Scratch location unavailable = ESXi meltdown

[scerazy]

If the sctarch locaction is configured to point to LUN, and this LUN disappear, ESXi 5 gets a fit & disconnects from VC (and can not be administered any more)

The VMs still run (can be acceessed directly), but they can NOT be migrated

esxcli on the server does NOT work any more (it gives Connection Failure to localhost)

The only option is to reboot the affected ESXi (switching off the VMs first by their own means)

"Tested" it on 3 ESXi

I think there should be a failsafe build in, so if the locaction is not there any more then it reverts back to /tmp/scratch

Seb

[buckmaster]

Not sure if this is the same as you experienced, but similiar.  I relocated /scratch to a VMFS volume thinking I was following best practice to save log files.  Received an email alert from my SAN I lost a drive, things still running but degraded performance as expected.  Shutdown VM's, ESXi and shut down SAN to replace drive, SAN is fine.  But I can not reconnect to iSCSI volumes?  esxcli is failing, df fails, iscsid is failing on connecting.  I have powered off esx and restarted but no good.  Can ping iscsi address both ways but unable to connect to iscsi volumes.  Running vSphere5 update1.