Upgrading

    Updated Edit on GitHub
    Copy page

    This page describes the steps for upgrading The Things Stack on Azure Kubernetes Service.

    Database Migration

    The Things Stack database schema is managed between minor versions using database migrations.

    Note:

    User are informed of required migrations for each version via the Release Notes. It’s mandatory to run required database migrations for The Things Stack to function properly.

    To migrate the database during an upgrade set the respective Terraform variables to true.

    Service Variable
    IS is-db-migrate
    NOC noc-db-migrate

    Note:

    For example to migrate Identity Server and Network Operations Center databases run the following command.

    $ terraform apply -var='is-db-migrate=true' -var='noc-db-migrate=true'

    Kubernetes version upgrade

    AKS Kubernetes version upgrades have to be performed sequentially for each minor version. For example to upgrade from 1.25 to 1.27 it’s necessary to upgrade to version 1.26 first as an intermediate step. For more information refer to the official documentation.

    Upgrading from Helm chart 1.x.x to 2.x.x

    New in 2.0.0 Workload Identity is no longer wired up automatically from global.blob.provider: azure. Users must now set the annotation and pod label explicitly in values:

    global:
  serviceAccount:
    annotations:
      azure.workload.identity/client-id: "<client-id>"
    podLabels:
      azure.workload.identity/use: "true"

    The pod label has also moved from Deployment selector labels to pod template labels. Existing Azure deployments must delete their Deployments before upgrading. Use --cascade=orphan to keep pods running during the migration:

    RELEASE="<your-release-name>"
NAMESPACE="<your-namespace>"
for component in is as ns gs js dcs gcs console noc pba; do # Drop pba and/or noc from the list if not enabled.
  kubectl delete deployment "${RELEASE}-${component}" -n "${NAMESPACE}" --cascade=orphan
done
helm upgrade "${RELEASE}" . -n "${NAMESPACE}" -f <your-values-file>

    This is required because Kubernetes does not allow changes to spec.selector.matchLabels on existing Deployments. The --cascade=orphan flag ensures pods continue running while the Deployment objects are recreated by Helm.