[4440] in Hotline Meeting
unhappy
daemon@ATHENA.MIT.EDU (David Krikorian)
Fri Apr 5 20:04:40 1991
Date: Fri, 5 Apr 91 20:04:15 -0500
From: David Krikorian <dkk@ATHENA.MIT.EDU>
To: hotline@ATHENA.MIT.EDU
Reply-To: dkk@mit.edu
The machine happy.mit.edu is unhappy:
------------
Apr 5 04:42:03 happy vmunix: uda0: soft error, <4>unknown error, unit 0, format 011, event 0350
Apr 5 04:42:03 happy vmunix: ra0a: hard error sn7176 pid 4153: exec: I/O errormapping pages
Apr 5 05:42:00 happy vmunix: ra0a: hard error sn7176 pid 4166: exec: I/O errormapping pages
[then a reboot at about 7am -dkk]
Apr 5 07:26:53 happy vmunix: ra0a: hard error sn7176 ra0a: hard error sn7176 pid 76: exec: I/O error mapping pages
Apr 5 07:26:54 happy vmunix: ra0a: hard error sn7176 pid 80: exec: I/O error mapping pages
[followed by 15 more of pretty much the same message]
------------
Tracking down the bad sector, I found that it was in the file
/bin/awk. (Didn't this happen to another server just a few weeks
ago?)
------------
happy# /etc/icheck -b 7176 /dev/rra0a
/dev/rra0a:
7176 arg; frag 0 of 8, inode=514, class=logical data block 12
bad mode 1350
bad mode 1351
files 688 (r=190,d=26,b=58,c=163,sl=251)
used 12707 (i=26,ii=0,b=1478,f=675)
free 2192 (b=268,f=48)
missing 0
happy# /etc/ncheck -i 514 /dev/rra0a
/dev/rra0a:
514 /bin/awk
------------
So I renamed the copy of awk that lives on the bad spot of the disk,
and copied in a new one.
------------
happy# attach -h -n m37-vssys
attach: filesystem m37-vssys (socrates.MIT.EDU:vssys) mounted on /srvd (read-only)
happy# mv /bin/awk /DONT_TOUCH
happy# cp -p /srvd/bin/awk /bin/
happy# detach /srvd
detach: m37-vssys detached
------------
The bad spot will remain a danger (in case /DONT_TOUCH is removed)
until rqbads is run on the machine. The fact that one bad block was
found also suggests that there may be more disk trouble hiding, or
about to happen.
...
I just checked the rest of the disk. There was only one bad block
(7176). The next update will presumably remove /DONT_TOUCH, so rqbads
should be run before then, to fix it.
