[XFS] xfs_repair time how long it takes??

[XFS] xfs_repair time how long it takes??

[Ryan Lee]

[-- Attachment #1.1: Type: text/plain, Size: 912 bytes --]

Hi, My name is Ryan.

I am a computer programmer who are making  embedded linux products.

Nowadays one of  my job is porting the XFS Filesystem in our embedded linux
products.

Everything is going well until the day before Yesterday.

The problem is the taking time with xfs_repair like this "]# xfs_repair -P
-L /dev/sda2" is to long to wait in our embedded system and for me, so the
booing time is increasing around 2 minutes totally.

If the HDD has a big size of data what  I wrote,  the xfs_repair takes time
more and more next booting time as I mention it.

So, it makes a big problem with my company products.

Would you mind helping me to solve this problem.

I think many company who use XFS Filesystem without  this kind problem.

Please sharing any information with XFS Filesystem.

I attached the console message what i saw.

I really appreciate you read my e-mail.

Have a good day.

Thanks, Ryan.

[-- Attachment #1.2: Type: text/html, Size: 1253 bytes --]

[-- Attachment #2: message.png --]
[-- Type: image/png, Size: 28511 bytes --]

[-- Attachment #3: Type: text/plain, Size: 121 bytes --]

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

Re: [XFS] xfs_repair time how long it takes??

[Brian Candler]

On Fri, Mar 02, 2012 at 10:09:56PM +0900, Ryan Lee wrote:
>    The problem is the taking time with xfs_repair like this "]# xfs_repair
>    -P -L /dev/sda2" is to long to wait in our embedded system and for me,
>    so the booing time is increasing around 2 minutes totally.

Why do you feel you need to run xfs_repair on every boot? Regular Linux
systems do not attempt this, the same as they don't force a full "e2fsck -f"
on every boot.

http://linux.die.net/man/8/fsck.xfs

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

Re: [XFS] xfs_repair time how long it takes??

[Ryan]

How about the dirty log. I use the xfs_repair because the dirty log must be removed when i mount HDD.

As you know xfs filesystem doen't repair the dirty log so i couldn't mount the HDD after when i do cold reset during writing the HDD. How i can solve this problem? 

Sent from my iPhone.



On Mar 2, 2012, at 11:01 PM, Brian Candler <B.Candler@pobox.com> wrote:

> On Fri, Mar 02, 2012 at 10:09:56PM +0900, Ryan Lee wrote:
>>   The problem is the taking time with xfs_repair like this "]# xfs_repair
>>   -P -L /dev/sda2" is to long to wait in our embedded system and for me,
>>   so the booing time is increasing around 2 minutes totally.
> 
> Why do you feel you need to run xfs_repair on every boot? Regular Linux
> systems do not attempt this, the same as they don't force a full "e2fsck -f"
> on every boot.
> 
> http://linux.die.net/man/8/fsck.xfs

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

Re: [XFS] xfs_repair time how long it takes??

[Brian Candler]

On Fri, Mar 02, 2012 at 11:26:30PM +0900, Ryan wrote:
> How about the dirty log. I use the xfs_repair because the dirty log must be removed when i mount HDD.

Google for "xfs dirty log" and I get this as the second hit:

http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/6/html/Storage_Administration_Guide/xfsrepair.html

What it says is the exact opposite of what you said above:

- the dirty log is automatically dealt with when you mount the filesystem
- xfs_repair will refuse to run if a dirty log is present

So running xfs_repair on boot is actually a bad idea.

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

Re: [XFS] xfs_repair time how long it takes??

[Eric Sandeen]

On 3/2/12 7:09 AM, Ryan Lee wrote:
> Hi, My name is Ryan.
>  
> I am a computer programmer who are making  embedded linux products.
>  
> Nowadays one of  my job is porting the XFS Filesystem in our embedded linux products.
>  
> Everything is going well until the day before Yesterday.
>  
> The problem is the taking time with xfs_repair like this "]# xfs_repair -P -L /dev/sda2" is to long to wait in our embedded system and for me, so the booing time is increasing around 2 minutes totally.

Why are you running it with -L?
And why are you running it with -P?

As others mentioned, there is no reason to run xfs_repair on every boot.
And if you run it with -L on every boot, you've ruined the whole reason for using a journaling filesystem in the first place.

