Hi.

I have a zpool made up of three 350GB Lacie USB 2.0 disks configured with a single parity RAIDZ. The zpool is 81% full and have 31 ZFS file systems in it. Every file system is snapshotted once a day and I retain one entire week of snapshots (217 snapshot in total). The scrubbing of the zpool is performed once a week and it takes more or less 5 hours. The machine is a Sun Ultra 20 M2 and both scrubbing and snapshotting are performed by cron jobs.

I don't know if it's just a coincidence, but I just upgraded OpenSolaris 2008.05 to build 93. What's happening these days is that during heavy activity, for example during a scrub or a scrub overlapped by snapshot creation, I find the system completely hung and I cannot even connect via ssh to it, and eventually I have to power it down. After some reboots I even noticed that the snapshot operation didn't even complete (I don't know about the scrub because it leaves no evidence, or at least I don't know where).

How can I try to identify the problem? I never observed this behavior before and I would expect the system not to hang, even during a disk failure, which is not even the case.

Any hints?

Any chance your system simply over heated ? The "heavy activity" clue points in that direction. Does OpenSolaris have something like Linux's sensor package ?

David

There could be chance... but the workstation works perfectly and the USB disks are external, how could this have a relationship with the OS hang? I mean: if some USB disk overheats, it's still strange that the system hangs!

If the kernel hangs, you can configure it to generate a crash dump after 50 seconds of inactivity.

Add this line to the /etc/system file:

It will be taken into account after the next boot.

Better to run the system in console (no X11) mode so you'll be able to see system messages.

When you reboot the OS, have a look at /var/crash/<hostname> for dump files.

If it's a heat issue, it almost certainly isn't the hard drives, external or not, that would be to blame. It would be the CPU. Too much or too little cpu paste, too much dust on the heat spreader. Improperly mounted heat sink. Could be lots of things, doesn't have to be the hard drives.

David

Oh, you might want to test your ram too.....

@budword:
Thanks, I'm checking if the Ultra 20 M2 has temperature sensors I can read from the OS but it seems it's not the case.

@jlliagre:
Thanks for the hint indeed.