I've got an embedded ESXi installation (installed on a USB thumb drive plugged into the internal USB port) on a Dell T710 with a few of SAS drives. I tried upgrading from 4.0-294855 to 4.1-348481 using the instructions here via:
esxupdate --bundle=/vmfs/volumes/datastore_name/path/to/upgrade-from-esxi4.0-to-4.1-update01-348481.zip update
Although it completed successfully, when I rebooted I received a purple screen with this message on it:
The system has found a problem on your machine and cannot continue.
Two file systems with the same UUID have been detected. Make sure you do not have two ESXi installations.
So, it tried updating one of the internal SAS drives versus the internal USB thumb drive...?
I researched this once before & found a thread, where posted for help, but it wasn't enough to get me going. This evening I found KB1035107 that details how to fix this but I've got some concerns:
Last, and perhaps more important than these other questions:
Is there a way for me specify which disk/drive/device to target when performing upgrades, specifically 4.0 to 4.1?
How I can perform upgrades in the future without having worry about this nonsense? I don't want to have to disconnect datastores/LUN's each time there's an upgrade (major or minor).
Many thanks.
Does the HD with the datastore show other partitions if your run fdisk -l?
Bump...?
Either no one has a solution or I've missed something ridiculously simple and am being shunned as a result.
You can use the option from the KB article to get the system booted. Once you have it up and running you can go to Configuration > Storage to find out where the 2nd install is located. Just make sure to set the correct boot order in the host's BIOS to make sure you're booting the embedded install.
Thanks for taking the time to respond Dave. I sincerely apologize for the delay.
With your guidance I was able to figure out which drive appears to have the other install. Apparently its the same drive as 'datastore1'. So eitther I goofed when setting up the box, or the bundle update may have goofed?
So basically I'll need to move the VM's and other data over to another datastore then format the drive & try again?
Does the HD with the datastore show other partitions if your run fdisk -l?
Is it possible that you have multiple paths to the same storage, and don't have proper multipathing setup (e.g. PowerPath etc) - so the same storage is being seen twice?
Thanks to all who took the time to look into this and respond.
I took a look at all the other drives on the machine and sure enough one of them had new partitions. (Partitions that weren't there when I set the drive up.) Some how, it seems, when doing the upgrade instead of partitioning the embedded USB drive it partitioned a random drive.
Following the KB article I was able to boot the server into the older version, migrate data off that drive then reformatted it
vmkfstools --createfs vmfs3 --blocksize <B> /vmfs/devices/disks/<drive>
I'm still concerned however about esxi arbitrarily targeting random drives vs the USB drive.
Again, thank you all for your support.
It would be impossible for an ESXi upgrade to have added partitions to an existing datastore drive since doing so would have wiped out the existing virtual machines. There must have been an ESXi install on the drive at some point. I don't think you need to worry about USB upgrades.