Hi all I’m running flink cluster in application mo...
# troubleshooting
c
Hi all I’m running flink cluster in application mode in aws EKS cluster, want to expose jmx metrics on job and task managers, can someone please help me with the MBean configurations, I don’t want that native flink metrics exporter.
p
Just out of curiosity, why don't you want to use the native exporter?
c
Hi @Pedro Mázala Actually I have already one jvm metrics exporter configured and exporting metrics at port 9898, I think that should also export the flink metrics but it's not. I think if we are already exporting jmx port shouldn't it be sufficient to export flink metrics too? or do I need to explicitly export flink metrics again?
After enabling the jmx native exporter I am seeing TaskManager memory spikes, upon investigation it results to be because of the JMX exporter. From a heap dump I found this class seems to be instantiating more objects and not garbage collected efficiently. Any suggestion on how to optimise the memory while using jmx native exporter.
java.util.zip.ZipFile$Source
and the RMI thread: