Host still thinks it has registered VMs after reboot

C

Child of Wonder

2[H]4U
Joined
May 22, 2006
Messages
3,270
Got a 5.0 cluster running on UCS and just put a host in maintenance mode and upgraded the firmware on it. It comes up fine but doesn't reconnect to vCenter on its own. vCenter reports the user name or password is incorrect.

No big deal. I choose "Reconnect" and input the root account info. As it shows the host summary, it shows a whole slew of powered down VMs attached to the host, VMs that had been VMotioned away when I put it in maintenance mode.

Anyone seen this before? Safe to continue adding the host and vCenter will set the inventory straight? vCenter is 5.5b.

I think I may have seen this before, but can't recall. If memory serves, reconnecting ended up being fine but I can't have any downtime on these VMs.
 
You should be fine. Worst case you know those VMs arent running on that host and you can just kill them via CLI on the new host.

Disclaimer: Of course i say worst case but we all know how it goes in our line of business.
 
Where is the storage for the VM's? Plus I have seen this when VM's have the cd rom mapped to a ISO on the host storage.
 
I'm guessing it will be fine, too. But these are the client's external facing websites and can't go down so I've got to be sure.

Just freaks me out seeing all those VMs listed even though they were all migrated away when I put the host in maint mode. The host's inventory file is probably stale and I'm reasonably sure vCenter will set it straight.
 
I just went ahead and connected directly to the host, removed the VMs from inventory, and added the host. Worked fine.
 
Second host came back after the firmware upgrade with 10 "Unknown" VMs. Annoying. Even after performing a clean shut down on the host after putting it into maintenance mode the inventory file still keeps those stale entries.
 
Child of Wonder said:
I'm guessing it will be fine, too. But these are the client's external facing websites and can't go down so I've got to be sure.

Just freaks me out seeing all those VMs listed even though they were all migrated away when I put the host in maint mode. The host's inventory file is probably stale and I'm reasonably sure vCenter will set it straight.
Click to expand...

Happens sometimes. Make sure it took the update - it almost sounds like the bootbank reverted for some reason (hence the password), and it has the last known list there.

Even if it's fine, VC will just tell it to drop that crap.
 
Child of Wonder said:
Second host came back after the firmware upgrade with 10 "Unknown" VMs. Annoying. Even after performing a clean shut down on the host after putting it into maintenance mode the inventory file still keeps those stale entries.
Click to expand...

Ok, THATS screwed up. It added the unknowns to VC?
 
lopoetve said:
Happens sometimes. Make sure it took the update - it almost sounds like the bootbank reverted for some reason (hence the password), and it has the last known list there.

Even if it's fine, VC will just tell it to drop that crap.
Click to expand...

The firmware updated fine. Just odd that the first two hosts had the "incorrect password" issue and the stale VM inventory list, but the third host didn't. It popped back into vCenter automatically without any intervention.

lopoetve said:
Ok, THATS screwed up. It added the unknowns to VC?
Click to expand...

No, it didn't. Only when hitting "Connect" in vCenter, entering root creds, then it displays the host's inventory before connecting to vCenter again. When that happened to the first two hosts I simply logged into the host directly, selected all VMs, and Removed from Inventory. After that I connected in vCenter and all was well.
 
Child of Wonder said:
The firmware updated fine. Just odd that the first two hosts had the "incorrect password" issue and the stale VM inventory list, but the third host didn't. It popped back into vCenter automatically without any intervention.



No, it didn't. Only when hitting "Connect" in vCenter, entering root creds, then it displays the host's inventory before connecting to vCenter again. When that happened to the first two hosts I simply logged into the host directly, selected all VMs, and Removed from Inventory. After that I connected in vCenter and all was well.
Click to expand...

ah. For the first two, when they connected, VC should have just ignored the VMs and told the host to clear them without you even noticing. (Like when a host fails and the VMs boot with HA, when it comes back if you peeked before it connected it'd still list the VMs it ~was~ running before VC clears them).
 
lopoetve said:
ah. For the first two, when they connected, VC should have just ignored the VMs and told the host to clear them without you even noticing. (Like when a host fails and the VMs boot with HA, when it comes back if you peeked before it connected it'd still list the VMs it ~was~ running before VC clears them).
Click to expand...

Yeah, I've seen that before. When the first host did this I figured maybe it hadn't shut down cleanly. When acknowledging a pending reboot in UCS it's "supposed" to perform a soft shut down so the OS goes down cleanly. For the second host I not only put it in maintenance mode but also shut down the host from vCenter and let it completely power down before starting the UCS firmware upgrade. Same problem.

Then the third host throws a monkey wrench in the whole thing and just works. :p
 
You must log in or register to reply here.
Back
Top