Hi,
I just set up a multiple node configuration with 1 MDS, 2 OSTs and 1
client. I gave the following for striping in the configuration file:
${LMC} --add lov --lov lovA --mds mdsA --stripe_sz 1048576 --stripe_cnt
0 --stripe_pattern 0
Now does it not mean 1 file will be striped across all OSTs
If I create a file of 1 GB should it not be striped across both the
OSTs.
but,
I get this when I use lfs
[root@srl-lustre1 lustre]# ls -lh
total 12G
-rw-r--r-- 1 root root 1.0G Jul 25 12:28 file1
-rw-r--r-- 1 root root 10G Jul 25 12:57 file2
-rwxr-xr-x 1 root root 25K Jul 25 12:08 fileop
[root@srl-lustre1 lustre]# lfs find -r --obd ost2_UUID /mnt/lustre/
/mnt/lustre/fileop
/mnt/lustre/file2
[root@srl-lustre1 lustre]# lfs find -r --obd ost1_UUID /mnt/lustre/
/mnt/lustre/file1
should''nt file2 and file1 have objects over both OSTs.
am i interpreting striping incorrectly?
Thanks
Priya
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://mail.clusterfs.com/pipermail/lustre-discuss/attachments/20060725/a5c86dc8/attachment.html
On Jul 25, 2006 13:55 -0700, Priya Muralidharan wrote:> I just set up a multiple node configuration with 1 MDS, 2 OSTs and 1 > client. I gave the following for striping in the configuration file: > > ${LMC} --add lov --lov lovA --mds mdsA --stripe_sz 1048576 --stripe_cnt 0 --stripe_pattern 0"stripe-cnt 0" means "use default striping" and in the current case this will be 1 object per file.> Now does it not mean 1 file will be striped across all OSTsNo, if you want each file to stripe over 2 OSTs use "--stripe_cnt 2". Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.