Hi everyone. We are using the official Flink K8s o...
# troubleshooting
b
Hi everyone. We are using the official Flink K8s operator and have a session cluster with 2 Job Managers for HA. When we submit a FlinkSessionJob with jarURI pointing to remote location, we intermittently see an error in flink session job status:
Copy code
org.apache.flink.runtime.rest.handler.RestHandlerException: Jar file /<jar-dir>/ce239a09-463f-4667-8f81-9cd6a76f5454_my-jar.jar
    does not exist\n\ta ...
Our understanding is this is happening because the jar is uploaded to one of JM's local FileSystem by the flink operator(our web.upload.dir is a k8s emptyDir atm) and if the job submit request goes to other JM, it fails as jar doesn't exist. One way to fix this is to use a Persistent Volume mounted to both JM's jar directory(web.upload.dir). I have a few questions: 1. we are wondering if is there any easier way to fix it? 2. Does the operator downloads the jar and upload it to JM or does the operator instruct the JM to pass jarURI to it downloading the jar?
o
hey πŸ™‚ did you manage to fix that one?
b
Hey, I heard it's fixed in 1.6 version. Btw we cloned the official operator and patched it at that time to make it work