I'm trying to start flink job cluster.
I've cofigured HA to zookeeper and placed job.jar to classpath.
I'm expecting the jobmanager starts with one job waiting to taskmanagers starts.
But when I starts jobmanager via ./flink/bin/standalone-job.sh start-foreground
jobmanager starts executing the job immediately without any taskmanager running. And WebUI says:
{
errors: [
"Service temporarily unavailable due to an ongoing leader election. Please refresh."
]
}
The job is written in Apache Beam and the behavior is similar to start the job manually with java -jar job.jar --runner=DirectRunner
The problem was in missing argument. Flink just run the main method and beam uses DirrectRunner as default. So the job does not run on flink. This command works: ./flink/bin/standalone-job.sh start-foreground --runner=org.apache.beam.runners.flink.FlinkRunner