Latest operator/helm chart release results in failed install due to helm secret > 1MB

[aesimard1]

Describe the bug
Latest release 3.2.0 is resulting in failed to install release: create: failed to create: Secret "sh.helm.release.v1.nginxingress-sample.v1" is invalid: data: Too long: must have at most 1048576 bytes when following the Getting Started documentation for OpenShift via OLM.

This error did not appear in 3.1.0, before this latest release.

To Reproduce
Follow the exact steps specified in Getting Started to install on OpenShift using OLM.

1. Install the operator via OLM
2. Create the certificate secret
3. Apply the SCC resource
4. Deploy the default NginxIngress resource

Expected behavior
Expect the NginxIngress resource to be installed successfully with the default configuration.

Your environment

• Version of the NGINX Ingress Operator - 3.2.0
• Version of the Ingress Controller - 5.1.0
• Version of Kubernetes - v1.31.8
• Kubernetes platform - OpenShift on prem
• Using NGINX or NGINX Plus - NGINX

Additional context
N/A

[grandjean244]

Hit by the exact same issue. Works fine with v. 3.1.0, but fails with v. 3.2.0, even for a simple sample configuration.
Please also consider keeping multiple entries/versions of the operator in the operator catalog to allow for smooth rollback

[achoimet]

@grandjean244 @aesimard1 do you know if the new 3.2.1 release fix it ?

[grandjean244]

@achoimet, same issue for me with version 3.2.1
failed to install release: create: failed to create: Secret "sh.helm.release.v1.nginxingress-sample.v1" is invalid: data: Too long: must have at most 1048576 bytes

[aesimard1]

@achoimet I can confirm as well, the issue is not fixed in 3.2.1.

[jmighion]

Looks like #531 might be the fix

[AlexFenlon]

Hey @aesimard1, thanks for the report,

We can confirm that there is an issue with Kubernetes Schema file sizes which is causing Helm to complain.

The Helm secret file size exceeds it limits. This is fixed by removing unused files in the Kubernetes local schema folder.
#531 solves this. Thanks @jmighion for linking it,

@achoimet @grandjean244 We should have a fix for this soon, it will be v3.2.2