xfs - fixing wrong xfs size

xfs - fixing wrong xfs size

[Nebojsa Trpkovic]

[-- Attachment #1.1: Type: text/plain, Size: 705 bytes --]

hello.

I had problem with raid5 on my system.
after recovering of the array, I can not mount XFS that resides on it.
I've run xfs_check and xfs_repair (which fixed some errors) and now I can
run them all day long without any errors.

but, when I try to mount that partition, I get:

mount: /dev/md0: can't read superblock


and there's

Filesystem "md0": Disabling barriers, not supported by the underlying device
attempt to access beyond end of device
md0: rw=0, want=123024384, limit=123023488
I/O error in filesystem ("md0") meta-data dev md0 block 0x75533f8
 ("xfs_read_buf") error 5 buf count 4096
XFS: size check 2 failed



in the dmesg.


is there any way to "fix" this wrong size of XFS ?

thx.

[-- Attachment #1.2: Type: text/html, Size: 1149 bytes --]

[-- Attachment #2: Type: text/plain, Size: 121 bytes --]

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs - fixing wrong xfs size

[Eric Sandeen]

Nebojsa Trpkovic wrote:
> hello.
> 
> I had problem with raid5 on my system.
> after recovering of the array, I can not mount XFS that resides on it.
> I've run xfs_check and xfs_repair (which fixed some errors) and now I
> can run them all day long without any errors.
> 
> but, when I try to mount that partition, I get:
> 
> mount: /dev/md0: can't read superblock
> 
> 
> and there's
> 
> Filesystem "md0": Disabling barriers, not supported by the underlying device
> 
> attempt to access beyond end of device
> md0: rw=0, want=123024384, limit=123023488
> I/O error in filesystem ("md0") meta-data dev md0 block 0x75533f8

       ("xfs_read_buf") error 5 buf count 4096
> 
> XFS: size check 2 failed

So, the superblock says that the fs is 896 1k-blocks longer than the
device actually is.

> in the dmesg.
> 
> 
> is there any way to "fix" this wrong size of XFS ?

hard to say, almost certainly sounds like an md problem; is there a
chance that your raid recovery led to a device which is somehow smaller
than it started?  You could change the superblock block-count value, but
I'm guessing that something else has gone wrong.

It might be useful to know what errors xfs_repair found.

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Fwd: xfs - fixing wrong xfs size

[Nebojsa Trpkovic]

[-- Attachment #1.1: Type: text/plain, Size: 2272 bytes --]

On Mon, Apr 13, 2009 at 5:02 PM, Eric Sandeen <sandeen@sandeen.net> wrote:

> Nebojsa Trpkovic wrote:
> > hello.
> >
> > I had problem with raid5 on my system.
> > after recovering of the array, I can not mount XFS that resides on it.
> > I've run xfs_check and xfs_repair (which fixed some errors) and now I
> > can run them all day long without any errors.
> >
> > but, when I try to mount that partition, I get:
> >
> > mount: /dev/md0: can't read superblock
> >
> >
> > and there's
> >
> > Filesystem "md0": Disabling barriers, not supported by the underlying
> device
> >
> > attempt to access beyond end of device
> > md0: rw=0, want=123024384, limit=123023488
> > I/O error in filesystem ("md0") meta-data dev md0 block 0x75533f8
>
>       ("xfs_read_buf") error 5 buf count 4096
> >
> > XFS: size check 2 failed
>
> So, the superblock says that the fs is 896 1k-blocks longer than the
> device actually is.
>
> > in the dmesg.
> >
> >
> > is there any way to "fix" this wrong size of XFS ?
>
> hard to say, almost certainly sounds like an md problem; is there a
> chance that your raid recovery led to a device which is somehow smaller
> than it started?  You could change the superblock block-count value, but
> I'm guessing that something else has gone wrong.
>
> It might be useful to know what errors xfs_repair found.
>
> -Eric
>


thx for quick reply.

yes, I definitely had md problem - my raid5 made of 6 drives connected to
ICH9R and 2 drives connected to the masterpeace of hardware called JMicron
fell apart when later one just stopped working out of nowhere. after reboot,
JMicron worked well, but raid was out of sync with 2 members failed. I've
zeroed superblocks of raid partitions and created new one with
"assume-clear". partition tables stayed intact - I had no need to change
anything in them. that worked fine with /dev/md1 wich had reiserfs: after
long and slow reiserfs --rebuild-tree I'm able to access all files on that
partition. unforutuately, I have no way to mount xfs partition although
xfs_check and xfs_repair give no errors (I had to run xfs_repair -L /dev/md0
to get in this error-free situation).

how can I set superblock block-count value (now I have realy nothing to
loose - the only other option is to give up of that data) ?

[-- Attachment #1.2: Type: text/html, Size: 2980 bytes --]

[-- Attachment #2: Type: text/plain, Size: 121 bytes --]

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Fwd: xfs - fixing wrong xfs size

[Rafal Blaszczyk]

On Mon, Apr 13, 2009 at 5:02 PM, Eric Sandeen <sandeen@sandeen.net> wrote:


> > attempt to access beyond end of device
> > md0: rw=0, want=123024384, limit=123023488
> > I/O error in filesystem ("md0") meta-data dev md0 block 0x75533f8
>
>       ("xfs_read_buf") error 5 buf count 4096
> >
> > XFS: size check 2 failed

Hi,
I've had similiar problems with md on top of lvm after converting from
single LV to mirrored (md device). I've managed to solve it by expanding
underlying devices by just a few megabytes. I suppose it's not your case
because you're dealing with bare devices.
But you could still try to experiment with xfs_growfs or other xfs_* tools.

