"Unable to start init, probably incorrect template" in certain circumstances (several moun [message #38946] |
Tue, 23 February 2010 19:52  |
januszzz
Messages: 50 Registered: January 2007 Location: Opole, Poland
|
Member |
|
|
Hi, I got real issue and please do not tell me this is feature, not bug either, things are very wrong here read on:
I got squashfs filesystem which is obviously mounted read-only. It icludes real openvz environment as well as single VE. The HE works perfectly, but Ve cannot start:
Warning: configuration file for distribution nx not found default used
Starting container ...
Container is mounted
Setting CPU units: 1000
Configure meminfo: 65536
Configure veth devices: veth102.0 veth102.1
Unable to start init, probably incorrect template
Container start failed
Stopping container ...
Container was stopped
Container is unmounted
I need to be able to write on some parts of squashfs, so I copy on boot /etc/ /var and some more directories. The rest of VE's directories are mounted with -o bind option, i.e.:
/mnt/livecd/vz/private/102/sbin on /vz/private/102/sbin type none (rw,bind)
/mnt/livecd/vz/private/102/sys on /vz/private/102/sys type none (rw,bind)
/mnt/livecd/vz/private/102/usr on /vz/private/102/usr type none (rw,bind)
etc.
You would say, that template is wrong, huh? its not, it is perfectly right as when I use rsync to copy the whole container to internal hdd, I can start the container without any problem:
rsync -avH /vz/private/102 /mnt/hdd
mount -o bind /mnt/hdd /vz/private/102
vzctl start 102
Warning: configuration file for distribution nx not found default used
Starting container ...
Container is mounted
Setting CPU units: 1000
Configure meminfo: 65536
Configure veth devices: veth102.0 veth102.1
Container start in progress...
What is happening then? a bug? a feature? maybe I shouldn't have mounted some dirs (proc,sys, I dunno what yet?)
I would be very thankful for any help!
[Updated on: Tue, 23 February 2010 19:54] Report message to a moderator
|
|
|
|
Re: "Unable to start init, probably incorrect template" in certain circumstances (several [message #38957 is a reply to message #38948] |
Wed, 24 February 2010 10:07   |
januszzz
Messages: 50 Registered: January 2007 Location: Opole, Poland
|
Member |
|
|
Yes, init is in place, is executable. As I've said, when I blindly copy all the files the new container starts OK and I can always chroot into the container - both mounted with -o binds and the one copied (chroot /vz/private/102 /bin/bash, all here seems right)
In my opinion some bind mounts prevents entering the container.
But there is easy way to try the situation - just take a container and remount -o bind every single directory to another, like this:
cd /vz/private/102
mkdir 103
ls -1 | while read x; do mkdir /vz/private/103/$x; mount -o bind /vz/private/102/$x /vz/private/103/$x; done
then starting new container fails with the above command.
I do bind mounts previously to starting container, I do not use 102.mount script. The question I don't know the answer is if I could mount almost every single dir using the 102.mount script (all except etc,dev and var)? I haven't tried that yet.
[Updated on: Wed, 24 February 2010 10:16] Report message to a moderator
|
|
|
|
|
|
|