networking and lvm mounts

Bug #37216 reported by IC Raibow
6
Affects Status Importance Assigned to Milestone
lvm-common (Ubuntu)
Invalid
Low
Unassigned

Bug Description

I have /var mounted on LVM. As `S0?loopback' is starting at very before S26lvm and S35mountall, it complains with can't create dir:
  [ -d /var/run/network ] || mkdir /var/run/network

So we got here non-working dbus and all that stuff, resulting in veeeeryyy looong logging in process and semi-working dbus-using apps.

Strangely, S40networking borken in right same way.

Solved it by putting "ifdown" right before "ifup" stuff in loopback and networking init scripts. Also moved S0?loopback to S36loopback in /etc/rcS.d/

Revision history for this message
ADORJÁNI Gábor (adorjanig-deactivatedaccount) wrote :

Same problem here, thanks wiz for the tips.

After a basic install I put /var onto a logical volume. Everything's okay, reboot and I just stared stupidly why dnscache went crazy saying there's no loopback interface to bind to. :o

IMHO it was quite a bad idea to mount a tmpfs on /var/run before any filesystem comes up, assuming the whole /var structure is residing on root. Since it's quite common on servers to put everything possible onto separate partitions/LVs, it should be expected that /var is one them. From a security standpoint I don't want the possibilty that logs fill up my root FS.

Oh, BTW and it's on a Kubuntu 6.06LTS install on a dual AMD Opteron box, amd64 architecture.

Revision history for this message
Rich Johnson (nixternal) wrote :

Is this issue still occurring with the latest release?

Revision history for this message
IC Raibow (icrbow) wrote :

As for xubuntu 6.10 order is:

  S01mountkernfs.sh
  S01readahead
  S02hostname.sh
  S06keyboard-setup
  S07linux-restricted-modules-common
- S08loopback
  S10udev
  S11mountdevsubfs.sh
  S13pcmciautils
  S15module-init-tools
  S17procps.sh
  S20checkroot.sh
  S22mtab.sh
  S25mdadm-raid
- S26lvm
  S30checkfs.sh
  S35mountall.sh
  S36mountall-bootclean.sh
  S39readahead-desktop
- S40networking
  S45waitnfs.sh

So, it looks like more Edgy users could confirm this tickets anytime.

Is there really need of loopback even before root is up?

Revision history for this message
Scott James Remnant (Canonical) (canonical-scott) wrote :

This implies missing mountkernfs

Changed in lvm-common:
status: Confirmed → Rejected
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.