Hi Gunnar. Answer 1.
I am not sure which image we should be talking about – operator or Flink sessions cluster, since I do not know who is interpreting the job instructions. Anyway, we are using an image we derive from Flink official image: flink:1.20.1-java17 Answer 2. We are using 1.12, which was an upgrade of 1.10. Maybe the upgrade didn’t clean everything up properly. I can try to do a clean install, since the JIRA ticket you mention is also stating something along those lines: setting default-opts. Maybe an old configuration was passed over to the new operator. Thanks for some hints and reassurances. Nikola. From: Gunnar Morling <gunnar.morl...@googlemail.com> Date: Wednesday, June 11, 2025 at 2:19 PM To: Nikola Milutinovic <n.milutino...@levi9.com> Subject: Re: Problems running Flink job via K8s operator Hey Nikola, > Problem 1: The first problem is that the operator doesn’t know about “local” > URI schema used in jarURI. "local" should work, e.g. see here for an example: https://github.com/apache/flink-kubernetes-operator/blob/main/examples/basic.yaml#L38. What's the container image you are using (the CR you shared doesn't tell)? It should be derived from the default Flink image. > Problem 2: With jarURI dropped I get further, but now Java 9+ modules bite me. Which version of the operator are you using? This should work OOTB as of 1.11 or newer (see https://issues.apache.org/jira/browse/FLINK-36646). I ran into this myself on an earlier version, discussing it here (see the repo linked from that post for a workaround, based on that 1.11 fix): https://www.decodable.co/blog/get-running-with-apache-flink-on-kubernetes-1 Hth, --Gunnar On Wed, 11 Jun 2025 at 09:46, Nikola Milutinovic <n.milutino...@levi9.com<mailto:n.milutino...@levi9.com>> wrote: Hello. I have problems trying to run a Flink session job using Flink Kubernetes operator. Two problems, so far. This is the Spec I am trying to use: apiVersion: flink.apache.org/v1beta1<http://flink.apache.org/v1beta1> kind: FlinkSessionJob metadata: name: nix-test spec: deploymentName: flink-cluster-session-cluster job: jarURI: "local:///opt/flink/opt/flink-python-1.20.1.jar" entryClass: "org.apache.flink.client.python.PythonDriver" args: [ "--python", "/opt/flink/workflows/CACHE/cache_valkey_updater.py", "--cdc_kafka_topic", "cdc-inspection-type", "--entity_type", "inspection-type", "--field_names", "Name_Enum" ] parallelism: 1 state: running upgradeMode: savepoint Problem 1: The first problem is that the operator doesn’t know about “local” URI schema used in jarURI. Could not find a file system implementation for scheme ''local''. The scheme is not directly supported by Flink and no Hadoop file system to support this scheme could be loaded. Is there something I should turn on in the config for this to be recognized? In any case, I have read that for PyFlink jobs, this is just a placeholder and can be dropped. So I did. Problem 2: With jarURI dropped I get further, but now Java 9+ modules bite me. Caused by: java.lang.reflect.InaccessibleObjectException: Unable to make field private final java.util.Map java.util.Collections$UnmodifiableMap.m accessible: module java.base does not \"opens java.util\" to unnamed module @4ba2ca36 at java.base/java.lang.reflect.AccessibleObject.checkCanSetAccessible(Unknown Source) at java.base/java.lang.reflect.AccessibleObject.checkCanSetAccessible(Unknown Source) at java.base/java I know that we must put a lot of those –add-opens to Flink config and I have them all. But this looks like it should be added to the Operator itself. Any advice here? Nikola.
