Linux SMF: cinema4d 2025.3.1 conda post-install fails with CloudFront 403 downloading "extra shared libs"

[ansata2001]

Summary

On a Deadline Cloud Linux service-managed fleet, installing Cinema 4D via the managed deadline-cloud conda channel fails in a post-install step with a CloudFront 403, so the environment is never usable and no render can run.

Repro

Job is a hand-authored OpenJD bundle submitted with deadline bundle submit, using the managed conda queue environment (conda-queue-env-enter).

• CondaPackages: cinema4d=2025 gtk3 atk pango gdk-pixbuf cairo libsoup=2.74
• CondaChannels: deadline-cloud conda-forge
• Region: us-east-1, GPU fleet (A10G / g5)

cinema4d-2025.3.1 resolves and downloads, and the conda transaction completes:

cinema4d   deadline-cloud/linux-64::cinema4d-2025.3.1-...
Preparing transaction: ...working... done
Verifying transaction: ...working... done
Executing transaction: ...working... done

Then a post-install step fails:

🌐 Downloading extra shared libs from CloudFront...
curl: (22) The requested URL returned error: 403
gzip: stdin: unexpected end of file
tar: Child returned status 1
tar: Error is not recoverable: exiting now
Process pid <n> exited with code: 2

The queue environment then exits with an error and the session fails before any render step.

Notes / questions

• cinema4d 2025.3.1 appears to be the only Cinema 4D build published for linux-64 on the deadline-cloud channel, so there is no alternate version to fall back to. (cinema4d=2026 and a bare cinema4d=2025 both return PackagesNotFoundError on linux-64; only the specific build resolves.)
• The same bundle logic works on the Windows fleet (C4D installs and launches), so this is Linux-package specific.
• I could not find this "extra shared libs" download documented in the adaptor README or the public conda recipe.

Question: Is a specific setup step / IAM permission / entitlement / region required for this post-install CloudFront download to succeed on a service-managed fleet? Or is the object/URL misconfigured for cinema4d 2025.3.1 on linux-64?

Environment

• Deadline Cloud service-managed fleet, Linux, us-east-1, NVIDIA A10G (g5)
• Managed deadline-cloud conda queue environment (conda-queue-env-enter)
• Submitted via deadline bundle submit

Full session logs and job IDs available privately on request.

[karthikbekalp]

Hmm, this is interesting. I'll take a look into this.
I just submitted a job on Linux for 2025 which was successfully completed. I'll try to reproduce this more as it looks like a troubling problem.

[karthikbekalp]

Job is a hand-authored OpenJD bundle submitted with deadline bundle submit

Is there a requirement or missing feature that led you to use a hand-authored bundle instead of the AWS Deadline Cloud Cinema 4D submitter?

The submitter provides the supported adaptor-based workflow for running Cinema 4D on service-managed fleets and supplies the expected conda configuration. See the Cinema 4D quick start.

Your package list is missing cinema4d-openjd, which is required to manage Cinema 4D on the worker. It also includes several additional GTK-related packages that the submitter does not normally require. Could you explain why those packages and conda-forge were added?

cinema4d 2025.3.1 appears to be the only Cinema 4D build published for linux-64...

Cinema 4D 2026 is supported on Linux and is available only through our managed conda channels. The submitter should automatically select the conda packages for you.

I could not find this "extra shared libs" download documented...

No additional libs are required.

We have not been able to reproduce this failure through the Cinema 4D submitter workflow. Please submit the same scene through the latest submitter.
If that submission also fails, share the submitter-generated job bundle and full session logs privately so that we can reproduce and investigate it. :)

[ansata2001]

Closing this out. We traced the CloudFront 403 to a stale channel left in our own queue environment (not the managed conda flow); removing it let the managed cinema4d package install cleanly. Thanks for the help.