OCFS2 showing "No space left on device" on a device with free space

Starting with SLES 11 SP2 a new feature has been introduced to avoid this issue. This is called discontig-bg (Discontiguos Block Groups). Further info and steps to enabled this feature are documented in the mkfs.ocfs2 and tunefs.ocfs2 manpages. For example (we will always use the device OCFS2 /dev/sdc1 as example here):

# tunefs.ocfs2 --fs-features=discontig-bg /dev/sdc1


In case the solution described here above is not applicable, an alternative possible solution which might be helpful is to recreate the file system with a higher data cluster size directive [i.e.: 64K (16 bits) or bigger, see man page for mkfs.ocfs2]. There is no way to change a block size to another value without recreating the file system.

The following command will show the Cluster Size value of the file system:

# debugfs.ocfs2 -n -R "stats" /dev/sdc1 | grep -i "Cluster Size"
    Block Size Bits: 12   Cluster Size Bits: 12

You could format it with a 64Kb block size as follows:

# mkfs.ocfs2 -C 64K /dev/sdc1 -F


Another help which may be worth to try to mitigate the issue, is about freeing contiguous space decreasing the number of node-slots (maximum number of nodes that can concurrently mount the partition) when possible. This space is always consumed by the journals for those slots, even if the nodes is not actually used.
In this way it is not necessary to recreate the file system. However, this is possible only if there are node-slots which are not actually needed.

For example, to know the current number of node-slots:

# debugfs.ocfs2 -n -R "stats" /dev/sdc1 | grep -i "Max Node Slots"
p, li { white-space: pre-wrap; }

Max Node Slots: 8

To change it:

# tunefs.ocfs2 -N 3 /dev/sdc1