The culprit was finally found. Prior to adding the 4 new nodes I
upgraded to pvfs2-1.5.1 from 1.4.2. This required me to run pvfs2-
migrate-collection which made a backup of the pvfs2 data by creating
hard links in another directory. So when I removed the original file
the bstream files from the backup remained and since the reference of
the inode wasn't 0 the backup data remained. Runnging pvfs2-migrate-
collection --all --clean on all the original nodes freed up lots of
space. So now I have the rebalance running again.
On Aug 28, 2006, at Mon,Aug 28, 1:20 PM, Julian Martin Kunkel wrote:
Hi,
Maybe the old datafiles are not deleted properly ? Maybe we could
try the
following to figure out if that is the case, choose a file on the
old setting
(4 servers) (or create a new one lets say 100MByte) printing out
the real file
names in the storage space with pvfs2-viewdist, in addition do a
statfs. Then
balance this file on the new setting with 8 servers. Then look at
the statfs
again. Also check the old bytestream files in the dataspace again.
Are they
still present ?
I think it might also be interesting to run the normal UNIX df
command before
and after the migration on the server which should delete the file
to verify
these results.
Julian
_______________________________________________
Pvfs2-developers mailing list
Pvfs2-developers@beowulf-underground.org
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
_______________________________________________
Pvfs2-developers mailing list
Pvfs2-developers@beowulf-underground.org
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers