During the cleanup from a recent DR test at work, I ran into an issue where I couldn’t migrate guests between different hosts. We have three hosts in a HA/DRS cluster, so if I couldn’t move the guests around there is no way that VMware could automatically do this either.
Not exactly sure why I couldn’t do this, I figured I would reboot each host one at a time. The first host was not able to be put into maintenance mode, since the guest vMotions failed. I also realized that this was the host I was trying to get guests off of when the vMotions failed earlier. Other hosts seemed to vMotion guests between them fine, it was only an issue with “host 1.” I also noticed that I could not open the console for any guest on this particular host.
I checked in the Event Logs in vCenter and found an error similar to (I forgot to document the EXACT error message wording): RAMDISK root is full on host1. I didn’t find much on the Internet regarding this error, other than the actual ESXi OS must have a full partition and I should clean it. I wasn’t quite sure what to delete, but I did find a command to identify what part of the host was full (vdf -h). I called VMware for support and, while helpful, they didn’t show me anything I didn’t already find on my own.
I found the partition that was full was only 32 MB and there was a 30 MB /var/log/hpHelper.log (we have HP hosts) file. I copied the log file using WinSCP, then deleted it from the host, and restarted the host services. After that, vMotions to/from that host were successful and I could put it in maintenance mode to reboot it.
Summary –
– I couldn’t vMotion guests between hosts at one of our datacenter locations.
– The event log in the vSphere client showed that ramdisk root is full on host1.
– SSHing into host1 and running `vdf -h` shows that ramdisk root has 0% free space out of 32 MB.
– I looked around and found that /var/log was very large.
– I found a 30 MB file in /var/log called hpHelper.log.
– I used WinSCP to copy that file to my desktop and read it, it’s just a log file.
– I ran `rm /var/log/hpHelper.log` and `touch hpHelper.log` to delete and re-create the log file.
– I restarted the management services on the host – `./sbin/services.sh restart`
– Now I can open the console on various guests and I can vMotion between hosts.
Can you check your hosts now whether your logs are full with
2013-04-15T15:18:24.907Z cpu56:3009109)WARNING: UserObj: 675: Failed to crossdup fd 0, /dev/char/pty/t0 type CHAR: Would block
Sorry for the late reply, they were full with SNMP-based messages. The message you sent above does not look familiar.
Hi Curtis, What’s the hardware you seeing this errors on? I’m only seeing them on HP server (G8).
The errors were on HP hosts, I believe they were G8s.
I notice if I flash the ILO firmware (1.22) the problem stops.. I’ll monitor and keep you posted.