Load balancer server pool is down for Kubernetes Services
search cancel

Load balancer server pool is down for Kubernetes Services

book

Article ID: 345587

calendar_today

Updated On:

Products

VMware

Issue/Introduction

Symptoms:
  • You are not able to communicate with Kubernetes services created in VMware PKS.

  • After upgrading to NCP 2.3.2, LB Server pool status is down for K8S services.

  • You see the K8S service endpoints created and Service IP assigned, however server pool status is still down.



Environment

VMware PKS 1.x

Cause

The issue was due to named targetPorts in a service. You should use the Port numbers for target ports instead of Names.

This is related to named targetPort, which was never supported in NCP. The scenario was working in previous NCP releases, because it was using port name to parse endpoints. If the service port name was found missing, ncp assumed this to be a single port service and added all the endpoints as members. 

protocol: TCP
port: 80
targetPort: http
name: http_port =====> port name

From NCP 2.3.2, we started parsing endpoints with targetPort instead of port name, as targetPort is always present in a service. So ncp is trying to match the targetPort number in endpoint spec and is not able to do so as a string is provided instead of an integer.

Resolution

This is a known issue with VMware NCP plugin 2.3.2, currently there is no resolution.


Workaround:

To work around this issue, use the Port numbers for targetport instead of names.

Ex; use port number 80 instead of http.