Our news server, which has near constant disk activity, was up for a
little over 2 days when we started getting the following errors in the
kernel messages file:
Aug 2 11:00:41 roch3 kernel: EXT2-fs error (device 8/33):
ext2_check_blocks_bitmap: Wrong free blocks count in superblock,
stored = 380041, counted = 380061
Aug 2 11:00:41 roch3 kernel: EXT2-fs error (device 8/33):
ext2_check_inodes_bitmap: Wrong free inodes count in group 14,
stored = 0, counted = 1
These errors repeat numerous times until we reboot (we had more than a
screenful of them for 11:00:41 alone). The also exist for device 8/17.
It seems that while running, our spool disks are becoming corrupt. INN
then complains that the disk is full, even with df shown only 65% usage,
and we have to reboot and fsck.
We've had this happen on two different news machines. Our old machine
had an PCI AHA-2940 with a Seagate 4G Barracuda. Our new machine has a
VL AHA-2842A with 3 Seagate 1G Hawks (2 for /var/spool/news). Both have
64M of core memory. We're currently running a 1.3.12 a.out kernel, but
have had similar problems with a 1.2.6 ELF patched for the 2940.
So, what's going on and how can we fix it?
TIA,
--Andy
-- ------------------------------------------------------------- Andrew Turnquist, System Administrator, E-Znet, Rochester, NY aturn@eznet.net "IP over everything" --Vint Cerf's T-Shirt