How can I move Traefik certificates to cert-manager without downtime?

November 19, 2026 3 answers
Asked by d.762+socketbadger

I have a single Traefik instance on a k3s cluster serving about 20 services. Traefik currently gets Let's Encrypt certificates using HTTP-01, and I want to move certificate management to cert-manager before making Traefik highly available.

Could Traefik use TLS-ALPN-01 temporarily while cert-manager uses HTTP-01? Some customer domains only point their A records at the cluster, so I can't rely on controlling all their DNS.

Answers

_c64!plunvexa.7

That split is a reasonable migration path. Keep Traefik serving its existing certificates, switch its resolver to TLS-ALPN-01 for any renewals, and let cert-manager solve HTTP-01 on port 80. cert-manager doesn't have a TLS-ALPN-01 solver, so that part needs to stay on the Traefik side.

For HTTP-01, cert-manager creates a temporary solver Ingress. Make sure Traefik's Kubernetes Ingress provider is enabled and the solver's ingressClassName selects it. Also check that requests to the challenge path reach the solver rather than being intercepted or redirected in a way that breaks validation.

Start with a staging ClusterIssuer and one hostname. Once validation works, issue a production certificate, configure the route to use its Secret, and verify the certificate being served before moving on to the other domains. Test renewal too, not just initial issuance.

_b46~tavmori.8:

Keep the old resolver and certificate storage available until the new Secrets are in use and verified. Switching the challenge type alone doesn't make the final cutover risk-free.

Reply to this answer
Nexqueli.7m

Use DNS-01 for the domains where you control DNS. It avoids sharing the HTTP challenge path with Traefik, and it can issue wildcard certificates too.

For customer domains, HTTP-01 is still useful if the hostname points at your cluster and the challenge request can reach the solver. You don't have to force one challenge method on every domain.

Reply to this answer
q.731~fendrix

I'd separate certificate migration from the Traefik HA rollout. First prove that the current instance can serve cert-manager's Secrets, then change the deployment topology. That gives you a smaller rollback if something goes wrong.

For the HA cutover, check that every replica can read the same Kubernetes Secrets and that traffic can reach healthy replicas throughout the rollout. A certificate being issued successfully doesn't by itself guarantee uninterrupted routing.

4n~parcelgauge_712:

Checking the served certificate through the public endpoint is useful here, since a ready Secret doesn't prove the route is using it.

Reply to this answer

Add your answer

Answers are reviewed before publication.

Related Discussions