Upgrading KubeArchive

This document details the process to upgrade your KubeArchive installation.

  • KubeArchive is not backward compatible with database changes. Both application and schema must be upgraded at the same time.

  • Backup your database before doing a KubeArchive upgrade.

  1. Scale down KubeArchive:

    kubectl scale -n kubearchive deployment -l app.kubernetes.io/part-of=kubearchive --replicas 0

  2. Download the KubeArchive repository and switch to the version you would like to upgrade to.

    git clone https://github.com/kubearchive/kubearchive.git --branch=v1.1.0 kubearchive-v1.1.0

  3. Run the database migrations:

    cd kubearchive-v1.1.0
migrate -verbose \
    -path integrations/database/postgresql/migrations/ \
    -database postgresql://<kubearchive-user>:<kubearchive-password>@<postgresql-host>:<postgresql-port>/<kubearchive-database> \
    up

  4. Apply the new KubeArchive manifests and scale up KubeArchive:

    kubectl apply -n kubearchive \
    -f https://github.com/kubearchive/kubearchive/releases/download/v1.1.0/kubearchive.yaml
kubectl scale -n kubearchive deployment -l app.kubernetes.io/part-of=kubearchive --replicas 1

  5. Check pods/deployments are running correctly:

    kubectl get -n kubearchive all
kubectl logs -n kubearchive deployment/kubearchive-sink
kubectl logs -n kubearchive deployment/kubearchive-api-server
kubectl logs -n kubearchive deployment/kubearchive-operator

Handling Upgrade Or Downgrade Failure

When a migration fails, the version of the database changes but it is left in a state with data and/or schema inconsistencies. That stops migrate from applying more migrations:

error: Dirty database version <VERSION>. Fix and force version.

When this happens inspect the state of the database to decide how to proceed. When you fix the problem, force the migration to the desired version with the following command (continuing the PostgreSQL example):

migrate -verbose \
    -path integrations/database/postgresql/migrations/ \
    -database postgresql://<kubearchive-user>:<kubearchive-password>@<postgresql-host>:<postgresql-port>/<kubearchive-database> \
    force <VERSION>

The schema version for migrate force must be: * <VERSION -1> when you rolled back the changes of migrate and you overrode the migration file. * <VERSION> when you manually completed the failed migrate execution. Then run migrate up again to make sure you apply the whole migration stack.