Jack Kielsmeier
2010-Jan-11 08:06 UTC
[zfs-discuss] ZFS import hangs with over 66000 context switches shown in top
Howdy All, I made a 1 TB zfs volume within a 4.5 TB zpool called vault for testing iscsi. Both DeDup and Compression were off. After my tests, I issued a zfs destroy to remove the volume. This command hung. After 5 hours, I hard rebooted into single user mode and removed my zfs cache file (I had to do this in order to boot up, as with the zfs cache file, my system would hang at reading zfs config). Now I cannot import my pool as the box always hangs after about 30 minutes. It''s not a complete hang, I can still ping the box, but I cannot do anything. The keyboard is still responsive, but the server will do nothing with any input I make. I cannot ssh to the box as well. The only thing I can do is hard reboot the box. At first I thought I was running out of RAM, because the hang always happened right when my free RAM hit 0 (still had swap available however), but I''ve made tweaks to /etc/system and now I get freezes with over a gig of RAM free (8GB total in the box). The strange thing is, the context switches shown in top skyrocket from about 2000-6000 to over 66,000 just before the freeze. Would anyone know what that would skyrocket like that? If I do a ps -ef before the freeze, there are a normal amount of processes running. I have also tried a zpool import -f vault using a snv_130 live CD, as well as trying a zpool import -fFX vault. The same thing happens. System Specs: snv_130 AMD Phenom 925 8GB DDR2 RAM 2x 500 GB rpool mirrored drives 4x 1.5TB vault raidz1 drives I have let the import run for over 24 hours with no luck. Thanks for the assistance. -- This message posted from opensolaris.org
Jack Kielsmeier
2010-Jan-11 08:17 UTC
[zfs-discuss] ZFS import hangs with over 66000 context switches shown in top
I should also mention that once the "lock" starts, the disk activity light on my case stays busy for a bit (1-2 minutes MAX), then does nothing. -- This message posted from opensolaris.org