IBM Support

ICP is not spreading the workload on all Master nodes

Troubleshooting


Problem

This technote describes some issues that you may be experiencing on one of the Master servers of your HA ICP cluster. You notice that most of the ICP workload in kube-system namespace is running on these servers, leaving the other two almost free: this simple command: kubectl -n kube-system get pods -o wide | grep MASTER-x | wc -l where x can be 0,1 or 2, returns 26 for MASTER-0, 38 for MASTER-1 and 54 for MASTER-2

Furthermore, you noticed also that some workloads like logging-elk-master or ibmcloud-image-enforcement, are spawning 3 pods all on the same MASTER-2 server.

N.B. For MASTER-x is a format to describe the master node. Node name is IP address in most of ICP installations. Also, node name in k8s concept is not the hostname of node. But in aws or vsphere cloud provider case, it is hostname.
User can run kubectl get nodes to get node name. See documentation

Document Location

Worldwide

[{"Business Unit":{"code":"BU048","label":"IBM Software"},"Product":{"code":"SSBS6K","label":"IBM Cloud Private"},"Component":"Kubernetes","Platform":[{"code":"PF016","label":"Linux"},{"code":"PF043","label":"Red Hat"}],"Version":"All Versions","Edition":"","Line of Business":{"code":"LOB77","label":"Automation Platform"}}]

Log InLog in to view more of this document

This document has the abstract of a technical article that is available to authorized users once you have logged on. Please use Log in button above to access the full document. After log in, if you do not have the right authorization for this document, there will be instructions on what to do next.

Document Information

Modified date:
16 October 2019

UID

ibm10964796

Page Feedback