[lxc-devel] Fedora containers don't appear to boot

Daniel Lezcano daniel.lezcano at free.fr
Mon Jun 22 20:55:33 UTC 2009


Matty wrote:
> Howdy,
>
> I grabbed the latest lxc bits from git, and installed them on my
> Fedora 11 host (this system is running a 2.6.30 kernel from rawhide).
> With the lxc-fedora changes I submitted to the list earlier today, I
> am able to install a Fedora container with lxc-fedora:
>
> $ lxc-fedora create
> What is the name for the container ? [fedora] fedora
> What hostname do you wish for this container ? [fedora] fedora
> What IP address do you wish for this container ? [172.20.0.21] 192.168.1.200
> What is the gateway IP address ? [172.20.0.1] 192.168.1.1
> What is the MTU size ? [1500]
> Checking cache download ...Found.
> Copying rootfs ...Done.
> Done.
>
> When I run 'lxc-start -n fedora', the container appears to hang (I
> left it running over night, and it didn't make any progress booting).
> The only active process I see is udevadm:
>
> root      1848  0.0  0.0  11112  1224 ?        Ss   12:37   0:00
> /bin/sh -e -c ?runlevel --set S >/dev/null ||
> true???/etc/rc.d/rc.sysinit ??runlevel --reboot || true? /bin/sh
> root      1853  0.0  0.0  11380  1680 ?        S    12:37   0:00
> /bin/bash /etc/rc.d/rc.sysinit
> root      1885  0.0  0.0  11380  1620 ?        S    12:37   0:00
> /bin/bash /sbin/start_udev
> root      1911  0.0  0.0  10324   524 ?        S<s  12:37   0:00 /sbin/udevd -d
> root      2009  0.5  0.0   8232   688 ?        S    12:37   0:00
> /sbin/udevadm settle
>
> Which appears to be repeatedly checking for the existence of
> /dev/.udev/queue and then opening the seqnum file (the following five
> lines loop over and over):
>
> nanosleep({0, 50000000}, NULL)          = 0
> stat("/dev/.udev/queue", 0x7fffd1dc2940) = -1 ENOENT (No such file or directory)
> open("/sys/kernel/uevent_seqnum", O_RDONLY) = 3
> read(3, "2039\n"..., 32)                = 5
> close(3)                                = 0
>
>
> I'm not entirely sure if there is a specific event that needs to occur
> to allow the daemon to continue, but the fact that it hangs seems odd.
> Has anyone on the list seen this behavior with their Fedora guests? I
> checked the archives for solutions, but didn't seeing anything that
> matched. :(
>
> Thanks for any insight,
> - Ryan
>   

I checked the lxc-fedora template and found there are a lot of useless 
services and which have no sense for the container.

chkconfig cpuspeed off
chkconfig acpid off
chkconfig cups off
chkconfig bluetooth off
chkconfig avahi-daemon off
chkconfig irqbalance off
chkconfig mdmonitor off
chkconfig messagebus off
chkconfig udev-post off

I commented the udev script launch in /etc/rc.sysinit

Changed the root password with
    chroot rootfs.fedora pwconv
    chroot rootfs.fedora passwd

Added the tty's in the /dev's rootfs

and launched the fedora, I booted but did not launched the network 
neither sshd but I was able to log to the console with the lxc-console 
command.

It looks like there is still a lot of things to do for this script :/
eg: make the installation more "minimal" :) , reduce the number services 
launched at boot time, etc ...,

pfou !

   -- Daniel




More information about the lxc-devel mailing list