The new Proxmox got released http://pve.proxmox.com/wiki/Roadmap#Proxmox_VE_4.0 on 05.10.2015. You probably ended up here by typing “Proxmox update cluster from 3 to 4” to google. In this article I will outline how i updated my “poor man’s cluster”. My proxmox cluster currently consists of 5 servers hosted @ 3 Datacentres. I will try to add screenshots etc because most people i know prefer a hands on description rather then the technical writeups.
According to the migration paper which can be found @ https://pve.proxmox.com/wiki/Upgrade_from_3.x_to_4.0
Proxmox VE 4.0 introduces major new features, therefore the upgrade must be carefully planned and tested. Depending on your existing configuration, several manual steps are required, including some downtime. NEVER start the upgrade process without a valid backup and without testing the same in a test lab setup.
Major upgrades for V4.0:
- OpenVZ is removed, a conversion via backup/restore to LXC is needed
- New corosync version, therefore clusters has to be re-established
- New HA manager (replacing RGmanager, involving a complete HA re-configuration)
Generally speaking there are two possibilities to move from 3.x to 4.0
- In-place upgrade via apt, step by step
- New installation on new hardware (and restore VM´s from backup) – safest way.
In order to keep the downtime minimal i will try the “In-place upgrad via apt, step by step guide”. So lets dive in
Mixed cluster of v3 and v4 not possible:
Why bother to update?
There are a lot of reasons why to update to the new versions but the key reasons that motivated me to update are:
- New Kernel Versions
- Put OpenVZ to rest
- Introduction of LXC
- High Availability ( HA ) that supposedly works out of the box
Quick summary of steps for proxmox update cluster
step 1 – 5 Servers in cluster, remove all vms from server1 move over to other servers.
step 2 – change dns of all vms that ran on server1 to point to the new locations
step 3 – update server1 to new proxmox 4.0 ( remove from current cluster )
step 4 – create new cluster on server1 using proxmox 4.0
step 5 – import vms that previously ran on server1 to the new cluster on server1
step 6 – repeat step 1-5 for every other server
1 Migrate all Openvz Containers to KVMs/LXC
OpenVZ is not available for Kernels above 2.6.32, therefore a migration is necessary. Linux Container technology is available in all mainline Linux kernels and a future proof technology introduced in Proxmox VE 4.x series.You have a few options here
- Move the applications that are running inside openvz containers over to KVMs
- Convert the openvz containers to LXC containers – tutorial
2 Backup all existing systems
I hope you already have an automated system to backup your virtual machines regularly, if not create a system of each VMknow, in case of a zombie apocalypse your should be prepared. You can create backups inside proxmox easily using the webinterface, your backups’ default location for kvms on your servers is /var/lib/vz/dump. Most server providers offer a few 100GB of backup space accesible via nfs/ftp etc, this backup space is ideal to place your backups.
3 Clean up your Test-Server
So from the 5 servers i have running i have physical access to 1 of those servers, this will be the server i ll be testing the updated from 3 to 4 on. The servername of that server is Berlin1. The other Servers are located in datacentres, because they need more traffic then my local ISPs can handle. All servers are connected via OpenVPN to allow multicasts to be sent between the servers.
So my first task is to move all virtual machines running on this server to other servers in my cluster, so these vms remain online and i can test this update process on berlin1 without affecting anybody else. Then once i know how the updates work, I can roll-out these changes on the other production systems during off-peak hours ( yes nightshift! 🙂 )
I do keep regular backups of my Berlin1 node on other Servers. So for me it was faster to use the last Backups of these and import them on the other servers. For you it might be faster to simply migrate the KVMs from one server to another. So now my Berlin1 node is free of any VMs – as naked as it hasn’t been since setting it up last year.
4 One last check before take off
Another peak at the Upgrade wiki of Proxmox tells us to ensure that we have met our preconditions for the in-place upgrade
- upgraded to latest V3.4 version ( Check )
- reliable access to all configured storages ( Check: A copy of each VM on the Server and 2 redundant off-site backups )
- healthy cluster ( Check )
- no VM or CT running (note: VM live migration from 3.4 to 4.0 node or vice versa NOT possible) ( Check )
- valid backup of all OpenVZ containers (needed for the conversion to LXC) ( Check )
- valid backup of all VM (only needed if something goes wrong) ( Check )
- Correct repository configuration (accessible both wheezy and jessie) ( Check )
- at least 1GB free disk space at root mount point ( Check )
5 Remove Proxmox VE 3.x packages
Since I use the same aptitute repository lists i have to ensure all servers have the same software installed. So we ll start my making sure they have the same software installed
Then we remove Proxmox VE 3.x packages in order to avoid dependency errors
6 Update the repository
We adapt repository locations and update the apt database, point all to jessie:
In case Ceph server is used: Ceph repositories for jessie can be found at http://download.ceph.com, therefore etc/apt/sources.list.d/ceph.list will contain
7 Install the new Kernel
A new version of OpenSSH will be installed, by default this newer version does not allow root logins with a password, so make sure you have a non root user to login later, this update process also removes the server from the cluster so if you plan on using the keys you added when you created a cluster this will not work. to add a user simply connect to your server and use “adduser newuser”
First you search for the new kernel by using
Then you install the latest version in my example 4.2.3-2
So all thats left to be done is run the dist-upgrade command to turn our debian wheezy into a debian jessie:
And finally reboot the server into the new kernel version:
After the reboot we check if we are now running the proper kernel by running
And then we remove the old kernel
and install our loved proxmox tools
It is not possible to mix Proxmox VE 3.x and earlier with Proxmox VE 4.0 cluster
Due to the new corosync 2.x, the cluster has to be re-established again. So we can remove berlin1 from the old cluster.
8 Create new Cluster
After the installation we can login to see that the cluster is there, but neither working nor healthy
Now we can create a new Cluster with our first Server by simply typing
Make sure you use the same cluster name as you had before
So you remember the beginning, my berlin1 has been running roughly 20 Virtual Machines, which i moved over to other hosts, so right now these VMs are running on the other hosts. I will in the next step import all VMs back to berlin1, shut down the clones on the other servers 1 by 1 and update dns records so we can slowly migrate users over. ( Good tip for this to lower your dns’s TTL during all these moving back and forth ). But eventually we want to get all the servers back to the newly created cluster – and to create/join a cluster each proxmox server has to have no vms running.
9 Cleanup entries from old Cluster
So at this moment initially it was a bit unclear to me if that now means that the old configuration was ported over to the new cluster. After checking pvecm nodes and pvecm status, i came to the conclusion that the other nodes are not actually linked to this server anymore and that i only have left over data inside my /etc/pve/nodes folder. So i changed to that directory and removed my old nodes ( proxmox1,2,5,6 )
The Webui then updated and now only shows me the one node.
We don’t use windows here – but we ll do a reboot anyways to it to recover without having to track down much here. After the server returns the light is now green:
But we are greeted by the lovely error message: “cluster not ready. no quorum? (500)”
Now that the quorum is there I can create the vm and start importing the last backups:
Next on the road: create new vms for each of the backups and start importing. Coffee++ time ….
10. Do it to the next servers
We do the same to the other servers in the cluster with 1 small difference instead of pvecm create IME we join the cluster using.
After running these commands on proxmox6.internetz.local the node was showing up red initially, but after 1 reboot we are in business: