1st, check the physical layer...uninstall, check the data and power pins and receptacles for bends, breaks, debris, even corrosion [eep].
Sniff the circuit board--any magic smoke get out recently?
put it all together again, sometimes doing nothing else but unplugging and plugging back in makes it all better..
failing that, you could be suffering from degenerative lubrication failure, an increase in the static friction coefficient, affectionately known as "stiction."
So if the replug fails, sometimes a good solid WHACK will get you back in business. Seriously. Scary, but true.
First, despite the device name, this really isn't a separate spindle. /dev/hdb1, /dev/hdb6, /dev/hdb7, /dev/hdb8, /dev/hdb9, and /dev/hdb10 are all on the same physical drive as /dev/hdc, which has a different major device number. No, I haven't any idea how on Earth I managed to make that happen several years ago when I moved the drive to its current location.
The other partitions were all working fine; only /dev/hdc failed.
Even more strangely, I could mount /dev/hdb5, which contained a copy of /home from 2002.
"1st, check the physical layer...uninstall, check the data and power pins and receptacles for bends, breaks, debris, even corrosion [eep]."
0th: clear a path through the assorted unsorted Pennsic stuff to reach the server ...
I unplugged things and plugged them back in. I couldn't see damage in that light but I'll look again in daylight.
"Sniff the circuit board--any magic smoke get out recently?"
Not that I could tell, but a) the drive was awfully warm, and b) my beard still smells of wood smoke, which is probabbly throwing my sniffer off.
"put it all together again, sometimes doing nothing else but unplugging and plugging back in makes it all better.."
Alas, after unplugging/replugging everything:
Now that box can see /dev/hdc but can't see /dev/hdb*. And 'dd if=/dev/hdc | od --strings=6 | more' found Stuff (but also kicked "lost interrupt" and "drive not ready for command" errors to the console). And, fsck hung.
"failing that, you could be suffering from degenerative lubrication failure, an increase in the static friction coefficient, affectionately known as 'stiction.'"
*nod* A great guess except for the (unknown to you until just now) fact that other partitions on the same spindle had been working fine until I powered down and unplugged things.
"Failing that...how's your backup?"
Uh, embarrasing. If I can't resuscitate this partition (and now /shared as well), this is gonna hurt a lot.
I had been planning to replace the nearly full drive with a larger one and get a free backup from the process of copying the files over, but month after month I come up just a little short of being able to buy a bigger drive. :-(
Sounds like a disk label/partition issue or perhaps a bad superblock (depending on how these filesystems are accessed). I'd image the whole drive and then try to dope out the original geometry and make a new label. Failing that, scan for superblocks and try to reverse engineer a label/partition map from that. Failing that, scan the section containing the missing filesystem and look for stuff that appears to be file-like (text, JPEG, PDF, and similarly structured data). This only works well if the data is mostly contiguous (i.e. unfragmented). Note that I've written utilities to do all of these things on various systems, and my file recovery tools tend to outperform even expensive payware.
There appears to be a hardware problem as well, since 'dd' sometimes gets stuff and sometimes doesn't, and there are all those "drive not ready" console messages, but I'd appreciate the use of your Linux/ext2 tools if I do manage to image the drive.
This afternoon I'll try more experiments, including moving the drive to another computer in case the IDE controller is flaky. I powered it down overnight to cool off. (And to help me resist the temptation to stay up until dawn dicking around with it.)
when a recently healthy drive suddenly disappears completely...
Sniff the circuit board--any magic smoke get out recently?
put it all together again, sometimes doing nothing else but unplugging and plugging back in makes it all better..
failing that, you could be suffering from degenerative lubrication failure, an increase in the static friction coefficient, affectionately known as "stiction."
So if the replug fails, sometimes a good solid WHACK will get you back in business. Seriously. Scary, but true.
Failing that...how's your backup?
Re: when a recently healthy drive suddenly disappears completely...
The other partitions were all working fine; only /dev/hdc failed.
Even more strangely, I could mount /dev/hdb5, which contained a copy of /home from 2002.
"1st, check the physical layer...uninstall, check the data and power pins and receptacles for bends, breaks, debris, even corrosion [eep]."
0th: clear a path through the assorted unsorted Pennsic stuff to reach the server ...
I unplugged things and plugged them back in. I couldn't see damage in that light but I'll look again in daylight.
"Sniff the circuit board--any magic smoke get out recently?"
Not that I could tell, but a) the drive was awfully warm, and b) my beard still smells of wood smoke, which is probabbly throwing my sniffer off.
"put it all together again, sometimes doing nothing else but unplugging and plugging back in makes it all better.."
Alas, after unplugging/replugging everything:
Now that box can see /dev/hdc but can't see /dev/hdb*. And 'dd if=/dev/hdc | od --strings=6 | more' found Stuff (but also kicked "lost interrupt" and "drive not ready for command" errors to the console). And, fsck hung.
"failing that, you could be suffering from degenerative lubrication failure, an increase in the static friction coefficient, affectionately known as 'stiction.'"
*nod* A great guess except for the (unknown to you until just now) fact that other partitions on the same spindle had been working fine until I powered down and unplugged things.
"Failing that...how's your backup?"
Uh, embarrasing. If I can't resuscitate this partition (and now /shared as well), this is gonna hurt a lot.
I had been planning to replace the nearly full drive with a larger one and get a free backup from the process of copying the files over, but month after month I come up just a little short of being able to buy a bigger drive. :-(
Re: when a recently healthy drive suddenly disappears completely...
/dev/penis: no such device or addressSounds like a disk label/partition issue or perhaps a bad superblock
(depending on how these filesystems are accessed). I'd image the
whole drive and then try to dope out the original geometry and make
a new label. Failing that, scan for superblocks and try to reverse
engineer a label/partition map from that. Failing that, scan the
section containing the missing filesystem and look for stuff that
appears to be file-like (text, JPEG, PDF, and similarly structured
data). This only works well if the data is mostly contiguous (i.e.
unfragmented). Note that I've written utilities to do all of these
things on various systems, and my file recovery tools tend to
outperform even expensive payware.
Re: when a recently healthy drive suddenly disappears completely...
This afternoon I'll try more experiments, including moving the drive to another computer in case the IDE controller is flaky. I powered it down overnight to cool off. (And to help me resist the temptation to stay up until dawn dicking around with it.)