-Eric

> If the HDD has a big size of data what  I wrote,  the xfs_repair takes time more and more next booting time as I mention it.
>  
> So, it makes a big problem with my company products.
>  
> Would you mind helping me to solve this problem.
>  
> I think many company who use XFS Filesystem without  this kind problem.
>  
> Please sharing any information with XFS Filesystem.
>  
> I attached the console message what i saw.
>  
> I really appreciate you read my e-mail.
>  
> Have a good day.
>  
> Thanks, Ryan.
>  
>  
>  
> 
> 
> _______________________________________________
> xfs mailing list
> xfs@oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs

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

Re: [XFS] xfs_repair time how long it takes??

[Ryan Lee]

[-- Attachment #1.1: Type: text/plain, Size: 2806 bytes --]

Hi, all.

Thanks a lot of your kindness for me.
I really realy appreciate it.

I am going to tell you what I have done so far porting  XFS in our system.

First of all, the resources of our system.

CPU : MIPS
KERNEL : 2.6.37-2.6
XFSPROGS : 3.1.7

Second, I did this step.

1. mounting an xfs filesystem after an unclean shutdown on a machine.
As a result i saw this error below.
------------------------------------------------------------------------
XFS: xlog_recover_process_data: bad clientid
XFS: log mount/recovery failed: error 5
XFS: log mount failed
------------------------------------------------------------------------

2. I patched for this error 1 from internet.
http://comments.gmane.org/gmane.comp.file-systems.xfs.general/35446

I did aging test by my self.
As a result, after one day overnight, the XFS Filesystem crashed.

3, I look for another solution for 1 problem. eventually I found using
xfs_repair to remove dirty log and recover.
But as i mentioned it, it takes time a lot.

You guys told me to use  the "fsck.xfs". but i think it couldn't solve the
problem 1 above.

Onething more, I already know about the problem with "xfs_repair", but I
cound't find any other good solution to fix it until now.

How can I fix it?
Please help me.

Thanks, Ryan.




On Sat, Mar 3, 2012 at 12:07 AM, Eric Sandeen <sandeen@sandeen.net> wrote:

> On 3/2/12 8:26 AM, Ryan wrote:
> >
> > How about the dirty log. I use the xfs_repair because the dirty log must
> be removed when i mount HDD.
> >
> > As you know xfs filesystem doen't repair the dirty log so i couldn't
> mount the HDD after when i do cold reset during writing the HDD. How i can
> solve this problem?
>
> If mount fails to replay the dirty log after a power loss, then I would
> suspect a misconfiguration (maybe mounting -o nobarrier with write caches
> enabled on the drives?) or a bug in your version of xfs, possibly unique to
> architecture (what architecture is it, what kernel is it, what xfs
> "porting" did you do?)
>
> -Eric
>
> > Sent from my iPhone.
> >
> >
> >
> > On Mar 2, 2012, at 11:01 PM, Brian Candler <B.Candler@pobox.com> wrote:
> >
> >> On Fri, Mar 02, 2012 at 10:09:56PM +0900, Ryan Lee wrote:
> >>>   The problem is the taking time with xfs_repair like this "]#
> xfs_repair
> >>>   -P -L /dev/sda2" is to long to wait in our embedded system and for
> me,
> >>>   so the booing time is increasing around 2 minutes totally.
> >>
> >> Why do you feel you need to run xfs_repair on every boot? Regular Linux
> >> systems do not attempt this, the same as they don't force a full
> "e2fsck -f"
> >> on every boot.
> >>
> >> http://linux.die.net/man/8/fsck.xfs
> >
> > _______________________________________________
> > xfs mailing list
> > xfs@oss.sgi.com
> > http://oss.sgi.com/mailman/listinfo/xfs
> >
>
>

[-- Attachment #1.2: Type: text/html, Size: 4273 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] xfs_repair time how long it takes??

[Ryan Lee]

[-- Attachment #1.1: Type: text/plain, Size: 3395 bytes --]

Eric,
I have one more qustion for  you.

You told me like this "then I would suspect a misconfiguration (maybe
mounting -o nobarrier with write caches enabled on the drives?) "

How can I check this misconfiguration? such as "-o nobarrier" is I couldn't
find this option with mount in my busybox and what do you mean the sentence
this " write caches enabled on the drives?"

