Resolving Image Pull Failures with OAuth Token

Author: Sravya Bandari

Date: November 2024

Audience: Everyone

Environmental details: Self-hosted (KOTS) v9.6.x

Summary

Deployment may experience Ingress image pull failures due to issues with the image registry or URL rewriting by KOTS. Below are solutions to resolve this issue.

Solution

This issue is often resolved after a retry, as requests may get routed to a functional server, or the registry issue may be fixed. However, if KOTS rewrote the image URLs, retries may continue to fail.

To address this:

Revert KOTS image URL changes and retry deployment:

• Make a slight change in the KOTS app (e.g., update a non-critical setting).
• Save the configuration.
• This will revert the image URLs to their original state.
• Deploy the latest sequence again.

Pull the image manually and retry deployment:

• Use the following command to test the image pull directly from the host:
  

  ctr image pull registry.k8s.io/ingress-nginx/kube-webhook-certgen:v20231226-1a7112e06

• If the pull succeeds, make a minor, no-operation configuration change in the KOTS App.
• Redeploy the latest sequence to apply the changes.

Cause

This issue is caused by:

1. Problems with the image registry (e.g., registry.k8s.io).
2. URL rewriting by KOTS.

Related Articles, Work orders, Zendesk tickets, Defects, etc.

• GitHub issue #395
• Replicated community article: Resolving unexpected image rewrites
• https://jamasoftware.zendesk.com/agent/tickets/81305

Please feel free to leave feedback in the comments below.