Data node JVM Heap not stable

Hi

I Have cluster with 3 data nodes + 1 master node.
I noticed that in Idle the JVM heap of data node is not stable and jumping between 16GB to 25GB. Please see below.

Why it is happening?
Is it normal behavior?

jvmHeap

Don't do that. If you only have 3-4 nodes just make them all do everything. Dedicated master nodes are unnecessary for such a small cluster. If you only have one master node then it will eventually fail, destroying your cluster in the process. See these docs for more info.

Not unexpected for sure, see e.g. this blog post.