pixop-test-live

Smoke-test the digest-pinned Pixop LIVE Docker image by running the container entrypoint with --version. Confirms that Docker, GPU passthrough, and the pinned image work together. This is not the same as running the full live pipeline (pixop-live, TSDuck, SRT listeners).

Usage

pixop-test-live

No flags. The script re-executes with sudo when not root.

Command reference

Step	Behavior
Config	Sources `/etc/pixop-live/live.env`
Image	Requires `LIVE_IMAGE=…@sha256:…` present locally
Run	`docker run --rm --gpus all --pull=never --log-driver=none` with `NVIDIA_DRIVER_CAPABILITIES=compute,utility,video`
Command	Container args: `--version` only

What changes / what never changes

Changes	Never changes
Nothing persistent	`live.env`, images, systemd units, SRT bindings

Does not start pixop-live.service, tsduck@*, or pull images from a registry.

Preconditions and scope

Environment	Notes
AWS / KVM guest	AL2023 appliance with NVIDIA container toolkit
GPU	`--gpus all` required; fails similarly to live service if GPU stack is broken
Any qualified appliance	Script shape matches the installed recipe; always verify output against `pixop-version` and your acceptance criteria

Contrast with pixop-nvencc, which runs NVEncC inside the same image, and with systemctl status pixop-live / pixop-stack for the actual pipeline.

Typical workflows

After instance boot or driver update:

pixop-test-live

Narrow GPU vs image issues:

pixop-test-live
pixop-nvencc --check-hw

When live fails but you suspect configuration—not Docker/GPU:

pixop-test-live
pixop-stack status
tsduckctl check main

(Use pixop-healthcheck on builds that ship it for broader readiness; not identical to this smoke test.)

Output interpretation

On success:

Testing image: <repo>@sha256:<digest>
<version string from Pixop binary>
OK: image runs and responds.

The version line is whatever the Pixop binary prints for --version inside LIVE_IMAGE—not the appliance AMI label, not Marketplace listing version, and not pixop-version packaging metadata unless they happen to match.

Exit 2 with ERROR: when:

LIVE_IMAGE missing or not digest-pinned
Image not present locally (This appliance does not pull images at runtime.)

Other failures (Docker daemon, GPU) surface as Docker/NVIDIA messages on stderr.

Security

Read-only regarding customer content; no network listeners opened.
Safe to run on a live appliance; does not stop or restart production services.

Common errors

Symptom	Next step
Image not present	Load/sync image per your deployment process; check `live.env`
GPU errors	GPU requirements (FAQ), `nvidia-smi`, driver toolkit
Permission errors	Run without manual sudo—the wrapper elevates once

pixop-nvencc — NVEncC inside the same image
pixop-stack — live stack orchestration
pixop-version — appliance and image version reporting
GPU requirements (FAQ)