Last update: 2008/04/30
Companies like 1and1 and Strato offer virtual servers based on the Virtuozzo virtualization technology. While these machines are quite cheap and provide a full linux work environment they run SUSE by default. Not my favorite linux distribution...
I was pretty certain that I could also switch the server to Gentoo. But when I asked the customer support they told me that they have no one running Gentoo on any of these machines. And that they would have no clue if that could work.
So I tried and it is definitely possible. Just in case there are others who would like to have a Gentoo vserver on a Virtuozzo system this HowTo will provide some instructions on how to achieve that.
Do I need to give the usual warnings? You'll completely wipe the old system and if something does not work, you will have to reinitialize the server. If you don't want to take that risk, do not continue.
Cleaning up
First you will have to log into your "Virtuozzo Power Panel" in order to switch the system into repair mode. The original system now resides in /repair and you work in a safety mode.
Now log into your system via ssh and make a backup copy the old /etc/mtab (this helps to have a working df command at a later time point, reported by Gian):
cp /repair/etc/mtab /root/mtab.old
Now remove the old suse system:
cd /repair rm -rf *
In case this results in a failure your repair directory might be mounted as read-only (reported by Ulrich):
mount -o remount,rw /repair
Install the basic Gentoo system
Now (still in /repair) start to download the stage and a portage snapshot from your nearest mirror:
wget ftp://linux.rz.ruhr-uni-bochum.de/gentoo-mirror/experimental/x86/vserver/stage3-i686-20060317.tar.bz2 wget ftp://linux.rz.ruhr-uni-bochum.de/gentoo-mirror/snapshots/portage-latest.tar.bz2 tar xvjpf stage3-*.tar.bz2 tar xvjf portage-*.tar.bz2 -C /repair/usr rm stage3-*.tar.bz2 portage-*.tar.bz2
The basic tools are now in place. Next we need the original network information:
cp /etc/resolv.conf /repair/etc/
In addition copy the original mtab back into place:
cp /root/mtab.old /repair/etc/mtab
And now we can chroot into the new Gentoo environment:
mount -t proc proc /repair/proc/ mount -o bind /dev /repair/dev chroot /repair
Time to fix the timezone information and sync the portage tree:
env-update source /etc/profile export PS1="(chroot) $PS1" cp /usr/share/zoneinfo/Europe/Berlin /etc/localtime emerge --sync
Set a root password:
passwd
Please note that this password becomes your new master password for the server!
Optional: Configure a build host
The vServers are not the most powerful machines and they definitely benefit from pulling packages from a central build host. If you have such a machine you should complete your /etc/make.conf with the following variables:
PORTAGE_BINHOST="http://buildhost.example.com/packages/i686/All" SYNC="rsync://buildhost.example.com/portage" EMERGE_DEFAULT_OPTS=" --usepkg --getbinpkg --getbinpkgonly"
Move to baselayout2
The old baselayout-vserver probably still works but the newer baselayout2 also copes for vServers and I recommend to use it.
First we should ensure that we link to the current Gentoo profile:
rm /etc/make.profile ln -s ../usr/portage/profiles/default-linux/x86/2007.0 /etc/make.profile
Now we unmask the newer baselayout and the OpenRC package:
echo "sys-apps/baselayout ~x86" >> /etc/portage/package.keywords echo "sys-apps/openrc ~x86" >> /etc/portage/package.keywords
In case the kernel of the system underlying your virtual server is somewhat older, you should also ensure that you do not use the newer glibc-2.4 and that nptl is disabled:
echo ">sys-libs/glibc-2.5-r4" >> /etc/portage/package.mask echo "sys-libs/glibc -nptl -nptlonly" >> /etc/portage/package.useTime to update the system:
emerge -uND world
Configure Gentoo as a virtual server
Now you can configure the network:
emerge iproute2 cd /etc/init.d rm net.eth0 ln -s net.lo net.venet0 rc-update add net.venet0 default rc-update add net.lo default
You will need to provide a static definition of your network parameters in /etc/conf.d/net. In order to determine the necessary parameters, follow the steps below:
# ip addr
326: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 brd 127.255.255.255 scope host lo
327: venet0: <BROADCAST,POINTOPOINT,NOARP,UP> mtu 1500 qdisc noqueue
link/void
inet 127.0.0.1/32 scope host venet0
inet 87.123.45.123/32 scope global venet0:0
From the output note the ip of the venet0 adapter. Here it is 87.123.45.123.
Now you need the routing information:
# ip route 191.255.255.0/24 dev venet0 scope link 127.0.0.0/8 dev lo scope link default via 191.255.255.1 dev venet0
The necessary parameters are the first netmask and the default gateway (191.255.255.0/24 and 191.255.255.1).
