Why can't I stop these arrays?

Why can't I stop these arrays?

[Richard Michael]

I am experimenting with various array configurations.  I have several
arrays running and I want to stop them.  They are not mounted; indeed,
they don't even have filesystems on them.  As far as I can tell, nothing
on the system is using them.

I have tried stopping the monitoring mdadm process, but I still cannot
stop the arrays.  (Although, if it was necessary to stop the monitoring
process for all arrays just to stop one of them, that would be a bug.)

There is a raid5 resyncing, which might be causing the problem (although
it shouldn't), but it needs hours to finish the resync.  Also, none of
the components in the raid5 array are in the arrays I want to stop, so
again, I don't see why there should be a problem.

(Though, md3 is also a raid5 and md5 is comprised of md3, so in some
sense md5 also "has something to do with" raid5.  Problem?)

Aside, how can I stop the resync to test if it is interferring?  Note,
the system is running of the resyncing raid5, so I need to stop the
resync but *not* stop the array.

Any advice?  Probably I've missed something obvious..


[root@server ~]# cat /proc/mdstat 
Personalities : [raid6] [raid5] [raid4] [raid1] 
md5 : active raid1 md3[0]
      1952475584 blocks [2/1] [U_]
      
md3 : active raid5 sdd2[0] sdg2[3](S) sdf2[2] sde2[1]
      1952475648 blocks level 5, 64k chunk, algorithm 2 [3/3] [UUU]
      
md0 : active raid1 sda1[0] sdb1[1]
      521984 blocks [3/2] [UU_]
      
md1 : active raid5 sdc2[3] sda2[0] sdb2[1]
      1952475648 blocks level 5, 256k chunk, algorithm 2 [3/2] [UU_]
      [=======>.............]  recovery = 35.9% (351008004/976237824)
finish=163.5min speed=63727K/sec
      
unused devices: <none>
[root@server ~]# mdadm --stop /dev/md5
mdadm: fail to stop array /dev/md5: Device or resource busy
[root@server ~]# mdadm --stop /dev/md3
mdadm: fail to stop array /dev/md3: Device or resource busy


Thanks,
Richard

Re: Why can't I stop these arrays?

[Roger Heflin]

Richard Michael wrote:
> I am experimenting with various array configurations.  I have several
> arrays running and I want to stop them.  They are not mounted; indeed,
> they don't even have filesystems on them.  As far as I can tell, nothing
> on the system is using them.
> 
> I have tried stopping the monitoring mdadm process, but I still cannot
> stop the arrays.  (Although, if it was necessary to stop the monitoring
> process for all arrays just to stop one of them, that would be a bug.)
> 
> There is a raid5 resyncing, which might be causing the problem (although
> it shouldn't), but it needs hours to finish the resync.  Also, none of
> the components in the raid5 array are in the arrays I want to stop, so
> again, I don't see why there should be a problem.
> 
> (Though, md3 is also a raid5 and md5 is comprised of md3, so in some
> sense md5 also "has something to do with" raid5.  Problem?)
> 
> Aside, how can I stop the resync to test if it is interferring?  Note,
> the system is running of the resyncing raid5, so I need to stop the
> resync but *not* stop the array.
> 
> Any advice?  Probably I've missed something obvious..
> 
> 
> [root@server ~]# cat /proc/mdstat 
> Personalities : [raid6] [raid5] [raid4] [raid1] 
> md5 : active raid1 md3[0]
>       1952475584 blocks [2/1] [U_]
>       
> md3 : active raid5 sdd2[0] sdg2[3](S) sdf2[2] sde2[1]
>       1952475648 blocks level 5, 64k chunk, algorithm 2 [3/3] [UUU]
>       
> md0 : active raid1 sda1[0] sdb1[1]
>       521984 blocks [3/2] [UU_]
>       
> md1 : active raid5 sdc2[3] sda2[0] sdb2[1]
>       1952475648 blocks level 5, 256k chunk, algorithm 2 [3/2] [UU_]
>       [=======>.............]  recovery = 35.9% (351008004/976237824)
> finish=163.5min speed=63727K/sec
>       
> unused devices: <none>
> [root@server ~]# mdadm --stop /dev/md5
> mdadm: fail to stop array /dev/md5: Device or resource busy
> [root@server ~]# mdadm --stop /dev/md3
> mdadm: fail to stop array /dev/md3: Device or resource busy
> 

Do they have volume groups on them?

