I came across a minor but annoying problem today after performing ESXi upgrades in my lab using VUM. Each of my ESXi hosts have a local 8GB boot disk (non-flash), and at install a local 4GB FAT32 partition was created automatically for scratch. However for some reason after upgrading the hosts to 6.0 U2 (from U1) this partition was removed for some of my hosts (but not all), therefore showing this ugly warning message in the web client for the hosts:
The fix was simple. In the web client:
- Create a new VMFS datastore on the boot disk (I made it 3 GB). E.g. “esx-01b_local”
- Create a new directory on the datastore for scratch. E.g. “.locker-esx-01b”
- Change the advanced property “ScratchConfig.ConfiguredScratchLocation” to the VMFS volume that was created. E.g. “/vmfs/volumes/esx-01b_local/.locker-esx-01b”
- Reboot the host (obviously make sure you’re in maintenance mode first!).
The reasoning for this occuring I will leave to the fact that my lab is running in a nested environment, which caused the ESXi upgrade process to be very slow (each ESXi upgrade took almost 30 minutes to complete).
Please note that this applies to local non-flash storage. If you are installing ESXi on Flash/SD cards, the above fix is not recommended due to the fact that Flash and SD cards can have potentially limited read/write cycles available (and having scratch on them can wear them out pretty quickly). In this case, you should use either a seperate local datastore that is on a HDD/SSD (but not VSAN, as VSAN is not supported), or a remote syslog server.