eftychia: Me in poufy shirt, kilt, and Darth Vader mask, playing a bouzouki (vader)
Add MemoryShare This Entry
posted by [personal profile] eftychia at 09:20pm on 2007-08-15

[root@richards ~]# fsck /dev/hdc
Parallelizing fsck version 1.14 (9-Jan-1999)
e2fsck 1.14, 9-Jan-1999 for EXT2 FS 0.5b, 95/08/09
fsck.ext2: Attempt to read block from filesystem resulted in short read while trying to open /dev/hdc
Could this be a zero-length partition?
[root@richards ~]# dd if=/dev/hdc | od -x | more
dd: /dev/hdc: Input/output error
0+0 records in
0+0 records out
0000000
[root@richards ~]# fdisk /dev/hdc

Unable to read /dev/hdc
[root@richards ~]#

Oh crap. /dev/hdc is usually mounted to /home on the file server. Ooooooh, f___ ...

There are 4 comments on this entry. (Reply.)
ext_4541: (Default)
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.

Failing that...how's your backup?
 
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. :-(
 
/dev/penis: no such device or address


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.)

Links

January

SunMonTueWedThuFriSat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24
 
25
 
26
 
27
 
28
 
29
 
30
 
31