What happened?
Hi Guys,
I’m using pomerium and I’ve installed it with our Gitops engine (flux). I’ve also forked the helm chart to replace the secret by a configmap (I did not know about some sensitive data like set_request_headers) and I’ve added another secret with the others properties (IDP settings, shared secrets, cookie…). I’ve also developed a small operator (before yours existed) whose goal is to generate the pomerium config and routes by using 2 CRs.
Now, if I want to go back to a Kubernetes secret instead of using a Configmap I’m faced to a big chicken-and-egg problem and I will try to explain it. We are using external-secrets which is linked to GCP Secret Manager . If we store all the configuration in GCP and let external-secret reconcile it and use the property useExistingSecret
in your helm charts we can think that is working. But, if we use our operator or maybe your ingress controller, whose goal is to update the pomerium config we will have an infinite update loop as described below:
FluxCD → reconcile/install → Pomerium
ExternalSecret → fetch GCP secret → Create secret with pomerium config (with no routes)
Operator or Pomerium Ingress Controller → Update the pomerium secret.
ExternalSecret → Recreate the secret (because we have some diffs)
…
I’m a little bit stuck with the management of this secret May be you have some idea to help me ?