Address family not supported by protocol error when deploying on K8S

[fkooistra]

Bug Overview

When I deploy the latest nginx-s3-gateway image (tag=unprivileged-oss-20260202 )in our single-stack (IPv4) K8S environment, the nginx-pods exit with error: 2026/02/02 09:35:10 [emerg] 1#1: socket() [::]:8080 failed (97: Address family not supported by protocol) which seems to be related with the new default dualstack setting in the nginx default.conf-file.

Expected Behavior

Expected bevavior is the pod starting as exepected, using IPv4

Steps to Reproduce the Bug

Please see bug overview. Deploying the latest image on a K8S environment which only allows IPv4 for pods will produce this result.

Environment Details

• Version of the S3 container used: [DockerHub/GCR] ghcr.io/nginx/nginx-s3-gateway/nginx-oss-s3-gateway:unprivileged-oss-20251201
• Commit/Tag (if building the NGINX S3 gateway from source): n/a
• Version of NGINX Open Source or NGINX Plus: [OSS/Plus]: OSS
• Version of NGINX JavaScript: [0.8.8/0.8.7/etc...]: 0.9.4
• Target deployment platforms [e.g. AWS/GCP/local cluster/etc...]: local cluster
• S3 backend implementation: [e.g. AWS, Ceph, NetApp StorageGrid, etc...]: local, MinIO
• Authentication method: [e.g. IAM, IAM with Fargate, IAM with K8S, AWS Credentials, etc...]: K8S

Additional Context

Currently the image defaults to IPv6 while we do not have dualstack enabled for our pods. It would be nice to have a variable to enable/disable IPv6 support instead of checking the kernel.

[fkooistra]

Note that issue was not present in the 20250101 build, which makes sense as IPv6 as default seems to be introduced in later builds.