[Ocfs2-users] Determining which version of ocfs2 tools a filesystem was created with.

Mikey Austin mikey at mikeyaustin.com
Wed Mar 16 23:19:56 PDT 2011


On Wednesday, February 09, 2011 11:40:01 AM Sunil Mushran wrote:
> On 02/07/2011 10:41 PM, Mikey Austin wrote:
> > We have been using the ocfs2 tools (version 1.4.4) on gentoo
> > (2.6.18-164.15.1 kernel) for around 6 months now. We would like to
> > re-create all of the ocfs2 filesystems that were not created with
> > mkfs.ocfs2 version 1.4.4, however we are not able to determine the
> > version of the ocfs2 tools our filesystems were created with .
> > 
> > Is this possible with version 1.4.4 of the tools?
> 
> The main difference in mkfs.ocfs2 1.4.4 is that it preallocates space
> in all extent_alloc system files.
> 
> Do:
> # debugfs.ocfs2 -R "ls -l //" /dev/sdX1 | grep extent_alloc
>      20              -rw-r--r--   1     0     0         8388608 22-Sep-2010
> 15:49 extent_alloc:0000 21              -rw-r--r--   1     0     0        
> 8388608 22-Sep-2010 15:49 extent_alloc:0001 22              -rw-r--r--   1
>     0     0         8388608 22-Sep-2010 15:49 extent_alloc:0002 23        
>      -rw-r--r--   1     0     0         8388608 22-Sep-2010 15:49
> extent_alloc:0003 24              -rw-r--r--   1     0     0        
> 8388608 22-Sep-2010 15:49 extent_alloc:0004 25              -rw-r--r--   1
>     0     0         8388608 22-Sep-2010 15:49 extent_alloc:0005 26        
>      -rw-r--r--   1     0     0         8388608 22-Sep-2010 15:49
> extent_alloc:0006 27              -rw-r--r--   1     0     0        
> 8388608 22-Sep-2010 15:49 extent_alloc:0007
> 
> The 6th column is size. Here we see all of the extent_allocs were created
> 8M in size. mkfs prior to 1.4.4 would not have preallocated space. Now do
> remember that as this system file is dynamically allocated, a volume
> formatted with the older mkfs can also show the same. But lower
> probability.

Thanks Sunil. I have run your command on the 10G filesystem that we were 
curious about:

# debugfs.ocfs2 -R "ls -l //" /dev/xvdd1 |grep extent_alloc
        20              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0000
        21              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0001
        22              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0002
        23              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0003
        24              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0004
        25              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0005
        26              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0006
        27              -rw-r--r--   1     0     0         4194304  9-Sep-2010 
12:16 extent_alloc:0007

I believe it was indeed created with mkfs.ocfs2 v1.4.4 (digging back into my 
memory). That being said, would that fact that all of the above extent_allocs 
are of the same size further point to pre-allocation?

The reason we want to know is that we were experiencing the disk free issues 
(as in http://oss.oracle.com/bugzilla/show_bug.cgi?id=915) with this 
filesystem. Have you (or anyone else for that matter) experienced this issue 
again with v1.4.4 of the tools?



More information about the Ocfs2-users mailing list