Isolation Response Clarification

[hutchingsp]

We're replacing our core LAN switch tomorrow. We have connected to it our two server HA cluster. The Isolation Response is the leave VM's powered on.

Do I need to do any more to ensure that when we disconnect the ESX boxes from the switch the VM's will stay on?

Granted without the switches nobody can access them anyway, but the point is obviously this is planned so I'm happy for the VM's to stay up/on.

Cheers!

[Troy_Clavell]

my advise to you would be to break your HA cluster during network maintenance. I have seen too many issues when all hosts go into isolation causing VM's to become orphaned, partially registered on other ESX Host, invalid.

If, you are not losing total network connectivity, you may think about using some advanced options in HA to maybe use your VMotion NIC as an isolation address, increase the isolation from it's default 15 seconds, or create a second COS.

[hutchingsp]

That seems a good option - like I said nobody is expecting those servers's to be available as it's literally our core LAN infrastructure that's being swapped out, all I want to do is make sure ESX stays happy and for me the simlpest thing to do is just leave it and all the VMs running.

Thanks.

[Troy_Clavell]

If no one is expecting your VM's to be available, I would just shut them down, break HA, shutdown the hosts, do you network maintenace, bring hosts back on-line, enable HA again and power on all your VM's.

To me that seems to be the safest route.

[hutchingsp]

Hmm I guess I could look at updating to 3.5 U3 whilst I'm about it as it's a ten minute thing if I'm bringing the boxes down anyway.

I'll look into that though as whilst it should be a "next, next, next, wait, done" job I don't want to rush it for the hell of it.

I'll do some digging.

[titaniumlegs]

Probably a bit late, as tomorrow is today, but...

Best practice is to have redundant service consoles on separate networks. If you have IP storage (NFS or iSCSI), you can put a service console port there. In advanced options, set an additional isolation address (or two) pointing to the back end switch or gateway (if the storage network has one, which it probably should not) or the IP of the storage device(s). To do this , right-click the cluster --> Edit settings --> VMware HA --> Advanced Options. Add entries like

das.isolationaddress1 | 192.168.42.201

das.isolationaddress2 | 192.168.42.202

If the back end network is going down, too, your best bet is to shut everything down.

If it's all FC SAN, then think about connecting a cheapo switch to an extra port on each ESX server and build a console port on that.

Enjoy!