1-5_upgrade.rst

Open Zaak 1.5 upgrade

Open Zaak 1.5.0+ corrected an oversight where the container was running as root. This is no longer the case, the image from 1.5.0 and newer drops to an unprivileged user with User ID 1000 and Group ID 1000.

Warning

The Open Zaak 1.5 update has an impact on existing installations!

If you are using the Documenten API with the default Open Zaak storage (so, not the CMIS adapter), then the directories in the storage are owned by the root user (or the user that the container root user maps to, in the case of podman for example).

After dropping privileges in the new version, this means that Open Zaak and nginx no longer have (write) access to these directories and files.

We have updated the deployment tooling to correct this where possible, but it's impossible to cover every case.

The instructions on what to check on how to handle this are provided below per supported environment.

We strongly advise to backup your data and test this upgrade on a test/staging environment before rolling it out in production!

Single server with Ansible collection

.. tabs::

  .. group-tab:: Docker

    If you are deploying with the ansible playbooks, then you must:

    * Ensure you are using version ``0.17.0`` or higher of the collection. We have
      updated the requirements to reflect this.
    * Specify the role variable ``openzaak_1_5_upgrade: true`` - this will fix the
      permissions of existing uploads. You can revert/remove this variable again
      after the ugprade has been deployed.

  .. group-tab:: Podman

     TODO - unclear what needs to be done since we don't have access to a podman
     testing environment.

Kubernetes

.. tabs::

  .. group-tab:: Ansible

     We have added an init container to the Ansible and Helm based deployments, which
     is enabled by default. This init container should correct the incorrect file
     system permissions, **provided that the pod is allowed to run containers as
     root**. It changes the owner and groupo of the ``/app/private-media`` directory
     to ``1000:1000``.

     The updated deployment tooling also includes a ``podSecurityContext`` which
     now specifies the ``fsGroup: 1000``. If your environment is different, you may
     have to specify ``openzaak_init_containers`` accordingly.

     It's possible that the PV provisioner causes problems, and it that case, please
     consult with your infrastructure provider on how to ensure the PV is writable
     by UID 1000 and/or GID 1000.

     Note that this init container slows down the application startup for every
     subsequent deployment, so after the migration you may want to disable it by
     setting the variable to no init containers:

     .. code-block:: yaml

         openzaak_init_containers: []

  .. group-tab:: Helm

     We have added an init container to the Helm charts by default to fix the file
     system permissions. For this you need to:

     * update the Open Zaak chart version to at least 0.5.0
     * have ``persistence`` enabled
     * check if there are any :ref:`installation_reference_1_5_upgrade_cloud_provider_notes`

     Once the upgrade is performed, you can skip this init container by setting the
     ``initContainers.volumePerms=false`` value.

     Note that the Open Notificaties update also requires a Helm chart version update
     to ``0.5.0``.

Cloud provider specific notes

Not all cloud providers are the same. The three big ones are arguably Azure, AWS and Google Cloud. Where applicable, we have provider-specific notes.

.. tabs::

    .. group-tab:: Azure

        **Storage classes**

        Persistent volumes on Azure are tricky. Out of the box only the
        ``kubernetes.io/azure-file`` provisioner works with ``ReadWriteMany`` mount
        mode, which Open Zaak requires.

        However, this filesystem gets mounted as ``root`` by default and it's not possible
        to correct the file permissions via an init container or the
        ``securityContext.fsGroup`` option. You must use a storage class with the
        correct mount options, for example:

        .. code-block:: yaml

            kind: StorageClass
            apiVersion: storage.k8s.io/v1
            allowVolumeExpansion: true
            reclaimPolicy: Delete
            volumeBindingMode: Immediate
            metadata:
              name: azurefile-openzaak
            provisioner: kubernetes.io/azure-file
            parameters:
              skuName: Standard_LRS
            mountOptions:
            - uid=1000
            - gid=1000

        Note the explicit ``uid`` and ``gid`` mount options which map to the user that
        Open Zaak runs as. For more information, see also
        `this related Kubernetes issue <https://github.com/kubernetes/kubernetes/issues/54610>`_.

        In our own testing, upgrading worked out of the box because the mounted volume
        results in ``777`` file permissions mode, while still being owned by the root
        user, which is functional but may not be what you want.

        .. note::

            On an existing installation you will probably have an existing PVC with
            incorrect mount options and changing the storage class after creation is
            not possible.

            We recommend backing up the uploaded files, deleting the PVC, modifying the
            storage class that Open Zaak uses and the restoring the backed up data on
            the new PVC.

    .. group-tab:: AWS

        No known challenges at the moment.

    .. group-tab:: Google Cloud

        No known challenges at the moment.