Subject: Re: [vserver] Guest vserver stops mysteriously
From: "Daniel Hokka Zakrisson" <daniel@hozac.com>
Date: Tue, 12 Aug 2008 18:34:21 +0200 (CEST)

Ravi Kumar wrote:
> Hello Friends
> I am running Ubuntu (64-bit) as guest OS. For Vservers i run
> kernel-2.6.22.19 with vserver patch vs2.2.0.7  .
>
> I want guest vservers to run on a partiton /dev/sda4 which i mount to
> /media/disk using the command :
>
> sudo mount /dev/sda4 /media/disk
>
> I dont use the normal ubuntu gui method to mount the disk as it mounts it
> with the option nodev, and i read somewhere that it may cause problems in
> vservers. Anyway before creating guest vservers i ran testme.sh and
> testfs.sh on /media/disk as root and all tests succeeded except xfs and jfs
> filesystems (in testfs.sh).
>
> I downloaded yum sources from git and reset it to the version 3.2.4 (for
> which the patch is available for vservers), and after applying the patch,
> did 'make install' on host (ubuntu). Now it doesnt complain about using
> special hacks in yum.

The 3.2.4 patch applies to (at least) 3.2.4-3.2.8.

> The guest OS i built is centos5 (32-bit) using following command :
>
> sudo linux32 vserver vserver1 build \
> -m yum \
> --rootdir /media/disk/vsrvs/vservers/.defaults/vdirbase \
> --pkgbase /media/disk/vsrvs/vservers/.pkg \
> --confdir /media/disk/vsrvs/vservers/vserver1 \
> --hostname vserver1.localdomain.com \
> --interface eth0:143.46.43.100 \
> -- \
> -d centos5
>
> after that i tried to start the guest OS by the command :
>
> cd /media/disk/vsrvs/vserver
> sudo vserver -v $PWD/vserver1 start
>
> but it shown the following :
>
> New network context is 40011
> Adding 143.46.43.100
> New security context is 40011
> Bringing up loopback interface:  Device lo does not seem to be present,
> delaying initialization.
>                                                            [FAILED]
> Mounting other filesystems:                                [  OK  ]
> $
>
> (you can see that it returned to shell prompt '$' without any furthur output
> or error).
> I also tried prepending 'linux32' to the command (after sudo), but got same
> message.
>
> and if i try to 'enter' into vserver1 or check its 'status' it says that it
> is stopped. (i always give full path to vserver1, as it is not at its
> standard location)

Install a recent snapshot of util-vserver. CentOS 5.2 includes rsyslog which is a
shorter name than sysklogd, but that initscript is not enabled by default. Changeset
2722 fixed the problem, so any snapshot newer than that will work fine.

> after that i did descended into vserver1/interfaces folder and did this :
>
> $ sudo mv 0 1
> $ sudo mkdir 0
> $ cd 0
> $ sudo bash -c "echo 'lo' > dev"
> $ sudo bash -c "echo '127.0.0.1' > ip"
> $ sudo bash -c "echo '255.0.0.0' > mask"
> $ sudo bash -c "echo 'lo' > name"
> $ cd ../..
>
> (simply i moved the original interface '0' to '1' and created new interface
> '0' with dev as 'lo', ip as '127.0.0.1', mask as '255.0.0.0' and name as
> 'lo')

As Herbert said, don't do that.

> [...]

-- 
Daniel Hokka Zakrisson