Hi All, I have deployed apache flink on EKS cluste...
# troubleshooting
c
Hi All, I have deployed apache flink on EKS cluster as a stand alone java process where flink jobmanager, task manager and actual job executed under same java process, now suddenly I observe there is a huge memory spike from the pod itself, and when I analyse the heap dump, it looks like very normal except below class I don't see any other objects consuming more space on heap, even the below class object occupy only 7MB, however the entire EKS pod memory usage is at ~ 2.5 GiB, now I am still not able to figure out how and where the memory is getting consumed more. 13 instances of java.util.zip.ZipFile$Source, loaded by *<system class loader>*occupy 74,10,944 (17.10%) bytes. I could not get much details from the java process heap dump, can some one suggest me how to figure out the memory leak from the flink job running on a minicluster from EKS.
d
How do you analyse heap dump?
n
What sink are you using?
c
@Davide Reich I am analysing the Java process heap dump.
@Nicholas Erasmus I am using Flink-Kafka connector to sink the events back to kafka.