[linux-lvm] unwanted autocreation of /dev/mapper/ entries while using parted on lvm devices

[linux-lvm] unwanted autocreation of /dev/mapper/ entries while using parted on lvm devices

[Oliver Rath]

Hi List,

Im not sure, if the udev-rules of lvm is creating this. Im Using Ubuntu
12.04.4 + lvm git yesterday, compiles with --enable-udev_sync
--enable-udev_rules. Here my testing results:


root@dmicrash\:rules.d# vgcreate myvg /dev/sdc1 /dev/sdd1
  Volume group "myvg" successfully created
root@dmicrash\:rules.d# lvcreate --name test --size 20G myvg
  Logical volume "test" created
root@dmicrash\:rules.d# parted /dev/myvg/test mklabel msdos

root@dmicrash\:rules.d# parted /dev/myvg/test mkpart primary  0% 40%
     
root@dmicrash\:rules.d# ls /dev/mapper/
control  myvg-test  myvg-test1

The device myvg-test1 was created automatically (normally this is done
by kpartx?). I disabled gnome-automounter, didnt help.

I tried to monitor this with udevadm --monitor, but i didnt get
displayed the rule-file who did this.

Any hints?

Tfh!
Oliver

Re: [linux-lvm] unwanted autocreation of /dev/mapper/ entries while using parted on lvm devices

[Peter Rajnoha]

On 02/23/2014 11:36 PM, Oliver Rath wrote:
> Hi List,
> 
> Im not sure, if the udev-rules of lvm is creating this. Im Using Ubuntu
> 12.04.4 + lvm git yesterday, compiles with --enable-udev_sync
> --enable-udev_rules. Here my testing results:
> 
> 
> root@dmicrash\:rules.d# vgcreate myvg /dev/sdc1 /dev/sdd1
>   Volume group "myvg" successfully created
> root@dmicrash\:rules.d# lvcreate --name test --size 20G myvg
>   Logical volume "test" created
> root@dmicrash\:rules.d# parted /dev/myvg/test mklabel msdos
> 
> root@dmicrash\:rules.d# parted /dev/myvg/test mkpart primary  0% 40%
>      
> root@dmicrash\:rules.d# ls /dev/mapper/
> control  myvg-test  myvg-test1
> 
> The device myvg-test1 was created automatically (normally this is done
> by kpartx?). I disabled gnome-automounter, didnt help.
> 
> I tried to monitor this with udevadm --monitor, but i didnt get
> displayed the rule-file who did this.
> 
> Any hints?

The parted itself creates a new device-mapper mapping that represents
the partition. Then it's like any other device-mapper device and so
the /dev content is created by 10-dm-disk.rules (the /dev/mapper
content) and 13-dm-disk.rules (the /dev/disk content).

Parted has this functionality integrated so there's no need to call
kpartx in addition.

-- 
Peter

Re: [linux-lvm] unwanted autocreation of /dev/mapper/ entries while using parted on lvm devices

[Oliver Rath]

Hi Peter,

thanks for your reply!

Am 24.02.2014 09:41, schrieb Peter Rajnoha:
> On 02/23/2014 11:36 PM, Oliver Rath wrote:
>> [..]
> The parted itself creates a new device-mapper mapping that represents
> the partition. Then it's like any other device-mapper device and so
> the /dev content is created by 10-dm-disk.rules (the /dev/mapper
> content) and 13-dm-disk.rules (the /dev/disk content).
>
> Parted has this functionality integrated so there's no need to call
> kpartx in addition.
>

Ist it possible to avoid this /dev/mapper-generation, if I partitioning
lvm-devices? My problem ist, that Im not able to export the whole device
i.e. as iscsi, if some sub-device is mapped in /dev/mapper/... (says,
its busy then ..).

At the moment I always have to delete these subdevices via kpartx -d
/dev/myvg/.. (or dmsetup remove ..) before exporting it, which is a bit
annoying.

I took a look into 10-dm.rules, but have no idea, what to do.

Regards,
Oliver

[linux-lvm] [SOLVED] unwanted autocreation of /dev/mapper/ entries while using parted on lvm devices

[Oliver Rath]

Hi Peter!

Am 24.02.2014 09:41, schrieb Peter Rajnoha:
> On 02/23/2014 11:36 PM, Oliver Rath wrote:
> The parted itself creates a new device-mapper mapping that represents
> the partition. Then it's like any other device-mapper device and so
> the /dev content is created by 10-dm-disk.rules (the /dev/mapper
> content) and 13-dm-disk.rules (the /dev/disk content). Parted has this
> functionality integrated so there's no need to call kpartx in addition. 

On my machine parted-2.3 was running. Upgrading to parted-3.1 didnt
help, BUT building parted-3.1 for myself with ./configure
--disable-device-mapper works! Then parted does his job without creating
these deices implicitly.

Maybe you know a less radical method? I.e. I could exclude some names in
lvm.conf or something like this? The names of the lvm whould should
create subvolumes are well defined.

Regards,
Oliver

Re: [linux-lvm] [SOLVED] unwanted autocreation of /dev/mapper/ entries while using parted on lvm devices

[Peter Rajnoha]

On 02/24/2014 01:40 PM, Oliver Rath wrote:
> Hi Peter!
> 
> Am 24.02.2014 09:41, schrieb Peter Rajnoha:
>> On 02/23/2014 11:36 PM, Oliver Rath wrote:
>> The parted itself creates a new device-mapper mapping that represents
>> the partition. Then it's like any other device-mapper device and so
>> the /dev content is created by 10-dm-disk.rules (the /dev/mapper
>> content) and 13-dm-disk.rules (the /dev/disk content). Parted has this
>> functionality integrated so there's no need to call kpartx in addition. 
> 
> On my machine parted-2.3 was running. Upgrading to parted-3.1 didnt
> help, BUT building parted-3.1 for myself with ./configure
> --disable-device-mapper works! Then parted does his job without creating
> these deices implicitly.

Yes, I was just about to write that :) I think there's no option in parted
to avoid creating these mappings for partitions once the parted is
compiled with device-mapper support. So yes, the only option is either
to recompile without devmapper support or always remove these mappings
manually...

> 
> Maybe you know a less radical method? I.e. I could exclude some names in
> lvm.conf or something like this? The names of the lvm whould should
> create subvolumes are well defined.

You would need to tell parted to stop syncing the state after partition
signatures are created/removed on/from disk. But parted always creates/removes
the partition mappings unconditionally (...or it calls BLKRRPART/blkpg for
non-dm devices).

You can try using fdisk instead which does not do this step and requires
calling partprobe or kpartx manually after partitioning.
-- 
Peter