If so you need to turn them off.

I would "vgdisplay" and if they show up in there they will need to be turned off 
there before they can be unmounted (vgchange -a y device).

                       Roger

Re: Why can't I stop these arrays?

[Mikael Abrahamsson]

On Sun, 22 Jun 2008, Richard Michael wrote:

> they don't even have filesystems on them.  As far as I can tell, nothing 
> on the system is using them.

Have you used the "lsof" utility to see what process might be accessing 
the md devices? That's where I start when I run into the problem you 
described (not knowing what is locking what I want to unmount or stop).

-- 
Mikael Abrahamsson    email: swmike@swm.pp.se

Re: Why can't I stop these arrays? PATCH

[Richard Michael]

>> Any advice?  Probably I've missed something obvious..

Yes, of course I did. :)  (To other poster, I checked lsof immediately.)

> Do they have volume groups on them?

Yes, I do.  I should know, since I created them.  I've been fooling
around all day with lvm, but somehow completely forget it would keep the
array active..

> I would "vgdisplay" and if they show up in there they will need to be 
> turned off there before they can be unmounted (vgchange -a y device).

I'm fairly certain here you mean: "vgchange -a n <vg_name>".

Thanks Roger!

Since this has come up on the mailing before, is there any interest in
this trivial one liner?  (Perhaps it needs to go elsewhere in the code
as well.)

diff -u -r mdadm-2.6.7/Manage.c mdadm-2.6.7-local/Manage.c
--- mdadm-2.6.7/Manage.c        2008-06-05 22:11:04.000000000 -0400
+++ mdadm-2.6.7-local/Manage.c  2008-06-22 20:35:25.000000000 -0400
@@ -115,6 +115,7 @@
                        if (quiet==0)
                                fprintf(stderr, Name ": fail to stop array %s: %s\n",
                                        devname, strerror(errno));
+                               fprintf(stderr, "Perhaps a running process, mounted filesystem or active volume group?\n");
                        return 1;
                }
                if (quiet <= 0)


Regards,
Richard

Re: Why can't I stop these arrays? PATCH

[Neil Brown]

On Sunday June 22, rmichael-raid@edgeofthenet.org wrote:
> 
> Since this has come up on the mailing before, is there any interest in
> this trivial one liner?  (Perhaps it needs to go elsewhere in the code
> as well.)
> 
> diff -u -r mdadm-2.6.7/Manage.c mdadm-2.6.7-local/Manage.c
> --- mdadm-2.6.7/Manage.c        2008-06-05 22:11:04.000000000 -0400
> +++ mdadm-2.6.7-local/Manage.c  2008-06-22 20:35:25.000000000 -0400
> @@ -115,6 +115,7 @@
>                         if (quiet==0)
>                                 fprintf(stderr, Name ": fail to stop array %s: %s\n",
>                                         devname, strerror(errno));
> +                               fprintf(stderr, "Perhaps a running process, mounted filesystem or active volume group?\n");
>                         return 1;
>                 }
>                 if (quiet <= 0)
> 

Somewhere way down on my TODO list is
  Improve error messages to try and really explain what is
  going on.

This is in line with that so I'm happy to take it.
However I would at least like the new message to be conditional on
   errno == EBUSY

I don't think any other error are possible, but it is good practice,
provides implicit documentation, and guards against possible future
changes.

NeilBrown