Difference between revisions of "Migration from Linux-VServer to OpenVZ"

From OpenVZ Virtuozzo Containers Wiki
Jump to: navigation, search
(Added note to remove udev from ubuntu guest)
m (/dev)
Line 198: Line 198:
  
 
Vserver mounts /dev/pts filesystem for container transparently, whereas openvz does not. To compensate the ommission, you need to move aside /dev directory in the vserver-originating container and copy /dev directory from openvz based container.
 
Vserver mounts /dev/pts filesystem for container transparently, whereas openvz does not. To compensate the ommission, you need to move aside /dev directory in the vserver-originating container and copy /dev directory from openvz based container.
 +
 +
=== Ubuntu udev ===
  
 
Additionally, Ubuntu based vservers have the udev package installed which prevents access to the console in openvz. This error message is an example of the problem:
 
Additionally, Ubuntu based vservers have the udev package installed which prevents access to the console in openvz. This error message is an example of the problem:
Line 234: Line 236:
 
root@test:/#
 
root@test:/#
 
</code>
 
</code>
 
  
 
=== /proc ===
 
=== /proc ===

Revision as of 23:52, 21 December 2009

This article describes the migration from Linux-VServer to OpenVZ.

Details of migration process

Initial conditions

The following example of Linux-VServer based solution was used for the experiment:

  • Kernel linux-2.6.17.13 was patched by the patch-2.6.17.13-vs2.0.2.1.diff and rebuild;
  • Util-vserver-0.30.211 tools were used for creating containers;
 # vserver-info
 Versions:
 Kernel: 2.6.17.13-vs2.0.2.1
 VS-API: 0x00020002
 util-vserver: 0.30.211; Dec  5 2006, 17:10:21

 Features:
 CC: gcc, gcc (GCC) 3.4.3 20041212 (Red Hat 3.4.3-9.EL4)
 CXX: g++, g++ (GCC) 3.4.3 20041212 (Red Hat 3.4.3-9.EL4)
 CPPFLAGS: 
 CFLAGS: '-g -O2 -std=c99 -Wall -pedantic -W -funit-at-a-time'
 CXXFLAGS: '-g -O2 -ansi -Wall -pedantic -W -fmessage-length=0 -funit-at-a-time'
 build/host: i686-pc-linux-gnu/i686-pc-linux-gnu
 Use dietlibc: yes
 Build C++ programs: yes
 Build C99 programs: yes
 Available APIs: v13,net
 ext2fs Source: kernel
 syscall(2) invocation: alternative
 vserver(2) syscall#: 273/glibc

 Paths:
 prefix: /usr/local
 sysconf-Directory: ${prefix}/etc
 cfg-Directory: ${prefix}/etc/vservers
 initrd-Directory: $(sysconfdir)/init.d
 pkgstate-Directory: ${prefix}/var/run/vservers
 vserver-Rootdir: /vservers
 #

VServer v345 was built using vserver vX build utility and populated by using the tarballed template of Fedora Core 4.

 # vserver v345 start
 Starting system logger:                                    [  OK  ]
 Initializing random number generator:                      [  OK  ]
 Starting crond: l:                                         [  OK  ]
 Starting atd:                                              [  OK  ]
 # vserver v345 enter
 [/]# ls -l
 total 44
 drwxr-xr-x    2 root     root         4096 Oct 26  2004 bin
 drwxr-xr-x    3 root     root         4096 Dec  8 17:16 dev
 drwxr-xr-x   27 root     root         4096 Dec  8 15:21 etc
 -rw-r--r--    1 root     root            0 Dec  8 15:33 halt
 drwxr-xr-x    2 root     root         4096 Jan 24  2003 home
 drwxr-xr-x    7 root     root         4096 Oct 26  2004 lib
 drwxr-xr-x    2 root     root         4096 Jan 24  2003 mnt
 drwxr-xr-x    3 root     root         4096 Oct 26  2004 opt
 -rw-r--r--    1 root     root            0 Dec  7 20:17 poweroff
 dr-xr-xr-x   80 root     root            0 Dec  8 11:38 proc
 drwxr-x---    2 root     root         4096 Dec  7 20:17 root
 drwxr-xr-x    2 root     root         4096 Oct 26  2004 sbin
 drwxrwxrwt    2 root     root           40 Dec  8 17:16 tmp
 drwxr-xr-x   15 root     root         4096 Jul 27  2004 usr
 drwxr-xr-x   17 root     root         4096 Oct 26  2004 var
 [/]# sh
 sh-2.05b#
 .........

As a result we obtain running virtual environment v345:

 # vserver-stat

 CTX   PROC    VSZ    RSS  userTIME   sysTIME    UPTIME NAME
 0       51  90.9M  26.3M   0m58s75   2m42s57  33m45s93 root server
 49153    4  10.2M   2.8M   0m00s00   0m00s11  21m45s42 v345

 # 

Starting migration to OpenVZ

Downloading and installing the stable OpenVZ kernel.

Install the OpenVZ kernel, as described in Quick installation.

After the kernel is installed, reboot the machine. After rebooting and logging in you will see the following reply on vserver-stat call:

 # vserver-stat
 can not change context: migrate kernel feature missing and 'compat' API disabled: Function not implemented

It is a natural thing that now virtual environment v345 is unavailable. The following steps will be devoted to making it work over OpenVZ kernel.

Downloading and installing vzctl package

OpenVZ solution requires installing a set of tools: vzctl and vzquota packages. Download and install it, as described in quick installation.

