[Bug 1185394] Re: systemd-udev fails when processing many logical volumes on boot

[Stefan Bader]

(1) Struggeling a bit with weirdly different behaviour between boots as
Xen host and "normal" boots. First there was "debug" added for the Xen
boot. This should only cause kernel messages of lower priority to be
sent to the console, but it seems to have the effect of not showing any
output of the initramfs hooks.

Second, for some reason a normal boot will take quite a while between
udevadm control --exit starting and finishing. But when it finishes, no
processes with udevd in they name remain. Booting in Xen mode there is
one systemd-udevd crashing immediatly after calling udevadm control
--exit and that ends the udevadm command (so it does not wait). After
that it takes another 25s until the remaining systemd-udevd processes.
That would explain that those missing /dev/null messages cannot be
reproduced without booting in Xen mode.

In both cases it seems that at least one (in the current case 2)
instance(s) of "sh -c vgscan; vgchange -ay" are still running (and the
actually stuck part seems to be vgchange). And that seems to trigger (in
both cases after udevadm control --exit has returned) the messages about
the timeout when trying to kill (one of?) these. That results in some of
the LVs not being properly set up. So far the /dev/dm-X entry was
present (dmsetup ls also shows them) but the symlinks from /dev/mapper/
and /dev/<vgname>/ pointing back to /dev/dm-X are missing. And that can
result in mountall waiting. Which in turn seems to fail showing or
accepting anything to skip when I redirect the console to a serial
device (if I don't it is hard to look at old messages and secondly I
usually prefer to use the serial over lan to interact with that server).

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to systemd in Ubuntu.
https://bugs.launchpad.net/bugs/1185394

Title:
  systemd-udev fails when processing many logical volumes on boot

Status in “systemd” package in Ubuntu:
  In Progress

Bug description:
  With recent systemd/udev updates in Saucy I see problems when /dev
  gets moved to the real root. It seems to be related to running
  vgscan/vgchange through watershed by systemd-udevd. The symptoms
  include a crash of one systemd-udevd process and several messages
  about systemd-udevd (or childs of it) not finding files in /dev.

  Setup:

  sda: contains the root fs (no separate /boot) and a swap partition
  sdb: contains vg1 with /home
  sdc and sdd: contain vg2 with 31 LVs, one of them mounted to /home/isos

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1185394/+subscriptions