What if HA didn't work?

[eagleh]

Environment: Two ESX3.5 hosts in one cluster, Vcenter server is a vm itself on one host. Both hosts connected to SAN.

Lets say, what if one host is gone, and HA didn't kick in for some reason, so the vms on the failed host were not shutdown gracefully or still alive but not accessible from network.

1> How to restart vms on the failed vms on another host?

2> What kind of challege I may come across?

Just want to be prepared for the worst. Thanks in advance.

[khughes]

If your host goes down and HA doesn't kick in for some reason you might have a little bit of residue on the VM's but you will be able to power them back on with the other working host. We recently caused a HA event but HA wasn't able to control everything due to the size of the event. The thing we ran into is that my boss powered down one of the hosts possibly while VM's were still running. I was able to restart them on a different host just fine.

1) Lets say your VC was on host1 and it fails. You would just need to login directly into host2, add the VM to the inventory and power it back on. Same for any other VM

2) You might come across some VM corruption but I wouldn't expect too much. We've done some stupid things here and our VM's are still up and running Just make sure you can login directly to the other host to manage the other VM's and restart the failed ones.

• Kyle

[eagleh]

Kyle, thanks for your quick response.

I guess my further question is what would happen when the lost host1 came back live. Would it mess up the database on VC? For example, the VC was on host1, now restarted on host2. thought? How VI handle the vmfs file locking?

[khughes]

Since you're using a SAN, it wouldn't be an issue. Host2 will have the file lock and virtual center will see what VM's are running on what hosts. Virtual Center will go right back to managing all the VM's and the hosts.

• Kyle

[Texiwill · Accepted solution]

Hello,

When VC starts and HA is enabled it goes out and checks the power state of each VM. If the VM is powered on nothing happens. If powered off, nothing happens as well.

HA handles this at the time of the event. So if VC is a VM and that host dies, it will boot VC on the other host. Now if VC does not boot, then that is still not an issue as VC will just get the state of the VMs and not necessarily start any VMs.

If HA fails however you will need to determine why, ensure there is no split-brained VMs (i.e running on both hosts), etc. I would create a DR/Business COntinuity document that covers all these cases. VMware HA failing is just one case.

Best regards,
Edward L. Haletky
VMware Communities User Moderator, VMware vExpert 2009
====
Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.
Blue Gears and SearchVMware Pro Blogs -- Top Virtualization Security Links -- Virtualization Security Round Table Podcast