Upgrade your cluster
Constellation provides an easy way to upgrade all components of your cluster, without disrupting it's availability.
Specifically, you can upgrade the Kubernetes version, the nodes' image, and the Constellation microservices.
You configure the desired versions in your local Constellation configuration and trigger upgrades with the apply
command.
To learn about available versions you use the upgrade check
command.
Which versions are available depends on the CLI version you are using.
Update the CLI
Each CLI comes with a set of supported microservice and Kubernetes versions. Most importantly, a given CLI version can only upgrade a cluster of the previous minor version, but not older ones. This means that you have to upgrade your CLI and cluster one minor version at a time.
For example, if you are currently on CLI version v2.6 and the latest version is v2.8, you should
- upgrade the CLI to v2.7,
- upgrade the cluster to v2.7,
- and only then continue upgrading the CLI (and the cluster) to v2.8 after.
Also note that if your current Kubernetes version isn't supported by the next CLI version, use your current CLI to upgrade to a newer Kubernetes version first.
To learn which Kubernetes versions are supported by a particular CLI, run constellation config kubernetes-versions.
Migrate the configuration
The Constellation configuration file is located in the file constellation-conf.yaml
in your workspace.
Refer to the migration reference to check if you need to update fields in your configuration file.
Use constellation config migrate
to automatically update an old config file to a new format.
Check for upgrades
To learn which versions the current CLI can upgrade to and what's installed in your cluster, run:
# Show possible upgrades
constellation upgrade check
# Show possible upgrades and write them to config file
constellation upgrade check --update-config
You can either enter the reported target versions into your config manually or run the above command with the --update-config
flag.
When using this flag, the kubernetesVersion
, image
, microserviceVersion
, and attestation
fields are overwritten with the smallest available upgrade.
Apply the upgrade
Once you updated your config with the desired versions, you can trigger the upgrade with this command:
constellation apply
Microservice upgrades will be finished within a few minutes, depending on the cluster size.
If you are interested, you can monitor pods restarting in the kube-system
namespace with your tool of choice.
Image and Kubernetes upgrades take longer. For each node in your cluster, a new node has to be created and joined. The process usually takes up to ten minutes per node.
When applying an upgrade, the Helm charts for the upgrade as well as backup files of Constellation-managed Custom Resource Definitions, Custom Resources, and Terraform state are created.
You can use the Terraform state backup to restore previous resources in case an upgrade misconfigured or erroneously deleted a resource.
You can use the Custom Resource (Definition) backup files to restore Custom Resources and Definitions manually (e.g., via kubectl apply
) if the automatic migration of those resources fails.
You can use the Helm charts to manually apply upgrades to the Kubernetes resources, should an upgrade fail.
For advanced users: the upgrade consists of several phases that can be individually skipped through the --skip-phases
flag.
The phases are infrastracture
for the cloud resource management through Terraform, helm
for the chart management of the microservices, image
for OS image upgrades, and k8s
for Kubernetes version upgrades.
Check the status
Upgrades are asynchronous operations.
After you run apply
, it will take a while until the upgrade has completed.
To understand if an upgrade is finished, you can run:
constellation status
This command displays the following information:
- The installed services and their versions
- The image and Kubernetes version the cluster is expecting on each node
- How many nodes are up to date
Here's an example output:
Target versions:
Image: v2.6.0
Kubernetes: v1.25.8
Service versions:
Cilium: v1.12.1
cert-manager: v1.10.0
constellation-operators: v2.6.0
constellation-services: v2.6.0
Cluster status: Some node versions are out of date
Image: 23/25
Kubernetes: 25/25
This output indicates that the cluster is running Kubernetes version 1.25.8
, and all nodes have the appropriate binaries installed.
23 out of 25 nodes have already upgraded to the targeted image version of 2.6.0
, while two are still in progress.
Apply further upgrades
After the upgrade is finished, you can run constellation upgrade check
again to see if there are more upgrades available. If so, repeat the process.