Thanks,



On Sat, Mar 3, 2012 at 12:48 AM, Ryan Lee <ryanlee.lahk@gmail.com> wrote:

>
> Hi, all.
>
> Thanks a lot of your kindness for me.
> I really realy appreciate it.
>
> I am going to tell you what I have done so far porting  XFS in our system.
>
> First of all, the resources of our system.
>
> CPU : MIPS
> KERNEL : 2.6.37-2.6
> XFSPROGS : 3.1.7
>
> Second, I did this step.
>
> 1. mounting an xfs filesystem after an unclean shutdown on a machine.
> As a result i saw this error below.
> ------------------------------------------------------------------------
> XFS: xlog_recover_process_data: bad clientid
> XFS: log mount/recovery failed: error 5
> XFS: log mount failed
> ------------------------------------------------------------------------
>
> 2. I patched for this error 1 from internet.
> http://comments.gmane.org/gmane.comp.file-systems.xfs.general/35446
>
> I did aging test by my self.
> As a result, after one day overnight, the XFS Filesystem crashed.
>
> 3, I look for another solution for 1 problem. eventually I found using
> xfs_repair to remove dirty log and recover.
> But as i mentioned it, it takes time a lot.
>
> You guys told me to use  the "fsck.xfs". but i think it couldn't solve the
> problem 1 above.
>
> Onething more, I already know about the problem with "xfs_repair", but I
> cound't find any other good solution to fix it until now.
>
> How can I fix it?
> Please help me.
>
> Thanks, Ryan.
>
>
>
>
> On Sat, Mar 3, 2012 at 12:07 AM, Eric Sandeen <sandeen@sandeen.net> wrote:
>
>> On 3/2/12 8:26 AM, Ryan wrote:
>> >
>> > How about the dirty log. I use the xfs_repair because the dirty log
>> must be removed when i mount HDD.
>> >
>> > As you know xfs filesystem doen't repair the dirty log so i couldn't
>> mount the HDD after when i do cold reset during writing the HDD. How i can
>> solve this problem?
>>
>> If mount fails to replay the dirty log after a power loss, then I would
>> suspect a misconfiguration (maybe mounting -o nobarrier with write caches
>> enabled on the drives?) or a bug in your version of xfs, possibly unique to
>> architecture (what architecture is it, what kernel is it, what xfs
>> "porting" did you do?)
>>
>> -Eric
>>
>> > Sent from my iPhone.
>> >
>> >
>> >
>> > On Mar 2, 2012, at 11:01 PM, Brian Candler <B.Candler@pobox.com> wrote:
>> >
>> >> On Fri, Mar 02, 2012 at 10:09:56PM +0900, Ryan Lee wrote:
>> >>>   The problem is the taking time with xfs_repair like this "]#
>> xfs_repair
>> >>>   -P -L /dev/sda2" is to long to wait in our embedded system and for
>> me,
>> >>>   so the booing time is increasing around 2 minutes totally.
>> >>
>> >> Why do you feel you need to run xfs_repair on every boot? Regular Linux
>> >> systems do not attempt this, the same as they don't force a full
>> "e2fsck -f"
>> >> on every boot.
>> >>
>> >> http://linux.die.net/man/8/fsck.xfs
>> >
>> > _______________________________________________
>> > xfs mailing list
>> > xfs@oss.sgi.com
>> > http://oss.sgi.com/mailman/listinfo/xfs
>> >
>>
>>
>

[-- Attachment #1.2: Type: text/html, Size: 5220 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] xfs_repair time how long it takes??

[Brian Candler]

On Sat, Mar 03, 2012 at 12:48:06AM +0900, Ryan Lee wrote:
>    You guys told me to use  the "fsck.xfs".

No I didn't - I linked to the manpage.  If you read it, you will see that it
says it does nothing at all.

It is purely a placeholder, for Linux systems which try to do fsck.<foo>
whenever they mount a filesystem of type <foo>.

The point is, you don't need it.

>    Onething more, I already know about the problem with "xfs_repair", but
>    I cound't find any other good solution to fix it until now.
> 
>    How can I fix it?

Probably you need to stop your journal from getting corrupted in the first
place.  See the XFS FAQ:

http://xfs.org/index.php/XFS_FAQ#Q:_What_is_the_problem_with_the_write_cache_on_journaled_filesystems.3F

and the following 5 questions and answers.  In particular make sure your
block driver has working write barrier support.

Final note: please don't reply to me without replying to the list as well. I
have been forwarding your replies to the list, but it is annoying to have to
do so.  Your mail client probably has a 'reply to list' or 'group reply'
function you can use.

Regards,

Brian.

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