On Premise¶
This documentation describes how to perform the version upgrade operation of an on-premise Traefik Enterprise cluster.
Backing Up the Outdated Cluster¶
First, follow the documentation on backing up a cluster to generate a backup archive, which will be used to configure a new cluster.
Upgrading traefikee
and teectl
¶
The teectl
and traefikee
tools must be upgraded before starting the cluster version upgrade: follow the installing teectl
guide to install the latest version of teectl
and use the on-premise install guide to get the latest version of the traefikee
binary.
Run teectl version
and traefikee version
to ensure the correct versions are installed.
Replacing the Outdated Cluster¶
To upgrade a cluster, install a new one running in parallel with the outdated cluster. Once the new cluster is reachable, the outdated cluster can be uninstalled. The replacement cluster will need to be created on separate hosts/VMs.
In order to install a second cluster without conflicting with the existing one, it has to be installed under a different cluster name. This means all teectl
commands must specify the --cluster
option with the new cluster name.
Version upgrade from v1 to v2
To update a v1 cluster to v2, first make sure that the existing cluster is running at least version v1.3. Back it up and then use the cluster migration tool to transform the v1 backup into a v2 backup that can be used in the upgrade procedure.
- Run the same
traefikee
commands as you did for your previous Traefik Enterprise installation, with a different--api.addr
option for thetraefikee controller
command. - Wait for your the cluster to be reachable.
- Redirect traffic to the new cluster. This could be done by updating the external DNS entry to point to the new load balancer service's IP, or by editing your L4 load balancer's configuration, for example.
- Ensure that your applications are reachable through the new cluster.
- Interrupt all processes of the previous cluster in order to uninstall it.
The new cluster should now be operating properly and the upgrade process should not have dropped any application traffic. The outdated cluster can safely be removed.