Hi Ramya,
The attached image cannot be displayed.
It does not sound normal to me that your job manager with only one job of 5
parallelisms needs 3GB. Have you enabled many per-record metrics? And have
you tried with the latest Flink versions (1.11 or 1.10)?
Thank you~
Xintong Song
On Fri, Jul 10, 2020 at 2:32 PM Ramya Ramamurthy <
[hidden email]> wrote:
> Hi,
>
> I have Flink 1.7 running on our production. I can see that the Memory used
> by the Job Managers are pretty high. Below is a snapshot of our pods
> running the JM.
> I had to commit 3GB memory, with a memory limit of 5GB. I have one cluster
> per job, to make it scalable based on our traffic rate and the benchmark
> per pod. So this cluster holds a job with 5 parallelism.
> I just wanted to know how to understand if this memory consumption is
> acceptable. I was assuming JM's might not need this much memory. Please
> advise.
>
> [image: image.png]
>
Locked
