No, this probably has to do with the fs unable to allocate inodes
because inode_alloc for its slot is full and that it is unable to
extend it because it could not find 4MB contiguous on disk. So, this
leads to a situation where one can create files on nodes that have
free space in their inode_alloc but not in others.
We are working on a solution to allow nodes to use free inodes
in inode_allocs for other slots. Hopefully will be in 1.2.9. Once
we do that, we could look into allowing users to preallocate space
in inode_allocs for all slots. This way users will be assured that
that many inodes will be created (and more if the inode_alloc
can be extended) and at the same time not worry about wasting space
if a slot or more remain unused.
If you (or anyone else) does run into this situation, run the script
stat_sysdir.sh and email me (or file a bugzilla) the output.
http://oss.oracle.com/~smushran/.debug/scripts/stat_sysdir.sh
Sunil
Brekeveld, Feite wrote:
Hello,
We have a 3-node cluster setup with ocfs2.
Since friday one of the nodes went down and would not become clustermember
after a reboot because it was unable to write to the ocfs2 filesystem. Message:
no space left on device.
There is plenty of diskspace though.
No problem whatsoever to create a file / directory on the filesystem using one
of the other nodes.
Today one of the remaining 2-nodes got the same symptoms.
We started deleting some directories after which we were able to get the
clusternodes up-and-running again.
Is there a limit on files / directories for an ocfs2 filesystem ? I've read
something about 32000 subdirectories but with 21450 directories total we
weren't even close to that number.
Thanks,
Feite Brekeveld
_______________________________________________
Ocfs2-users mailing list
Ocfs2-users@oss.oracle.com
http://oss.oracle.com/mailman/listinfo/ocfs2-users
_______________________________________________
Ocfs2-users mailing list
Ocfs2-users@oss.oracle.com
http://oss.oracle.com/mailman/listinfo/ocfs2-users
