Bug 581590

Hi, 

same behaviour on OpenSUSE 11.3. (We didn't use 11.2. The error didn't occur in 11.1)
We use snapshots to backup Xen-DomU's, so a patch would be desirable.

(In reply to comment #1)
> Hi, 
> 
> same behaviour on OpenSUSE 11.3. (We didn't use 11.2. The error didn't occur in
> 11.1)
> We use snapshots to backup Xen-DomU's, so a patch would be desirable.

Hi,

  I tried once to reproduce this on my machine, but failed.
  Could you provide me a step-by-step instruction on how to reproduce this ?

  Thanks.

Hi,

thanks for help.
I also tried to reproduce the error in a test environment, but this fails. In our productive environment the error occur unsteady. Since last week we use 11.3 on one server. The migration of other servers is in planning. 

We do the following:

We have some physical server with Xen paravirtualized machines. 
These DomUs have partitions mounted like this: 
disk = [ 'phy:/dev/system/helios-dns_root,hda1,w',
         'phy:/dev/system/helios-dns_swap,hdb1,w']

We backup the DomUs with the following procedure:
1. Shutdown the DomU
  xm shutdown -w ${DomU}
2. Create a snapshot
  lvcreate --snapshot --size 2000M --name ${Partition}.backup ${Partition}
3. Start the DomU
  xm start ${DomU}
4. Mount the snapshot
5. Create a tar archive of the mounted snapshot
6. Remove the snapshot
  lvremove -f ${Partition}.backup


Since yesterday our backup script use a workaround (he tries to remove the snapshot up to 20 times :-). I will do some more testing and will wait until we migrate the other servers. Maybe the error depends on the server. If i have new informations, i will send a comment...

regards Stefan

It depends from the server. I can reproduce the error only on one of two servers. 

Redhat has the same problem: https://bugzilla.redhat.com/show_bug.cgi?id=577798

When i stop boot.udev the problem doesn't occur.

Hi Stefan,
  This is very helpful info.
  Could you try to 
"""
grep -r watch /etc/udev/rules.d/ /lib/udev/rules.d/
"""
  and see if there's a 80-udisks.rules ?
If yes, could you delete it and retest ?

  Thanks.

(In reply to comment #4)
> It depends from the server. I can reproduce the error only on one of two
> servers. 
> 
> Redhat has the same problem: https://bugzilla.redhat.com/show_bug.cgi?id=577798
> 
> When i stop boot.udev the problem doesn't occur.

Hi Wei Hu,

i found the file /lib/udev/rules.d/80-udisks.rules.
If i move it away and restart /etc/init.d/boot.udev , the error occurs anyway.
If i move some other rules (1*) away, then lvcreate hangs.

regards
Stefan

(In reply to comment #6)
> Hi Wei Hu,
> 
> i found the file /lib/udev/rules.d/80-udisks.rules.
> If i move it away and restart /etc/init.d/boot.udev , the error occurs anyway.

Can you move the 80-udisks.rules to where it was, and comment out following line
"""
KERNEL=="dm-*", OPTIONS+="watch"
"""
I see someone reports that worked for them.

> If i move some other rules (1*) away, then lvcreate hangs.

10-dm.rules       11-dm-lvm.rules   13-dm-disk.rules  
These 3 files are critical for lvm2 to working properly on 11.3
You should not remove any of them.

Thanks

(In reply to comment #6)
> Hi Wei Hu,
> 
> i found the file /lib/udev/rules.d/80-udisks.rules.
> If i move it away and restart /etc/init.d/boot.udev , the error occurs anyway.

Can you move the 80-udisks.rules to where it was, and comment out following line
"""
KERNEL=="dm-*", OPTIONS+="watch"
"""
I see someone reports that worked for them.

> If i move some other rules (1*) away, then lvcreate hangs.

10-dm.rules       11-dm-lvm.rules   13-dm-disk.rules  
These 3 files are critical for lvm2 to working properly on 11.3
You should not remove any of them.

Thanks

No change in behaviour.

(In reply to comment #9)
> No change in behaviour.

So, by running
"""
grep -r watch /etc/udev/rules.d/ /lib/udev/rules.d/
"""
Do you find any other rules watching on device-mapper devices ?

keto-0:~ # grep -r watch /etc/udev/rules.d/ /lib/udev/rules.d/
/lib/udev/rules.d/80-udisks.rules:KERNEL=="dm-*", OPTIONS+="watch"
/lib/udev/rules.d/60-persistent-storage.rules:# watch for future changes
/lib/udev/rules.d/60-persistent-storage.rules:KERNEL!="sr*", OPTIONS+="watch"
/lib/udev/rules.d/13-dm-disk.rules:OPTIONS+="watch"

if i comment out the lines in 80-udisks.rules and 13-dm-disk.rules, then the error doesn't occur. :-)

Is this a possible workaround?

(In reply to comment #12)
> if i comment out the lines in 80-udisks.rules and 13-dm-disk.rules, then the
> error doesn't occur. :-)
> 
> Is this a possible workaround?

Good to know ;)
Yes, I think this can be a workaround.

It's always a problem for device mapper to interact with udev watch.
We are going to follow upstream to resolve this issue.

Thanks for testing this.