Doesn't attach to the shell

[hcnp]

When I run it as described the following happens:

> k browse-pvc -n tcs-s3-backup s3-backup-restic
Browsing PVC s3-backup-restic in namespace tcs-s3-backup
✓ Attached to browse-s3-backup-restic-fjgvj
> k browse-pvc -n tcs-s3-backup s3-backup-restic -- sh
Browsing PVC s3-backup-restic in namespace tcs-s3-backup
chmod: /etc/profile.d/ps1.sh: No such file or directory
Processes: 3
> k browse-pvc -n tcs-s3-backup s3-backup-restic -- sh
Browsing PVC s3-backup-restic in namespace tcs-s3-backup
>

Is it supposed to work like this? I can attach the running pod afterwards, but then I have to use kubectl attach or k9s.

And it doesn't clean up the pod aftwards. It just keeps running.

[clbx]

Can you provide the image you're trying to attach to and some information about your cluster (Distro, Version, OS) and what OS you're running the tool on?

I'm not sure why its showing you the output of the startup script and not attaching to the pod.

[hcnp]

No. Its a private image, but this is the dockerfile:

# https://hub.docker.com/r/rclone/rclone/tags
# https://github.com/rclone/rclone/releases
FROM rclone/rclone:1.70

LABEL org.opencontainers.image.title="TCS S3 Backup" \
      org.opencontainers.image.authors="TCS Team" \
      org.opencontainers.image.url="https://github.com/trifork/tcs-s3-backup" \
      org.opencontainers.image.documentation="https://github.com/trifork/tcs-s3-backup" \
      org.opencontainers.image.source="https://github.com/trifork/tcs-s3-backup" \
      org.opencontainers.image.vendor="Trifork Cloud Stack" \
      org.opencontainers.image.description="Backup of S3 using rclone and restic to a Persistent Volume"

# Prepare the container for backup from S3 by syncing a bucket to local storage
# and the doing a backup with restic to an existing restic repository.
RUN apk add --no-cache \
    restic=0.18.0-r2

WORKDIR /home/rclone

COPY backup.sh .

# container has runAsNonRoot and image has non-numeric user
# https://github.com/kubernetes/kubernetes/pull/56503
# 1009 = rclone
USER 1009

ENTRYPOINT ["./backup.sh"]

... and I'm running:

• Ubuntu 24.04 LTS
• Zsh 5.9

[clbx]

Sorry for my late response, but the image that your workload is using shouldn't matter.

I spun up a Ubuntu 24.04 VM and set my shell to Zsh 5.9, but im not able to reproduce your first issue where it doesn't give you a shell

The second command you run with sh, executes the startup args. That has to do with how i'm constructing the startup arguments in the job, which could probably use some work, but I'm more interested in your first problem. What version of kubernetes is your cluster?

[hcnp]

> k browse-pvc -n tcs-s3-backup -u 1000 s3-backup-restic
Browsing PVC s3-backup-restic in namespace tcs-s3-backup
✓ Attached to browse-s3-backup-restic-fnmb7
> k version
Client Version: v1.34.1
Kustomize Version: v5.7.1
Server Version: v1.31.7+rke2r1
Warning: version difference between client (1.34) and server (1.31) exceeds the supported minor version skew of +/-1

I'm still experiencing the same. The ns is configured with pod-security.kubernetes.io/enforce: restricted. Could that be it?

It's a RKE2 with k8s 1.31

[clbx]

Possibly, can you look to see what’s happening to the pod? Is it even scheduling?

[hcnp]

Yes. It starts up as a job and pod, and I can shell into it, if I'm not too slow.

So it doesn't attach for some reason.

[clbx]

Coming back around to this. Even when I use the Dockerfile you put in a previous comment and replicate the client/server skew you have, I'm still not able to reproduce.

Can you try other shells? Other OS? A different cluster? I want to try where our systems are different.