Blueshell
February 4th, 2012, 04:37 AM
update-manager just blew up on me because it claims it ran out of disk space while trying to install linux-headers-2.6.32-38. The complaint appears to be totally wrong, but I'm now totally stuck. Can anyone help? Is the package perhaps defective?
The machine in question has everything (including /boot) in one partition, which currently has 2.1 GB free. I verified that I could write all of that by doing "dd if=/dev/zero of=ZERO" and dd reported writing 2.1 GB, and the file was that size. (I deleted it again, of course.) Just to be on the safe side, I deleted about 300 meg of other files and also did "apt-get clean"---no change in behavior (except perhaps blowing out at a slightly different place in the directory tree it's trying to create).
Yet I cannot finish updating my packages because both update-manager and synaptic instantly blow up claiming "unable to create `/usr/src/linux-headers-2.6.32-38/arch/arm/mach-ixp23xx/include/mach/uncompress.h.dpkg-new [blah blah]': No space left on device."
I even tried doing "while [ 1 ]; do echo "$(date) $(df / | grep sda)"; sleep .1; done" in another window while the attempted update was running. Used space went from 79% to 80% for a fraction of a second, right when synaptic blew out. No way did something write 2.1 GB---it took dd almost 3 minutes to do so (it's a slow disk). Unless perhaps there's something busted in that rpm that tries to write an enormous sparse file or something---how can I find out?
/usr/src/linux-headers-2.6.32-38-generic exists, but -not- the version without -generic, and the former currently has a zillion broken symlinks pointing to the nongeneric directory, but that never got created.
More to the point, how can I get -out- of this fix? Right now, synaptic claims that linux-headers-2.6.32-38-generic is broken (probably because it depends on 2.6.32-38, which blew up), and I don't even if the system is safe to reboot! An initrd was written into /boot and lots of /boot/grub got updated -before- it blew out, and I can't afford to guess wrong on this---this is a server that I cannot get physical access to except in extreme circumstances.
Help?
The machine in question has everything (including /boot) in one partition, which currently has 2.1 GB free. I verified that I could write all of that by doing "dd if=/dev/zero of=ZERO" and dd reported writing 2.1 GB, and the file was that size. (I deleted it again, of course.) Just to be on the safe side, I deleted about 300 meg of other files and also did "apt-get clean"---no change in behavior (except perhaps blowing out at a slightly different place in the directory tree it's trying to create).
Yet I cannot finish updating my packages because both update-manager and synaptic instantly blow up claiming "unable to create `/usr/src/linux-headers-2.6.32-38/arch/arm/mach-ixp23xx/include/mach/uncompress.h.dpkg-new [blah blah]': No space left on device."
I even tried doing "while [ 1 ]; do echo "$(date) $(df / | grep sda)"; sleep .1; done" in another window while the attempted update was running. Used space went from 79% to 80% for a fraction of a second, right when synaptic blew out. No way did something write 2.1 GB---it took dd almost 3 minutes to do so (it's a slow disk). Unless perhaps there's something busted in that rpm that tries to write an enormous sparse file or something---how can I find out?
/usr/src/linux-headers-2.6.32-38-generic exists, but -not- the version without -generic, and the former currently has a zillion broken symlinks pointing to the nongeneric directory, but that never got created.
More to the point, how can I get -out- of this fix? Right now, synaptic claims that linux-headers-2.6.32-38-generic is broken (probably because it depends on 2.6.32-38, which blew up), and I don't even if the system is safe to reboot! An initrd was written into /boot and lots of /boot/grub got updated -before- it blew out, and I can't afford to guess wrong on this---this is a server that I cannot get physical access to except in extreme circumstances.
Help?