Kubernetes worker node saturated with pods

[Sifu Tian]

Hi,

Is there a configuration setting where, using the Kubernetes elastic agent plugin, the go server calls for an agent pod, however, the pods only get spun up on one worker node?  This happens to the point where that worker node is 100% CPU and memory used, which makes the pipelines run to a crawl.

I have a 3 node Kubernetes cluster and there are 2 workers nodes sitting idle and the pipelines die due to one k8 worker node having all the pods running.

I there something that I need to configure that allows the pods to be evenly distributed on the other 2 worker nodes? 

[Sifu Tian]

Hi Everyone

This issue is now resolved.  I added CPU and Memory request and limits to the pod.yaml file.  I also added add inter-pod anti-affinity constraints to the pods as well.

I noticed that I didnt have enough CPU and Memory to the pods which was causes some issues.

Thanks,