Rawdisk VMDK I/O Errors

[drum]

I have an Ubuntu Host, with an Ubuntu Guest running, the Guest has a 4Tb Sata Hard Drive Mounted as a VMDK raw disk (secondary disk). The setup runs quite happily for 6+ hours, then the drive appears to crash and go read only. I initially thought it was a faulty hard disk, but I see the same fault with a second disk, and checking on the host service, no errors are recorded for disk io.

Rebooting the guest does not solve the problem, but restarting the VM does.

Anybody any ideas, what could be causing the problem?

[scottgus1]

Though I can't necessarily help diagnose so far, you might clarify one thing:

the drive appears to crash and go read only.

Crashing and going read-only are two different things, which one is it?

Additionally, we'd need a guest vbox.log where this happens. Start the guest from cold boot, not save-state, run until the problem happens, then shut down inside the guest OS, then zip and post the log using the forum's Upload Attachment tab.

Is it always 6+ hours?, never 5 or 1 for instance?

[drum]

Okay the mount goes read only, which causes my software to crash, not the vm itself.

Never seen it do it in under 6 hrs, so the log files could be huge, currently been running 9hrs, without a hiccup. Typical.

Next run, I'll capture the logs.

[mpack]

Most of the log is written while starting up and when shutting down, so size wise it probably won't be so bad. Just make sure to zip the log before attaching it, as we deliberately enforce a quite modest attachment size limit. And make sure the VM is shut down before you grab the log.

[socratis]

I remember that the Linux kernel mounts drives as read-only when it detects long delays and timeouts. It (the kernel) believes that there's no other explanation for the timeouts, except that the drive is going bad, and in order to prevent further corruption it "pulls" a read-only on the disk.

A proper VBox.log (ZIPPED) will contain the info...