If rpm complains about unresolved dependencies, you'll have to satisfy them first, then repeat the installation. Then launch the OpenVZ:

 # /sbin/service vz start
 Starting OpenVZ:                                           [  OK  ]
 Bringing up interface venet0:                              [  OK  ]
 Configuring interface venet0:                              [  OK  ]

Currently vzlist utility is unable to find any containers:

 # vzlist
 Containers not found

Updating different configurations

Move the existing guest OSs to the right place:

 # cd /vz
 # mkdir private
 # mkdir private/345
 # mv /vservers/v345 /vz/private/345

Now it is time for creating configuration files for OpenVZ container. Use the basic sample configuration presented in /etc/sysconfig/vz-scripts/ve-vps.basic.conf-sample file:

 # cd /etc/sysconfig/vz-scripts
 # cp ve-vps.basic.conf-sample 345.conf

Now, let's set some parameters for the new container.

First, we need to tell which distro the container is running:

 # echo "OSTEMPLATE="fedora-core-4" >> 345.conf

Then we set a few more parameters:

 vzctl set 345 --onboot yes --save # to make it start upon reboot
 vzctl set 345 --ipadd 192.168.0.145 --save
 vzctl set 345 --hostname test345.my.org --save

Testing how the guest OSs successfully work over OpenVZ

Now you can start a container:

# vzctl start 345

and see if it's running:

 # vzlist -a
 CTID      NPROC  STATUS  IP_ADDR         HOSTNAME
 345          5   running 192.168.0.145   test345.my.org

You can run commands in it:

 # vzctl exec 345 ls -l
 total 48
 drwxr-xr-x    2 root     root         4096 Oct 26  2004 bin
 drwxr-xr-x    3 root     root         4096 Dec 11 12:42 dev
 drwxr-xr-x   27 root     root         4096 Dec 11 12:44 etc
 -rw-r--r--    1 root     root            0 Dec 11 12:13 fastboot
 -rw-r--r--    1 root     root            0 Dec  8 15:33 halt
 drwxr-xr-x    2 root     root         4096 Jan 24  2003 home
 drwxr-xr-x    7 root     root         4096 Oct 26  2004 lib
 drwxr-xr-x    2 root     root         4096 Jan 24  2003 mnt
 drwxr-xr-x    3 root     root         4096 Oct 26  2004 opt
 -rw-r--r--    1 root     root            0 Dec  7 20:17 poweroff
 dr-xr-xr-x   70 root     root            0 Dec 11 12:42 proc
 drwxr-x---    2 root     root         4096 Dec  7 20:17 root
 drwxr-xr-x    2 root     root         4096 Dec 11 12:13 sbin
 drwxrwxrwt    2 root     root         4096 Dec  8 12:40 tmp
 drwxr-xr-x   15 root     root         4096 Jul 27  2004 usr
 drwxr-xr-x   17 root     root         4096 Oct 26  2004 var

Issues

Networking

Starting networking in VEs

The vserver-originating containers do not initialize network at all. Thus one needs to use following command to enable networking start (inside of the migrated container):

cd /etc/rcS.d
ln -s ../init.d/networking S40networking

Migrating your VServer Shorewall setup

If you had the Shorewall firewall running on the hardware node to route traffic to and from your guests, here are a couple of advices, provided you want a networking setup close to what you had with Vserver (i.e. running vnet interfaces, not veth ones) :

  • do not use the venet0 interface in Shorewall's configuration as the vz service starts after Shorewall (at least on Debian) and thus the interface does not exist when Shorewall starts. Do not use detect for the broadcast in /etc/shorewall/interfaces.
  • for your VEs to be able to talk to each other, use the routeback option for venet0 (and others) in /etc/shorewall/interfaces.

IP src from VEs

If you run a mail server in a VE, and if the hardware node has multiple network interfaces, you may have mail routing issues because of the originated IP address of the packets coming from the hardware node. Simply specify an interface in /etc/vz/vz.conf :

VE_ROUTE_SRC_DEV="iface_name"

Disk space information

Disk space information is empty. Do the following to fix:

rm /etc/mtab
ln -s /proc/mounts /etc/mtab

/dev

Vserver mounts /dev/pts filesystem for container transparently, whereas openvz does not. To compensate the ommission, you need to move aside /dev directory in the vserver-originating container and copy /dev directory from openvz based container.

Ubuntu udev

Additionally, Ubuntu based vservers have the udev package installed which prevents access to the console in openvz. This error message is an example of the problem:

  1. vzctl enter 101

enter into CT 101 failed Unable to open pty: No such file or directory

The fix is to remove the udev package from the guest:

  1. vzctl exec 101 'dpkg --force-depends --purge udev'

(Reading database ... dpkg: udev: dependency problems, but removing anyway as you request:

initramfs-tools depends on udev (>= 117-5).

15227 files and directories currently installed.) Removing udev ... Purging configuration files for udev ... dpkg - warning: while removing udev, directory `/lib/udev/devices/net' not empty so not removed. dpkg - warning: while removing udev, directory `/lib/udev/devices' not empty so not removed.

Now restart the container, you should now be able to use the console.

  1. vzctl restart 101

Restarting container ...

<SNIP>

... Container start in progress...

  1. vzctl enter 101

entered into CT 101 root@test:/#

/proc

The /proc filesystem is not automatically mounted by openvz. So the vserver needs to mount it itself. The simplests (not the best) way it can be done, is by sticking following command at the end of /etc/init.d/bootmisc.sh:

mount /proc