-
Notifications
You must be signed in to change notification settings - Fork 663
Description
Hi,
i am doing a scale test in controlled environment on kafka-rest-proxy in EKS setup and i have set heapopts as
- name: KAFKAREST_HEAP_OPTS
value: -Xms512M -Xmx512M -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/var/log/kafka-rest-dump/heapdump-1.hprof
and during scale test we faced java.lang.OutOfMemoryError: Java heap space
error, but when we increase the limits to -Xms1G -Xmx1G, it was fine..
i took a peak at heapdump.hprof file and it seems ingestion sent to particular topic has at least 2 threads holding 150Mb on an average at any time. after some googling i found maybe if i set a limit on jetty thread pool that would let ingestion not fail when heap limit set at 512M.
Q1: i tried using KAFKAREST_JETTY_THREADS_MAX as pod env var, it didnt reflect in pod - is there a configurable variable that can help set jetty threads?
Q2: will setting jetty threads would help here?
Q3: is setting higher limits the only solution? in production we will hit larger limits though so i want to try various options