Unique, just like all the others

Found myself back on the tools this weekend, patching an ESX cluster consisting of a pair of hosts that hadn't been rebooted since the Turnbull administration.

I wish this was a joke.

Attempting to update said hosts to the latest & greatest version of ESXi 6.5 threw mysterious failures. No worries, that's why we log everything that moves.

Attempting to read the esxupdate.log file turned into quite the adventure. After 45 minutes of failing to retrieve the log via the vSphere client & SCP, I ended up SSH'd to the box trying to reading it in situ only to be presented with:

"Device or resource busy"

This is not an ideal property for a logfile to have. I've never seen a system unable to cat a logfile before.

But I have seen a system locked out of a file by another host on a shared datastore, and it does look a lot like this. But that couldn't be the case here, surely? No one would be silly enough to assign multiple hosts to the same unique storage location? Surely?

Sure enough, both hosts in the cluster are writing their logs to the same location, similar to putting two possums in the same pillowcase. Fortunately, the fix is straightforward.

The setting in question lives under Advanced System Settings. To fix it, change the logDirUnique value to true and reboot hosts. That will make the hosts use a unique subdirectory for logging.