I'm having a short general question, regarding automatic load balancing of openshift to avoid overloading particular nodes.
Right now we are having a small system with 5 nodes, which are more or less similar in capabilities.
For some reason openshift always seems to start with node0 and fills it up to it's max pods and than moves on to node1, node2, etc. Instead of trying to balance things out.
This causes the first node to have rather high load of 70+, while the last node is completely idle.
Is there some way to specify, on a node basis, to not schedule further nodes, if the load is over X?
thanks
Today, when scheduling a pod and not specifying limits or request in the resources section of the pod, there is not enough information for the scheduler to balance the workload. If you set up a LimitQuota in the namespace to default limits, you'll see the pods get more evenly spread. This issue in Kubernetes is the tracking issue for improving the scenario where no resources are known and the pods aren't evenly spread.