LVM/chroot: WARNING: Device /dev/vg_sys/swap not initialized in udev database even after

[Tobias Unsleber]

Hi,

sorry for the bad formatting. This should be better.

I'm preparing Debian Buster for use with fai, and I'm getting a very
long delay for the installation when using LVM partition layout.

The following process takes a long time(the vgs command is ren multiple
times)

ps ax --forest 

\_ /bin/sh /usr/sbin/grub-mkconfig -o /boot/grub/grub.cfg 
 \_ /bin/sh /etc/grub.d/00_header 
  \_ /usr/sbin/grub-probe --device /dev/mapper/vg_sys-root --target=compatibility_hint 
   \_ vgs --options vg_uuid,pv_name --noheadings --separator : |

So the prozess vgs --options ... ist stuck.

When I'm executing the vgs command manually it's stuck too.

Using verbosity and debug switches, I get this output:

time vgs -vvvddd --options vg_uuid,pv_name --noheadings --separator :
  Parsing: vgs -vvvddd --options vg_uuid,pv_name --noheadings --separator: 
... 
  Device /dev/vg_sys/swap not initialized in udev database (1/100, 100000 microseconds). 
  Device /dev/vg_sys/swap not initialized in udev database (2/100, 100000 microseconds). 
... 
  Device /dev/vg_sys/swap not initialized in udev database (100/100, 9900000 microseconds). 

WARNING: Device /dev/vg_sys/swap not initialized in udev database even after waiting 10000000 microseconds. 

  filter md deferred /dev/vg_sys/swap
  filter cache deferred /dev/vg_sys/swap
... 
  Device /dev/vg_sys/root not initialized in udev database (1/100, 0 microseconds). 
  Device /dev/vg_sys/root not initialized in udev database (2/100, 100000 microseconds). 
... 
  Device /dev/vg_sys/root not initialized in udev database (100/100, 9900000 microseconds). 

WARNING: Device /dev/vg_sys/root not initialized in udev database even after waiting 10000000 microseconds.
...
rAGmpb-AnKr-Iemp-2WPv-I37a-N7C8-aUcJ2F:/dev/xvdb2 
Completed: vgs -vvvddd --options vg_uuid,pv_name --noheadings --separator : 

real 1m30,375s 
user 0m0,064s 
sys 0m0,132s 

I also checked what's going on with strace. I see a lot of Errors like this:

openat(AT_FDCWD, "/run/udev/data/b254:1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (Datei oder Verzeichnis nicht gefunden) 
openat(AT_FDCWD, "/run/udev/data/b254:1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (Datei oder Verzeichnis nicht gefunden) 
openat(AT_FDCWD, "/run/udev/data/b254:1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (Datei oder Verzeichnis nicht gefunden)
openat(AT_FDCWD, "/run/udev/data/b254:1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (Datei oder Verzeichnis nicht gefunden)
openat(AT_FDCWD,"/run/udev/data/b254:1", O_RDONLY|O_CLOEXEC) = -1 ENOENT (Datei oder Verzeichnis nicht gefunden)

After checking the documentation I wondered if using the lvm options
"use_lvmetad = 0" or "obtain_device_list_from_udev = 0" might help.
After setting them and make sure they are active(via lvm dumpconifg)
I saw that this was not the case.

On the internet there are advices to make /run/lvm available within chroot
via mount --bind. Since there is no /run/lvm on the installation system and
since the path causing the error is /run/udev, I tried to bind-mount
/run/udev.
It seems to fix the problem. The command runs fast again.

I'm not sure if this is the best way to go or if this is just a usable
workaround.

These are the installed packages on the fai server:

fai-client 5.3.6
fai-quickstart 5.3.6
fai-server 5.3.6

Regards,
Tobias