Upgrade Oathkeeper helm chart 0.41 causes 503

[WoodyWoodsta]

Preflight checklist

• I could not find a solution in the existing issues, docs, nor discussions.
• I agree to follow this project's Code of Conduct.
• I have read and am following this repository's Contribution Guidelines.
• I have joined the Ory Community Slack.
• I am signed up to the Ory Security Patch Newsletter.

Ory Network Project

N/A

Describe the bug

Upgrading the Oathkeeper chart to 0.41.0 causes oathkeeper to restart after 503 health check. I've not changed anything in the chart, and I don't use the secret (so it's secret.false for me).

Reproducing the bug

Upgrade to chart 0.41.0, with secret.false

Relevant log output

No response

Relevant configuration

oathkeeper:
  oathkeeper:
    config:
      log:
        level: debug
      authenticators:
        noop:
          enabled: true
        cookie_session:
          enabled: true
          config:
            check_session_url: http://ory-kratos-public:80/sessions/whoami
            preserve_path: true
            subject_from: "identity.id"
            only:
              - ory_kratos_session
        oauth2_client_credentials:
          enabled: true
          config:
            token_url: http://ory-hydra-public.ory.svc.cluster.local:4444/oauth2/token
            cache:
              enabled: true
      errors:
        handlers:
          redirect:
            enabled: true
            config:
              to: ***/login
          json:
            enabled: true
          www_authenticate:
            enabled: true
            when:
              - error:
                  - unauthorized
        fallback:
          - json
      authorizers:
        allow:
          enabled: true
        remote_json:
          enabled: true
          config:
            remote: http://ory-keto-read.ory.svc.cluster.local:80/relation-tuples/check
            payload: ""
      mutators:
        noop:
          enabled: true
      serve:
        proxy:
          trust_forwarded_headers: true
          timeout:
            write: 1m
            read: 1m
            idle: 1m
          cors:
            enabled: true
            allowed_origins:
              - ***
    managedAccessRules: false
  maester:
    enabled: false
  serviceMonitor:
    enabled: false
  deployment:
    resources:
      requests:
        cpu: 10m
        memory: 100Mi
      limits:
        memory: 1Gi

Version

0.41.0

On which operating system are you observing this issue?

None

In which environment are you deploying?

Kubernetes with Helm

Additional Context

No response

[Demonsthere]

Hi there! Can you share/take a look at the logs of the container? We could see some more info there. The upgrade itself passed, but for tests we use a very simplistic set of configs so we could easily miss some corner case 😞

[WoodyWoodsta]

Hey :)

I'd have to re-upgrade my running cluster to get the logs, but I didn't collect or post any because there was nothing related to the 503, even with log level on debug. If I get a chance I'll find a time appropriate to break my environment to get any log output.

[Demonsthere]

😅 Yeah, obviously it's better not to break you env, but a failing health check should be logged as an event on the deployment/pod object, did you maybe take a look at that? Asking because right now I can't really do more then try to reproduce that upgrade with a minimal config as close to yours as possible 😞

[WoodyWoodsta]

I've done a search through our monitoring history to see if I could get the kubernetes events around the time I attempted the upgrade, but with no luck. I'll see if I can reproduce and report back with the k8s events.

[sebboer]

i encounter the same error after updating to 0.41. It uses the same values.yaml as before the update.

{
  "audience": "application",
  "error": {
    "message": "The requested resource could not be found",
    "stack_trace": "stack trace could not be recovered from error type *healthx.swaggerNotReadyStatus"
  },
  "http_request": {
    "headers": {
      "accept": "*/*",
      "connection": "close",
      "user-agent": "kube-probe/1.29"
    },
    "host": "127.0.0.1",
    "method": "GET",
    "path": "/health/ready",
    "query": null,
    "remote": "<cluster-ip>:57138",
    "scheme": "http"
  },
  "http_response": {
    "status_code": 503
  },
  "level": "error",
  "msg": "An error occurred while handling a request",
  "service_name": "ORY Oathkeeper",
  "service_version": "v0.40.7",
  "time": "2024-04-29T12:28:50.400938908Z"
}

[sebboer]

Same error for 0.40.1. Update to 0.39.1 resolved into no error.
Coming from 0.38.0

[siredmar]

Same error here. I tried using the 0.42.0 helm chart that uses oathkeeper v0.40.7 by default. I got rid of the 503 errors by using the oathkeeper v0.39.4 with the 0.42.0 helm chart.

[Demonsthere]

If this is a bug in Oathkeeper itself and not the chart or upgrade process, then maybe we should move this to the oathkeeper repo, as this could be regressions in the code itself, similar to #1161