Skip to content

PingFederate Configuration Management

PingFederate has a variety of operating patterns. These patterns typically involve a trade-off between ease of implementation and mitigation of deployment risks.

To simplify the moving parts we will categorize PingFederate Configuration into three categories:

Infrastructure Config

  • Examples: Resource allocation (CPU/Memory/Storage), Client Ingress (Access and Hostnames), Image Version, Exposed Ports, Environment Variable Definition, Secrets Definition
  • Orchestration - These items are defined in the release's values.yaml and any change triggers an update.

Server Config

  • Examples: *.properties files, Integration Kits, HTML Templates, logs formatting (log4j2.xml). This can be oversimplified to everything besides the /instance/server/default/data folder or /instance/bulk-config/data.json.
  • Orchestration - These items are stored in the Server Profile and any change should trigger an update. It is up to the implementor to ensure that happens. This can be done by adding a non-functional variable in values.yaml to track the current profile "version". Example: `SERVER_PROFILE_VERSION: v1.1"

App Config

  • Examples - Core PingFederate configuration. Changes that are typicaly made through the UI or Admin APIs. This can be oversimplified to the /instance/server/default/data folder or /instance/bulk-config/data.json.
  • Orchestration - Dependent of your operating pattern this changes may be delivered via a rolling update, or by configuration replication.

PingFederate Data Mount

In the most common pattern, we attach a persistent volume to /opt/out/instance/server/default/data on the PingFederate Admin Console only.

This pattern is intended to be used when PingFederate Administrators need to deliver configuration through the UI in each environment, including production. Another reason for this may be if SP connections are allowed to be created by app developers via Admin API. In both of these scenarios, the defining factor is that there are mutations in the production Admin console that are not being tracked in any other way, like source control, and therefore must be persisted.

Attributes of this pattern:

  • App Config is persisted in each SDLC environment (e.g. Dev, QA, Prod).
  • App Config promotion is done manually or via Admin API.
  • App Config is replicated from Admin Console to Engines.
  • Server Config is maintained and delivered via server profile.
  • Server profile does not include App Config.
  • Backups are taken regularly in case of Persistent Volume loss or corruption.

Data Mount Example

Helm values relevant to this configuration may look like:

  enabled: true
    replicaCount: 1
    SERVER_PROFILE_PATH: pf-data-mount/pingfederate
    type: StatefulSet
        enabled: true
            ## NOTE THIS PVC DEFINITION ##
            mountPath: /opt/out/instance/server/default/data
              - ReadWriteOnce
                  storage: 8Gi

  enabled: true
    SERVER_PROFILE_PATH: pf-data-mount/pingfederate
    replicaCount: 3
    type: Deployment
        type: RollingUpdate
          maxSurge: 1
          maxUnavailable: 0

The key aspect here is pingfederate-admin.workload.statefulset.persistentvoume.volumes.out-dir.mountPath=/opt/out/instance/server/default/data. This is where all UI configuration (App Config) is stored as files. Because this is the mountPath, PingFederate admins have the freedom to deliver any files not used in /opt/out/instance/server/default/data via Server Profile.

For example, adding a new IDP adapter requires a restart of the service in order for the adapter to be identified and available to App Config. The steps in this case would be:

  1. Add the adapter at
  2. update SERVER_PROFILE_VERSION: v1.1 -> SERVER_PROFILE_VERSION: v1.2 on both the admin and engine
  3. run helm upgrade --install myping pingidentity/ping-devops -f /path/to/values.yaml

If the release already existed, the variable change signifies that the definition has mutated, and thus must be re-deployed. The admin pod will be deleted and recreated, while the engines will surge and roll one by one.

Reference links:

Data Mount Pros and Cons

Values with this approach

  • Managing App Config is more familiar to PingFederate admins with traditional experience.
  • Less to account for when building a CI/CD pipeline because there is no config export + templating.
  • Ability to have configurations different in each environment

Cautions with this approach

  • More room for user configuration error and outage because configurations are not promoted with automated testing.