Pods hang on AMD processor architecture

Learn how to resolve the issue of pods not starting in deployments that run on AMD processors.

Problem

During the startup of a IBM® Netcool® Operations Insight® on Red Hat® OpenShift® deployment it was found that the waitforaggp init container hangs, due to a problem in the container where the ObjectServer is located.

On startup of the ObjectServer containers ncoprimary and ncobackup, the docker-entrypoint.sh script hangs at the nco_keygen command and does not proceed further. This issue is due to a bug in the GSKit version provided in the ObjectServer containers.

The ObjectServer logs contain the following error message.
Error: E-UNK-000-000: srv_negotiate(CS_GET,32785) failed

Error: E-UNK-000-000: Invalid challenge response type

Error: E-OBX-102-057: User secure-login@noi-test-ea-noi-layer-easetupo failed to login: General failure

This problem is specific to IBM Tivoli® Netcool/OMNIbus code that runs on hardware that contains an AMD processor.

Resolution

The ICC_SHIFT=3 environment variable changes the behavior of GSKit, which bypasses the problem.

As a workaround, add an environment variable called ICC_SHIFT with a value of 3. Add the variable to the following pods in the environment where the nco_keygen command is run:
  • ncoprimary
  • ncobackup
  • ea-noi-layer-eanoigateway
  • topology-noi-probe
  • topology-noi-gateway