Replies: 7 comments 8 replies
-
|
To be honest, it is really hard to understand what the problem you are experiencing really is without the full configurations, full logs from the operator, and all components etc. So sharing that would be a good start. As for your questions:
I'm not sure what exactly you mean with this. The
No. But there were other things that changed how certificates are mounted and used which happened in recent versions.
Not really. One thing to keep in mind is that the certificates you provide should be in the PKCS#8 format. In older versions, PKCS#1 worked fine pretty much by coincidence, but they do not work in the newer versions. Could that be related? The correct ones - PKCS#8 - start with |
Beta Was this translation helpful? Give feedback.
-
|
Thank you @scholzj, for the quicky response. The cluster-ca and client-ca keys are PKCS#8 version. So I don't think this the issue. Here it the logs of the strimzi operator: |
Beta Was this translation helpful? Give feedback.
-
|
Kafka Broker Logs |
Beta Was this translation helpful? Give feedback.
-
|
Application logs |
Beta Was this translation helpful? Give feedback.
-
|
Is there any secure channel to share this kind of information? |
Beta Was this translation helpful? Give feedback.
-
|
Let me try and add some context to this. Only working part-time in proximity to this, but I am well acquainted with the setup. In v0.47.0, the init scripts added both the broker certificate from the broker secret and the cert-chain from the cluster-ca secret to the p12 store. The brokers present the full chain up to the root certificate and since that is what the clients trust, it works. In v0.49.0 the default behavior seems to be to read the contents of the broker secret directly into the Is this the expected behavior? |
Beta Was this translation helpful? Give feedback.
-
|
I have opened the bug report for it #12364. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
After upgrading the Strimzi Kafka Operator from 0.47.0 to 0.49.1, all applications connecting via internal DNS services (e.g., my-kafka-bootstrap.kafka.svc.cluster.local) started failing with TLS SSL_HANDSHAKE errors.
Our Kafka deployment does not use Strimzi‑managed CAs:
We supply our own certificates via Kubernetes Secrets.
All KafkaUsers are already in API version v1 and correctly use .spec.authorization.acls[].operations.
The issue appears only after upgrading the operator — no changes were made to certificates or KafkaUser resources. When we rollback to version 0.47.0 the issue disappears
Expected behavior
TLS clients connecting internally using Kubernetes DNS (e.g.,
my-kafka-bootstrap.kafka.svc.cluster.local:9093) should successfully authenticate.
No certificate regeneration should occur because Strimzi CA generation is disabled.
Internal brokers should continue presenting SANs matching:
Actual behavior
Immediately after the upgrade:
This began exactly when the operator reconciled after upgrading to 0.49.1.
Environment
Cluster Operator version:
Kafka version: 4.0.0
Kubernetes/OpenShift version: (Kubernetes 1.27+ required by Strimzi 0.48+)
CAs: Provided externally, not generated by Strimzi
Listeners:
What we need help from Strimzi maintainers with
Beta Was this translation helpful? Give feedback.
All reactions