Red Hat OpenShift

Camunda 8 can be deployed using Helm on Red Hat OpenShift with proper configurations. The primarily difference from general Helm deployment guide is related to the Security Context Constraints (SCCs) you have in your cluster.

Compatibility

We test against the following OpenShift versions and guarantee compatibility with:

OpenShift version	Supported
4.10.x	✅
4.11.x	✅
4.12.x	✅
4.13.x	✅

Any version not explicitly marked in the table above is not tested, and we cannot guarantee compatibility.

Pitfalls to avoid

For general deployment pitfalls, visit the deployment troubleshooting guide.

Security Context Constraints (SCCs)

Much like how roles control the permissions of users, Security Context Constraints (SCCs) are a way to control the permissions of the applications deployed, both at the pod and container level. It's generally recommended deploying your application with the most restricted SCCs possible. If you're not familiar with security context constraints, refer to the OpenShift documentation.

Permissive SCCs

Out of the box, if you deploy Camunda 8 (and related infrastructure) with a permissive SCCs, there is nothing particular for you to configure. Here, a permissive SCCs refers to one where the strategy for RunAsUser is defined as RunAsAny (including root).

Non-root SCCs

If you wish to deploy Camunda 8 but restrict the applications from running as root (e.g. the nonroot built-in SCCs), you will need to configure each pod and container to run as a non-root user. For example, when deploying Zeebe using a stateful set, you would add the following, replacing 1000 with the user ID you wish to use:

spec:
  template:
    spec:
      securityContext:
        runAsUser: 1000
      containers:
        securityContext:
          runAsUser: 1000

note

As the container user in OpenShift is always part of the root group, it's not necessary to define a fsGroup for any Camunda 8 applications pod security context.

This is necessary for all Camunda 8 applications, as well as related ones (e.g. Keycloak, PostgreSQL, etc.). This is notably crucial for stateful applications which will write to persistent volumes, but it's also generally a good idea to set.

Restrictive SCCs

The following is the most restrictive SCCs you can use to deploy Camunda 8. Note that this is, in OpenShift 4.10, equivalent to the built-in restricted SCCs (which is the default SCCs).

Allow Privileged: false
Default Add Capabilities: <none>
Required Drop Capabilities: KILL,MKNOD,SYS_CHROOT,SETUID,SETGID
Allowed Capabilities: <none>
Allowed Seccomp Profiles: <none>
Allowed Volume Types: configMap,downwardAPI,emptyDir,persistentVolumeClaim,projected,secret
Allow Host Network: false
Allow Host Ports: false
Allow Host PID: false
Allow Host IPC: false
Read Only Root Filesystem: false
Run As User Strategy: MustRunAsRange
SELinux Context Strategy: MustRunAs
FSGroup Strategy: MustRunAs
Supplemental Groups Strategy: RunAsAny

When using this, you must take care not to specify any runAsUser or fsGroup in either the pod or container security context. Instead, let OpenShift assign arbitrary IDs.

note

If you are providing the ID ranges yourself, you can configure the runAsUser and fsGroup values yourself as well.

The Camunda Helm chart can be deployed to OpenShift with a few modifications, primarily revolving around your desired security context constraints. You can find out more about this in the next section.

Deployment

As discussed in the previous section, you need to configure the pod and container security contexts based on your desired security context constraints (SCCs).

The Elasticsearch, Keycloak, and PostgreSQL charts all specify default non-root users for security purposes. To deploy these charts through the Camunda Helm chart, these default values must be removed. Unfortunately, due to a longstanding bug in Helm affecting all Helm versions from 3.2.0 and greater, this makes the installation of the chart (when deploying any of these sub-charts) more complex.

Note that this is only an issue if you are deploying Elasticsearch, Keycloak (via Identity), or PostgreSQL (via Keycloak). If you are not deploying these, or not via the camunda-platform chart, or you are using permissive SCCs, this issue does not affect your deployment.

note

This also affects installations done through the OpenShift console, as it still uses Helm under the hood.

Permissive SCCs

To use permissive SCCs, install the charts as they are. Follow the general Helm deployment guide.

Restrictive SCCs

To use more restrictive SCCs, configure the following minimum set of values for the various applications. The recommendations outlined in the sections are relevant here as well. As the Camunda 8 applications do not define a pod or security context, follow these recommendations, or simply omit defining any.

If you are deploying with SCCs where RunAsUser is MustRunAsRange (e.g. the default restricted SCCs), and are deploying at least one of Elasticsearch, Keycloak, or PostgreSQL, it's necessary to unset the default security context of these charts. If this does not apply to you, you can stop here.

Now this depends on which Helm version you use: 3.1.3 and lower, or 3.2.0 and greater (i.e. one affected by Helm's nested sub-charts). Find out your Helm version by running the following:

helm version --short

v3.8.1+g5cb9af4

Helm 3.1.3 or lower

If you're running on Helm 3.0.0 up to 3.1.3, you need to add these values to your values.yaml file, or save them to a new file locally, e.g. openshift.yaml:

note

These values are also available in the Camunda Helm chart.

