From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0; Q312461)
Description of problem:
I have 10 72GB 10000RPM drives connected via fibrechannel using software RAID
on an 8-CPU Dell 8450 box with 4GB of RAM and plenty of internal hard disk
space (>18GB). I setup 5 mirrors (sda1-sdf1, sdb1-sdg1, etc.), let their
mirrors sync, then RAID0 them. End result is one 350GB md device. Then I run
mke2fs on that 350GB device to create the filesystem. Around inode 250 or so,
it begins to slow down. Around inode 430, it stops altogether. The machine
isn't locked up, and it reports no CPU usage, no I/O usage across the FC
drives. mke2fs is in an uninterruptible sleep state. This is a fresh RH7.2
install, no errata, no changes from a default installation. No errors appear
in dmesg or /var/log/messages. mdstat reports no failures in the drives.
I have another 8-CPU Compaq 8500, 4GB RAM, almost identical setup, with the
same problem. This box is also RedHat 7.2 with all current (as of today)
errata in place. I have the same problem with this machine, although this
machine is in production on internal drives, so I can't test with it anymore.
Furthermore, when I was testing with it, sometimes the freeze up would affect
other I/O on the system, causing other programs to pause or hang.
Machine 1 has a QLA2200 FC card
Machine 2 has a QLA2300 FC card
Syncing of the drives in both cases worked just fine. There were no pauses or
hangups, so it's likely not a driver issue. I debated putting this under
kernel, because it smells of a kernel issue, but the direct problem is related
right now to mke2fs.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
Actual Results: mke2fs stops
Expected Results: mke2fs finishes
This has todo with kernel VM problems and newer kernels should work better
Thanks a lot,
Florian La Roche