[Ocfs-users] 15 minute favor?

Jeremy Schneider jer1887 at asugroup.com
Thu Mar 4 16:33:34 CST 2004


Anyone have 15 minutes on a development installation of ocfs 1.0.9-12 to
do me a favor?  Follow these 8 steps and see if you have the same
problem I'm having?  It doesn't look to me like a hardware issue but if
someone else does NOT have the same problem then I can rule ocfs out.

1. install ocfs-2.4.9-e-enterprise-1.0.9-12 and 1.0.9-12 tools
2. create a new filesystem
node1# mkfs.ocfs -b 128 -F -g 0 -L dc1:/u02 -m /u02 -p 755 -u 0
/dev/sdc
3. mount the fs
node1# mount /dev/sdc /u02 -t ocfs
4. make a subdir, change into it
node1# mkdir /u02/test
node1# cd /u02/test
5. fill the directory with 406 70M files
node1# /usr/bin/time -v /bin/sh -c '
for ((X=1; X \<= 406; X=((X + 1)) )); do
echo test$X.dbf;
dd if=/dev/zero of=./test$X.dbf bs=131072 count=560
done
'
6. mount the volume on THE OTHER NODE
node2# mount /dev/sdc /u02 -t ocfs
7. create a file in the test dir (from either node)
node2# /bin/echo test > /u02/test/file1
8. create a file in the test dir from the opposite node as step 7
node1# /bin/echo test > /u02/test/file2

step 8 causes my system to hang. if you try to "kill" the process (from
another login), it will get stuck in a "D" disk wait state. if you go
back to the first node and delete the first file you created then all
hanging processes on the second node should resume.  is anyone able to
reproduce this error?  following these steps seems to consistently
generate the error on my system.

many thanks,
Jeremy




More information about the Ocfs-users mailing list