Hi Guys, happy Friday~ We are testing out Autoscaler (https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-release-1.4/docs/custom-resource/autoscaler/) with a very simple Kafka reading and sinking Job, with the Kafka side data generation rate (1000 records per second) and Flink Job Reading rate at 8 records per second, we would expect Autoscaler should scale up to catch-up with the speed, but nothing happens in our env. Could anyone help to figure out why the backlog-based scaling (https://cwiki.apache.org/confluence/display/FLINK/FLIP-271%3A+Autoscaling) is not working in our case? Thank you! Here is the screenshot of the Job Metric: And Here is our flinkConfiguration:

apiVersion: <http://flink.apache.org/v1beta1|flink.apache.org/v1beta1>

kind: FlinkDeployment

metadata:

name: myauto

spec:

image: <http://ghcr.io/apache/flink-docker:1.17-snapshot-scala_2.12-java11-debian|ghcr.io/apache/flink-docker:1.17-snapshot-scala_2.12-java11-debian>

flinkVersion: v1_17

flinkConfiguration:

kubernetes.operator.job.autoscaler.enabled: "true"

kubernetes.operator.job.autoscaler.scaling.enabled: "true"

kubernetes.operator.job.autoscaler.stabilization.interval: "1m"

kubernetes.operator.job.autoscaler.metrics.window: "1m"

kubernetes.operator.job.autoscaler.vertex.max-parallelism: "16"

kubernetes.operator.job.autoscaler.vertex.min-parallelism: "1"

kubernetes.operator.job.autoscaler.catch-up.duration: "1m"

pipeline.max-parallelism: "24"

taskmanager.numberOfTaskSlots: "8"

execution.checkpointing.interval: "1m"

serviceAccount: flink

It would be best to check the operator logs for clues . All autoscaling metrics and decisions are logged there

Thanks @Gyula Fóra Cool! May I know where to check the operator logs?