IO error on raw disk causes VM crash

[dgasaway]

I have a number of disks attached to a virtual SATA controller as raw disks using VMDK files. The guest uses the disks in a md raid5 array (2 data + 1 parity + 1 spare). This setup has worked well for a long time. Recently, there was a read error on one of the active drives, and that caused the VM to crash (see log). Ideally, the read error should "pass through" to the guest so that it can act on it accordingly. In my case, md on the guest would have failed drive, activated the spare, and resynced the array. Only, the VM crashed so that couldn't happen. Restarting the guest would initiate a resync (because of an unclean shutdown) which would eventually hit the VERR_DEV_IO_ERROR error and crash the VM again. The attached log is from a restart attempt.

In the case of raw disks, anyway, it seems the guest should always be allowed to handle conditions such as this.

Host is Funtoo x64 stable running VBox 4.2.18 and kernel 3.10.7. Guest is Funtoo x64 stable running VBox Additions 4.2.18 and kernel 3.10.17.

[dgasaway]

Log of guest restart / IO error / crash

[aeichner]

Please reopen if still relevant with a recent VirtualBox release.