infrastructure shutdown best practice

[smitty380]

I would like to know the best practice for completely shutting down all ESX hosts in our environment for some scheduled maintenance. Once all of the VMs have been shutdown and I am ready to bring down the ESX hosts, should I first put the hosts in maintenance mode or simply shut them down. I have four ESX 3.5 servers in a cluster with HA enabled.

[scerazy]

If you shutdown the LOT then no point putting in Maintenance mode, as they will stay this way when back online (which is not what you want for autostart VMs - if you have any such ones)

Just power off (from VC or from console using shutdown command)

Seb

[DCTony]

Just curious, why would you do this? It's counter to what VMWare brings you. Why not move guests off of one host, maintenance mode that host, do your maintenance and then bring it back online, repeat as necessary. This way there is no down time for your guest VMs.

[khughes · Accepted solution]

If all the VM's are already powered off and you really want to shut down the hosts, then you don't have to put it into maintance mode. Personally I don't do it all the time but most of the time I try to make sure that the host is in maint mode before I restart or shutdown just to make sure I have all the VM's shutdown and nothing powers on without me telling it to when it powers back up.

• Kyle

[smitty380]

the maintenance is being done to the server room UPS's ... no power to the building.

[DCTony]

ah, yes without a secondary site you are pretty much hosed on that one

[Troy_Clavell]

One other thing you need to consider. If you have HA enabled, before you start shutting down ESX hosts, I would disable HA. Also, DRS may want to be disabled as well.

Then when the enviornment is live again and all hosts and guests are back on-line, turn back on HA and DRS.

if you feel this or any other answer was "helpful" or "correct" please consider awarding points.

[williambishop]

Hey DCTony! Wow, I can see you're going to be present more than hhtony was

To the main question, Troy's right, turn off HA, but you'll have to turn off DRS as well, since it will start queuing the vmotions as soon as you start turning off hosts.

[Hairyman]

Your UPS panel doesn't have a bypass mode so that you can isolate the UPS and provide live power to the room whilst under maintenance ?

[khughes]

While most UPS's have a bypass so work can be done on the UPS while keeping everything online, he said the building wont have power, so I'll assume that they're goign to be working on the power lines into the UPS. Also when work is being done on the UPS and you have a window where you can bring things down, its better to have them shut down incase the power is suddenly cut and you run the risk of corruption.

• Kyle

[williambishop]

Agreed, especially if they're willing to grant him this down time, it makes more sense to bring them down, get a bit of maitenance in as well.....Nothing's worse than losing your power unexpectedly. Knowing the when of it is pretty valuable. Plus they could probably use the reboot (guests). I've got hosts that have been up for a year, I can't imagine what some of my guests are, because I migrate them off to do maitenance on the host...those things have been up a LONG while. This is your golden opportunity.