" stuff. I appreciate VMotion is reliable and I must say I'm hugely impressed with it, but what scenarios would there be where VMotion could go wrong and what impact could this have on the VM's or Host ESX boxes?

I supose a lot of this would come down to understanding the entire VMotion process and how it works which I assure you I am reading up on, but a brief list of potential issues and what to look at to fix them would be of great benefit to me while I move these VM's around and hope nothing goes wrong.

Thanks a lot

Tom

[a2alpha]

If vMotion fails then the impact on the hosts and vms could be:

• DRS can't migrate your virtual machines to less used hosts, so a host could get resource contention if all the vm's want loads of resources. If this happens then your vms might not perform as well as they could.

• You wouldn't be able to manually or otherwise migrate the vms off to enter maintenance mode for patching or other work on the hosts.

HA would still work as it does not use vMotion.

The main thing we see vMotion failing on is DNS issues, so always a good start. Also check connectivity of your vmkernel port to other vmkernel port ips from the commandline using a vmkping <ip address of another vmk>

Other impacts on hosts / vms depends on what caused vMotion to fail, this list is endless though like has host has lost its network card etc...

Hope this helps,

Dan

[millardus]

Thank you Dan, greatly appreciated response. It gives me topics to read up on.

I supose my main question revolves around whether the VM would normally still be in a working state if VMotion failed. i.e. let's say it turns out to be DNS issue because one of our AD DC's is rebooted and doesn't come back up, and at that moment in time I am migrating a VM, the migration gets to 60% and then experiences a problem because of the DC. Would the VM just be available once again on source or would it likely need a reboot etc

Tom

[a2alpha]

Almost entirely if the vMotion fails during the migration then the vm remains on the source host from my experiences. If DNS failed half way through, as all the paths etc are already setup this would not affect it. vMotion wouldn't start the majority of the time there is something wrong. I haven't seen it actually affect the running of the underlying guest o/s. I suppose that if at the final switch over at 90 odd percent one of the hosts died, there would be an issue, but then you have bigger problems!

Hopefully this helps,

Dan

[a2alpha]

You might find the Availability Guide useful:

This is for version 4 but there is also one for 3.5.

Dan