2

I have an ESXi 5 host running 3 virtual servers. 1 of these servers uses a datastore via Software iSCSI on a separate piece of hardware.

When the two hardware servers (storage and host) experience an outage, the ESXi host is the first to boot up and fails to find the iSCSI datastore, which results in none of the virtual servers to starting. If I log into the vSphere client and rescan for datastores after both servers are up, it will find the datatore. Then I can manually start each virtual machine.

I'd like everything to boot up clean automatically when a power outage occurs. What am I missing?

As always, thanks for the help.

Ben B
  • 31
  • 5
  • What kind of server hardware? Most allow you to set a delay for auto-power on after failure. – SpacemanSpiff Dec 02 '13 at 19:18
  • The host is a Dell 610 and the Storage is a Dell 710. Are you saying I should set a delay for the 610 (host) within the bios to give the 710 (storage) time to boot into Windows? – Ben B Dec 02 '13 at 19:26
  • Try to boot in with a live cd/pendrive and see if you can see the iscsi devices on the network. You will at least know, which device has the problem (the iscsi or the esxi). – peterh Dec 02 '13 at 19:26
  • 1
    Yes Ben, exactly. Set the storage host to boot immediately, and the virtual hosts to wait say... 90 seconds. Also keep in mind though, that if this was a hard power-off, a disk-check might be forced. – SpacemanSpiff Dec 02 '13 at 19:28
  • Both of these servers are plugged into an APC UPS. When they lose power that can keep them going for about 15 minutes, then everything is powered off. If that counts as a hard power-off, then it probably forces the disk check. I guess I should set the delay on the host a little longer in that instance? – Ben B Dec 02 '13 at 19:32

1 Answers1

1

Before anything, please understand that you shouldn't be experiencing loss of power in this manner.

Do you have any ability to add some level of power protection to the environment? A basic UPS is the easiest way to deal with this...

I think you're focusing on the boot order rather than remediating the real problem; unstable power.

ewwhite
  • 194,921
  • 91
  • 434
  • 799
  • The rack has a APC UPS that can support the rack for 15 minutes or so. Our power grid is a little shaky and we sometimes lose power at night for 30+ minutes. Its after those extended outages that this problem occurs. I very recently ran a check on the UPS load and batteries. Both looked fine. – Ben B Dec 02 '13 at 19:34
  • 1
    You should set up a script to power off the ESXi host and datastore if the power stays out longer than ~5 minutes. Otherwise, you're going to get corrupted VMs at some point. – jlehtinen Dec 02 '13 at 20:32
  • @jlehtinen Exactly. $1200 can get you 1.5-2 hours of runtime these days. – ewwhite Dec 02 '13 at 20:49
  • @jlehtinen I like the idea. If its not too much to ask can you provide a link on how to do this? – Ben B Dec 02 '13 at 22:38
  • @BenB How it can be done will depend on your hardware. You could use the info in this thread for ideas and starting points, several options are discussed: http://serverfault.com/questions/462993/vmware-esxi-shutdown-triggered-by-apc-ups-connected-via-usb – jlehtinen Dec 03 '13 at 15:16