On Thu, Nov 15, 2007 at 09:11:29AM -0800, Lan Barnes wrote:
During the fires, we lost power and our Mailman server (FC6) ended not cleanly. Now the image is refusing to boot -- kernel panic. I suspect that the superblock on the very-non-virtualized hard drive is corrupted.
My condolances. My personal experience is the the VMWare virtual drives break all of the ordering constraints filesystems typically use to ensure consistency.
fsck -p /dev/hda1
except without the -p. The -p is what the system will try doing on its own. If you're getting a panic, it might even be partition tables and such messed up. Do you know what kind of filesystem you have? ext3 will usually recover data, but it may not be what you want. reiserfsck, has some rebuild options, such as --rebuild-sb, and --rebuild-tree, although you would most likely get a bunch of unnamed files as a result of this.
... and so on, but probably just for the root/boot partition. Then it will tell me if the superblock is corrupt and suggest how to find alternate superblocks. I believe that this will be mkfs with a switch that says "don't do anything, just give me the information." But I'm pretty sure that fsck will tell me what to do. Then my memory is that I return with fsck -p -b <superblock> /dev/hda1
If your superblock is corrupted, you're probably hosed. Are these ext3 filesystems, or some other type. Other types will have their own way of recovering.
I write this seeking confirmation or correction. And to ask if anyone has personal experience with doing this in VMWare.
My experience with a crashed VMWare system was that I had to re-install the VM. I did recover one once, but I had to use the package system to figure out which random files on the system had gotten silently corrupted, and re-install those packages. You should backup VMWare systems as if they were real systems. In fact, it seems to be more needful. Dave -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