Adapt the following settings to your specific parameters and echo it into your network configuration file:
echo '
modules="iproute2"
modules="!ifconfig"
config_venet0="87.123.45.123 netmask 255.255.255.0 broadcast 0.0.0.0"
routes_venet0="191.255.255.0/24 scope link
default via 191.255.255.1"
' >> /etc/conf.d/net
I am not an expert on the network settings and the proper routing on a vserver but these settings did work for me. Please send me a mail if you have suggestions on how to improve the configuration.
Another comment by Ulrich:
I did put spaces between the config_venet0, routes_venet0 and the equal sign. It's not allowed to do so. Adding this as a don't to your explanation might save an hour or two to some guys out there.
Finally you need to add the ssh server to the default services so that you will be able to log into the system:
rc-update add sshd default
Reboot into Gentoo
Now you should be able to end the repair mode. Log into your Virtuozzo Power Panel, select "Finish repair" and try to log into your vserver via ssh a short while later.
ChangeLog
- 2008/04/30: Included moving to baselayout2
Thanks for this nice howto, worked w/o problems for me.
ReplyDeleteWhile digging through the 1&1 customer mirror I found a gentoo distfile and gentoo-portage rsync mirror! :)
http://update.onlinehome-server.info/distribution/gentoo/gentoo
rsync://update.onlinehome-server.info/gentoo-portage
The hosts seem only to be reachable from 1&1 ranges ..
also thanks from me...
ReplyDeletei was always looking for something like that!!!
unfortunatelly, for me it doesn't seem to work, because i always get an error telling me i cannot allocate enough memory, which is true. i just wonder why it doesn't seem to use the swap. in your tutorial i haven't found anything concerning swap. do i miss something?
cheers!
alex
OR you could just move to a provider that offers Gentoo Virtual Private Servers.
ReplyDelete@alex: I am using a buildhost structure with the vServers. Otherwise I'd get the same memory problems on the smaller hosts.
ReplyDeleteThat way I avoid problems with swap and I believe you don't get swap with the vServers from 1&1. But since I rely on my buildhost I was never really pressured into asking them. Maybe it is possible to activate swap somehow.
@anonmyous: The krombie.net link is definitely interesting. I always enjoy it to see Gentoo being used in a commercial environment.
But you must admit that they are extremely expensive. At these rates it quickly becomes worthwhile to convert a cheap vServer to Gentoo yourself.
See http://linux-vserver.org/VServer_Hosting and search for 'gentoo', there are some hosting provider which offer gentoo vservers out of the box.
ReplyDeleteYou should do:
ReplyDeletemount -o bind /dev/pts/ dev/pts/
If you get PTS errors
And don't forget to edit your fstab ;)
ReplyDeleteI've used this guide for two Hosting Europe servers. With VPS 3.0 SuSE, everything worked fine. With VPS 4.0 Ubuntu, you also had to copy over the /etc/network directory and its content from the rescue/repair system to Gentoo.
ReplyDeleteBTW: Don't get confused when the repair systems routing table and network config looks bit confusing (like 127.0.0.1 on loopback and as a second IP for venet0). Just do it as described above.
I managed to do this with a Host Europe virtual server. Like in the comment above, I had to copy over the /etc/network directory, otherwise the system refused to boot.
ReplyDeleteIt might be worth noting that chrooting to the Gentoo directory and emerging stuff often failed on the rescue system, sometimes it told me that it “Could not allocate memory”, sometimes portage failed with not finding some Python classes without any reasonable explanation for that. In both cases it helped to reboot the rescue system or to just try again at a different time of day, as the rescue system at Host Europe does not seem to run on the same machine as the virtual server usually does (it seems to be a much faster machine but you probably share the memory with more people).
Edit /etc/conf.d/net:
ReplyDeleteconfig_venet0=( "YOUR IP netmask 255.255.255.255" )
routes_venet0=( "191.255.255.1/32" "default via 191.255.255.1" )
and don't forget:
ln -s /etc/init.d/net.lo /etc/init.d/net.venet0
rc-update add net.venet0
rc-update add sshd
cp -R /etc/network /repair/etc/
After that you can start without repair mode and without merging iproute2 so you can avoid out of memory errors.
This guide seems to be outdated since profile is selected via eselect and openrc and baselayout2 are stable now (distributed with the stage archive).
Besides I used http://gentoo.arcticnetwork.ca/releases/amd64/current-stage3/stage3-amd64-20111208.tar.bz2 and no vserver stage which seemed to be outdated, as well.