# omit this section if elasticsearch.enabled is false
elasticsearch:
  securityContext:
    runAsUser: null
  sysctlInitContainer:
    enabled: false
  podSecurityContext:
    fsGroup: null
    runAsUser: null

# omit this section if identity.enabled is false
identity:
  # omit this section if identity.keycloak.enabled is false
  keycloak:
    containerSecurityContext:
      runAsUser: null
    podSecurityContext:
      fsGroup: null
      runAsUser: null
    postgresql:
      # omit this section if identity.keycloak.postgresql.primary.enabled is false
      primary:
        containerSecurityContext:
          runAsUser: null
        podSecurityContext:
          fsGroup: null
          runAsUser: null
      # omit this section if identity.keycloak.postgresql.readReplicas.enabled is false
      readReplicas:
        containerSecurityContext:
          runAsUser: null
        podSecurityContext:
          fsGroup: null
          runAsUser: null
      # omit this section if identity.keycloak.postgresql.metrics.enabled is false
      metrics:
        containerSecurityContext:
          runAsUser: null
        podSecurityContext:
          fsGroup: null
          runAsUser: null

When installing the chart, run the following:

helm install camunda camunda/camunda-platform --skip-crds --version "$CHART_VERSION" -f values.yaml -f openshift.yaml

Helm 3.2.0 and greater

If you must deploy using Helm 3.2.0 or greater, you have two options. One is to use a SCCs which defines the RunAsUser strategy to be at least RunAsAny. If that's not possible, then you need to make use of a post-renderer.

danger

If using a post-renderer, you must use the post-renderer whenever you are updating your release, not only during the initial installation. If you do not, the default values will be used again, which will prevent some services from starting.

While you can use your preferred post-renderer, we provide one (included in the chart archive) which requires only bash and sed to be available locally:

#!/bin/bash -eu
# Expected usage is as an Helm post renderer.
# Example usage:
#   helm install my-release camunda/camunda-platform --post-renderer ./patch.sh
#
# This script is a Helm chart post-renderer for users on Helm 3.2.0 and greater. It allows removing default
# values set in sub-charts/dependencies, something which should be possible but is currently not working.
# See this issue for more: https://github.com/helm/helm/issues/9136
#
# The result of patching the rendered Helm templates is printed out to STDOUT. Any other logging from the
# script is thus sent to STDERR.
#
# Note to contributors: this post-renderer is used in the integration tests, so make sure that it can be used
# from any working directory.

set -o pipefail

# Perform two passes: once for single quotes, once for double quotes, as it's not specified that string values are
# always output with single or double quotes
sed -e "s/'@@null@@'/null/g" -e 's/"@@null@@"/null/g'

You also need to use a custom values file, where instead of using null as a value to unset default values, you use a special marker value which will be removed by the post-renderer.

Copy these values to your values file or save them as a separate file, e.g. openshift.yaml:

note

These values are also available in the Camunda Helm chart.

# omit this section if elasticsearch.enabled is false
elasticsearch:
  securityContext:
    runAsUser: "@@null@@"
  sysctlInitContainer:
    enabled: false
  podSecurityContext:
    fsGroup: "@@null@@"
    runAsUser: "@@null@@"

# omit this section if identity.enabled is false
identity:
  # omit this section if identity.keycloak.enabled is false
  keycloak:
    containerSecurityContext:
      runAsUser: "@@null@@"
    podSecurityContext:
      fsGroup: "@@null@@"
      runAsUser: "@@null@@"
    postgresql:
      # omit this section if identity.keycloak.postgresql.primary.enabled is false
      primary:
        containerSecurityContext:
          runAsUser: "@@null@@"
        podSecurityContext:
          fsGroup: "@@null@@"
          runAsUser: "@@null@@"
      # omit this section if identity.keycloak.postgresql.readReplicas.enabled is false
      readReplicas:
        containerSecurityContext:
          runAsUser: "@@null@@"
        podSecurityContext:
          fsGroup: "@@null@@"
          runAsUser: "@@null@@"
      # omit this section if identity.keycloak.postgresql.metrics.enabled is false
      metrics:
        containerSecurityContext:
          runAsUser: "@@null@@"
        podSecurityContext:
          fsGroup: "@@null@@"
          runAsUser: "@@null@@"

Now, when installing the chart, you can do so by running the following:

helm install camunda camunda/camunda-platform --skip-crds --version "$CHART_VERSION" \
    -f values.yaml -f openshift.yaml --post-renderer ./patch.sh

Configuring Ingress using routes for Zeebe Gateway

OpenShift version	Supported
4.10.x	limited
4.11.x	✅

The Ingress on OpenShift works slightly different from the Kubernetes default. The mechanism is called routes.

To use these routes for the Zeebe Gateway, configure this through Ingress as well.

Alternatives

An alternative to using routes is to install and use one of the Kubernetes Ingress controllers instead, for example, the ingress-nginx controller.

danger

Do not confuse the ingress-nginx controller with the NGINX Ingress Controller that is endorsed by Red Hat for usage with OpenShift. Despite very similar names, they are two different products.

