[Ocfs2-users] Deleting large numbers of files hangs

[Angelo McComis]

I have a 6-node cluster, this is on SLES 10-2+updates (almost ready to
move to sp3 shortly)

I have a directory where my users uploaded about 50,000 or so files of
about 50K each, resulting in about 5-6GB of space being utilized in my
OCFS2 filesystem.

The application that was migrating to this environment, which needed
these files was being backed out, and being redeployed at a different
time. When the application team attempted to delete the files, they
did a simple "rm *" in the directory, which gave them a "too many
arguments" error, so they tried a carefully crafted find . -type f -
exec rm {}  \; statement. That one hung. By the time they got me
involved, the server was practically unresponsive, and the result of a
ps faux command showed several rm processes in an uninterruptible
sleep state.   Knowing that is not resolvable, I had to reboot the
server to clear the uninterruptible sleep.  A colleague of mine was
watching the outcome and trying different things also ended up with
processes in this state.

We were finally able to go through and remove files chunk by chunk
(e.g. rm a*, rm b*, etc.) to clear out the files.  The space, however,
is still not released in the filesystem. The 5.5GB that was consumed
by these files still shows as being consumed.

Question: I've read about unclaimed inodes and running out of space --
is this the same issue?   -- And will the SP3 kernel update (which
presumably includes new ocfs2 binaries) resolve these issues?

Thanks!

Angelo