I’m running an Apache Flink cluster on multiple ma...
# troubleshooting
a
I’m running an Apache Flink cluster on multiple machine , one of them has 250 GB of memory and currently using 200 GB JVM heap. Has anyone experienced issues with such a large heap (e.g., long GC pauses or difficulty with fault isolation)? Would it be better to split the resources into multiple TaskManagers (for example, 8 instances with 25 GB each) to improve performance and resilience? I’d love to hear your experiences and any tuning recommendations!
p
Hey there! It all depends on the driver to the question. Usually it's easier to replace small bits than a larger one. In the other hand, smaller pieces may require more serde time. Do you want to achieve something specific in this scenario?