If you should decide to use the Red Hat endorsed NGINX Ingress Controller, you would require additional adjustments done to the Camunda 8 ingress objects and the NGINX Ingress Controller itself to make gRPC and HTTP/2 connections work. In that case, please refer to the example and the prerequisites.

Prerequisite

As the Zeebe Gateway uses gRPC (which relies on HTTP/2), this has to be enabled.

Required steps

Provide TLS secrets for the Zeebe Gateway, the Cert Manager might be helpful here:

One issued to the Zeebe Gateway Service Name. This must use the pkcs8 syntax as Zeebe only supports this, referenced as Service Certificate Secret or <SERVICE_CERTIFICATE_SECRET_NAME>. For more details, review the OpenShift documentation.
One that is used on the exposed route, referenced as External URL Certificate Secret or <EXTERNAL_URL_CERTIFICATE_SECRET_NAME>.

Configure your Zeebe Gateway Ingress to create a re-encrypt route:

zeebe-gateway:
  ingress:
    annotations:
      route.openshift.io/termination: reencrypt
      route.openshift.io/destination-ca-certificate-secret: <SERVICE_CERTIFICATE_SECRET_NAME> # this is not supported on Openshift 4.10
    className: openshift-default
    tls:
      enabled: true
      secretName: <EXTERNAL_URL_CERTIFICATE_SECRET_NAME>

Mount the Service Certificate Secret to the Zeebe Gateway Pod:

zeebe-gateway:
  env:
    - name: ZEEBE_GATEWAY_SECURITY_ENABLED
      value: "true"
    - name: ZEEBE_GATEWAY_SECURITY_CERTIFICATECHAINPATH
      value: /usr/local/zeebe/config/tls.crt
    - name: ZEEBE_GATEWAY_SECURITY_PRIVATEKEYPATH
      value: /usr/local/zeebe/config/tls.key
  extraVolumeMounts:
    - name: certificate
      mountPath: /usr/local/zeebe/config/tls.crt
      subPath: tls.crt
    - name: key
      mountPath: /usr/local/zeebe/config/tls.key
      subPath: tls.key
  extraVolumes:
    - name: certificate
      secret:
        secretName: <SERVICE_CERTIFICATE_SECRET_NAME>
        items:
          - key: tls.crt
            path: tls.crt
        defaultMode: 420
    - name: key
      secret:
        secretName: <SERVICE_CERTIFICATE_SECRET_NAME>
        items:
          - key: tls.key
            path: tls.key
        defaultMode: 420

Mount the Service Certificate Secret to the Operate and Tasklist pods and configure the secure TLS connection. Here, only the tls.crt file is required.

For Operate:

operate:
  env:
    - name: CAMUNDA_OPERATE_ZEEBE_SECURE
      value: "true"
    - name: CAMUNDA_OPERATE_ZEEBE_CERTIFICATEPATH
      value: /usr/local/operate/config/tls.crt
  extraVolumeMounts:
    - name: certificate
      mountPath: /usr/local/operate/config/tls.crt
      subPath: tls.crt
  extraVolumes:
    - name: certificate
      secret:
        secretName: <SERVICE_CERTIFICATE_SECRET_NAME>
        items:
          - key: tls.crt
            path: tls.crt
        defaultMode: 420

The actual configuration properties can be reviewed in the Operate configuration documentation.

For Tasklist:

tasklist:
  env:
    - name: CAMUNDA_TASKLIST_ZEEBE_SECURE
      value: "true"
    - name: CAMUNDA_TASKLIST_ZEEBE_CERTIFICATEPATH
      value: /usr/local/tasklist/config/tls.crt
  extraVolumeMounts:
    - name: certificate
      mountPath: /usr/local/tasklist/config/tls.crt
      subPath: tls.crt
  extraVolumes:
    - name: certificate
      secret:
        secretName: <SERVICE_CERTIFICATE_SECRET_NAME>
        items:
          - key: tls.crt
            path: tls.crt
        defaultMode: 420

The actual configuration properties can be reviewed in the Tasklist configuration documentation.

Configure all other applications running inside the cluster and connecting to the Zeebe Gateway to also use TLS.

Compatibility​

Pitfalls to avoid​

Security Context Constraints (SCCs)​

Permissive SCCs​

Non-root SCCs​

Restrictive SCCs​

Deployment​

Permissive SCCs​

Restrictive SCCs​

Helm 3.1.3 or lower​

Helm 3.2.0 and greater​

Configuring Ingress using routes for Zeebe Gateway​

Alternatives​

Prerequisite​

Required steps​

Compatibility

Pitfalls to avoid

Security Context Constraints (SCCs)

Permissive SCCs

Non-root SCCs

Restrictive SCCs

Deployment

Permissive SCCs

Restrictive SCCs

Helm 3.1.3 or lower

Helm 3.2.0 and greater

Configuring Ingress using routes for Zeebe Gateway

Alternatives

Prerequisite

Required steps