>From my case - this is not md's fault. XFS wants to have more underlying
space than it has and it cannot be mounted.
-- 
View this message in context: http://old.nabble.com/xfs---fixing-wrong-xfs-size-tp23023562p28465863.html
Sent from the Xfs - General mailing list archive at Nabble.com.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Fwd: xfs - fixing wrong xfs size

[Eric Sandeen]

On 05/05/2010 03:09 PM, Rafal Blaszczyk wrote:
> 
> 
> 
> On Mon, Apr 13, 2009 at 5:02 PM, Eric Sandeen <sandeen@sandeen.net> wrote:
> 
> 
>>> attempt to access beyond end of device
>>> md0: rw=0, want=123024384, limit=123023488
>>> I/O error in filesystem ("md0") meta-data dev md0 block 0x75533f8
>>
>>       ("xfs_read_buf") error 5 buf count 4096
>>>
>>> XFS: size check 2 failed
> 
> Hi,
> I've had similiar problems with md on top of lvm after converting from
> single LV to mirrored (md device). I've managed to solve it by expanding
> underlying devices by just a few megabytes. I suppose it's not your case
> because you're dealing with bare devices.
> But you could still try to experiment with xfs_growfs or other xfs_* tools.
> 
> From my case - this is not md's fault. XFS wants to have more underlying
> space than it has and it cannot be mounted.

xfs wants only as much space as it has when it was created :)

If that changes such that it is smaller, you'll get this warning at
mount time.  This is almost certainly the result of something that
happened outside xfs's control.

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Fwd: xfs - fixing wrong xfs size

[Eric Sandeen]

On 04/13/2009 11:33 AM, Nebojsa Trpkovic wrote:

> how can I set superblock block-count value (now I have realy nothing to
> loose - the only other option is to give up of that data) ?

Sorry, I never replied to this :(

You can use xfs_db to set it; or you could comment out the kernel
check... and mount readonly, and copy off the data you can get to?

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Fwd: xfs - fixing wrong xfs size

[johahoff]

> You can use xfs_db to set it; or you could comment out the kernel
> check... and mount readonly, and copy off the data you can get to?

I have got the same problem after recovering a xfs volume using ddrescue. It
seems that the new 1 TB drive I dd'ed to is a few blocks short ...

I had a look at the xfs_db man page, but wasn't able to figure out which
command would allow me to set the block count of the fs. Can you please give
the appropriate commands? I think I'll be able to figure out the correct
counts on my own.

Thanks a lot!

-Johannes
-- 
View this message in context: http://old.nabble.com/xfs---fixing-wrong-xfs-size-tp23023562p30211747.html
Sent from the Xfs - General mailing list archive at Nabble.com.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Fwd: xfs - fixing wrong xfs size

[Eric Sandeen]

On 11/14/10 5:16 AM, johahoff wrote:
> 
>> You can use xfs_db to set it; or you could comment out the kernel
>> check... and mount readonly, and copy off the data you can get to?
> 
> I have got the same problem after recovering a xfs volume using ddrescue. It
> seems that the new 1 TB drive I dd'ed to is a few blocks short ...
> 
> I had a look at the xfs_db man page, but wasn't able to figure out which
> command would allow me to set the block count of the fs. Can you please give
> the appropriate commands? I think I'll be able to figure out the correct
> counts on my own.
> 
> Thanks a lot!
> 
> -Johannes

# xfs_db -w /dev/blah
xfs_db> sb 0
xfs_db> p
...
magicnum = 0x58465342
blocksize = 4096
dblocks = XXXXXX
...
xfs_db> w dblocks <NEW VALUE>
xfs_db> quit

You may have to do it for each of the superblocks but I think the first
should suffice.

If files reference the blocks past the device then of course you'll hit
errors trying to read them as well ...

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Fwd: xfs - fixing wrong xfs size

[johahoff]

That worked perfectly, thanks a lot, Eric!

One minor issue: I had to use 'xfs_db -x', not '-w', but probably that's
because of an old version of xfs-utils on the SystemRescueCD I was using.
Just a hint if anybody is wondering.

Thanks again,
Johannes
-- 
View this message in context: http://old.nabble.com/xfs---fixing-wrong-xfs-size-tp23023562p30492876.html
Sent from the Xfs - General mailing list archive at Nabble.com.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: Fwd: xfs - fixing wrong xfs size

[Eric Sandeen]

On 12/19/10 8:37 AM, johahoff wrote:
> 
> That worked perfectly, thanks a lot, Eric!
> 
> One minor issue: I had to use 'xfs_db -x', not '-w', but probably that's
> because of an old version of xfs-utils on the SystemRescueCD I was using.
> Just a hint if anybody is wondering.

Nah that was me misremembering debugfs vs. xfs_db arguments :)

-Eric

> Thanks again,
> Johannes

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs