This is a good read. The moral of the story is to never never do a hard nfs mount within the service console. If the nfs server goes away it can cause serious heartburn for the ESX host....
I've resolved the issue and did not need to restart the host. The reason why VPXA died and did not exit was that the process was waiting for a stale NFS handle. For various reasons, the Service Console had an NFS mount of a remote server that had gone offline. As a consequence, several processes including VPXA had gone into "waiting for IO" state and were never woken up again.
Since NFS mounts are totally unkillable once they go stale, I had to configure another server with the IP of the old NFS server, and create an equivalent path which I shared out to my hung service console. After restarting the nfs server, the service discovered that my mounted share was stale, and allowed a umount.
Immediately after the umount, the defunct vpxa process disappeared. And after restarting hostd and vpxa, the system came straight back into Virtual Center.
So the moral of the story? Never _ever_ do a hard NFS mount inside your service console. You *will* regret it
Post a Comment