Advertisement
Guest User

Untitled

a guest
Oct 27th, 2013
57
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 2.56 KB | None | 0 0
  1. My headless server running Ubuntu 12.04 stopped responding to ping,
  2. and I could use some help diagnosing the problem. Here is what I know
  3. so far.
  4.  
  5. The first thing I did was plug it into a keyboard and monitor and try
  6. to boot it up. It gets to the GRUB kernel selection screen and picks
  7. the first kernel on the list automatically. After displaying the
  8. Ubuntu boot splash screen for a few seconds, it drops me to an
  9. (initramfs) prompt.
  10.  
  11. A bit about the HDDs: this machine contains 4 2TB hard drives
  12. connected via SATA. Each hard drive has a 3GB partition at the
  13. beginning, a 25GB partition following that, and a large partition
  14. taking up the rest of the device. The 3GB partitions on each drive are
  15. in a software RAID 1 configuration (/dev/md0) which was used as the
  16. swap device. The 25GB partitions on each drive are in a software RAID
  17. 1 configuration (/dev/md1) which was used as the root filesystem. The
  18. large partitions on each drive are in a software RAID 6 configuration
  19. (/dev/md2) which was mounted on /srv and used for user data.
  20.  
  21. Checking dmesg from the (initramfs) prompt showed that all of the
  22. partitions on one device had been "kicked" from their arrays because
  23. the partitions were not "fresh". My first suspicion was that this
  24. drive had died, but googling this error message turned up the fact
  25. that this can happen after an unclean shutdown and can be fixed by
  26. running
  27.  
  28. # mdadm /dev/md0 --fail /dev/sda2
  29. # mdadm /dev/md0 --remove /dev/sda2
  30. # mdadm /dev/md0 --add /dev/sda2
  31.  
  32. I ran these three commands, and found that the first two did not
  33. complete because /dev/sda2 had already been removed by the array.
  34. Perhaps this is what "kicked" meant? The third command succeeded and
  35. /dev/md0 was able to resilver quickly. I did the same thing with
  36. /dev/md1 by adding /dev/sda3 again. Checking /proc/mdstat showed that
  37. both arrays had recovered. I decided to leave /dev/md2 for later.
  38.  
  39. I tried rebooting, but it got stuck at the (initramfs) prompt again.
  40. Going back to Google, I found that a system stopping at the
  41. (initramfs) prompt can indicate a problem with the root filesystem, so
  42. I booted from a SysRescue USB flash drive and ran fsck on /dev/md1. It
  43. indicated that the journal needed to be replayed on the ext4
  44. filesystem and did that for me. I tried rebooting again, but ran into
  45. the same problem. Then I booted back into the rescue environment and
  46. ran fsck -f /dev/md1. It checked the filesystem more thoroughly, but
  47. that didn't end up helping either.
  48.  
  49. Just now I tried booting from one of the slightly older kernels in the
  50. GRUB list, but I got the same result.
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement