[Graphite MQ] Draft PR GROUP:spec_c48c80 (PRs 4181)

[graphite-app]

This draft PR was created by the Graphite merge queue.
Trunk will be fast forwarded to the HEAD of this PR when CI passes, and the original PRs will be closed.

The following PRs are included in this draft PR:

• #4181

[railway-app]

This PR was not deployed automatically as @graphite-app[bot] does not have access to the Railway project.

In order to get automatic PR deploys, please add @graphite-app[bot] to your workspace on Railway.

[claude]

PR Review

This is a Graphite merge queue draft for PR #4181 ("fix: docker cache issues, dashboards"). The changes are small and focused. Here is my review:

---

engine/packages/gasoline/src/db/kv/mod.rs — Divide-by-zero fix (line 1177)

Good fix. The change from active_worker_ids.len() as u64 to active_worker_ids.len().max(1) as u64 correctly guards against a panic when active_worker_ids is empty and the result is used as a modulus denominator (lines 1254, 1259).

One observation: when the list is empty the worker-idx comparison at line 1162 will have already exited early, so the .max(1) is a defensive safeguard rather than a semantic choice. Consider adding a short comment explaining why .max(1) is used, since a future reader may wonder whether the 1 has semantic meaning or is just a guard:

// Use max(1) to guard against an empty worker list; the modulus result
// is unused in that case because wf_worker_idx will never match this worker.
let active_worker_count = active_worker_ids.len().max(1) as u64;

---

engine/packages/util/build.rs — Incremental build fix (line 12)

Correct fix. Without cargo:rerun-if-env-changed=OVERRIDE_GIT_SHA, incremental builds would silently ignore a changed OVERRIDE_GIT_SHA, baking in stale git SHAs into binaries. The placement (before the if let Ok check) is correct per Cargo's build script protocol.

---

engine/docker/universal/Dockerfile — Debug echo

Minor concern. The echo "Copying binary" line appears to be a temporary debug statement left in. If there's intent to keep it as a permanent progress indicator, it should be a more informative message (e.g. echo "Copying rivet-engine binary to /app/dist/"). If it was just for debugging, it should be removed.

Also noted (pre-existing): the LABEL org.opencontainers.image.source in the engine-full stage still references https://github.com/rivet-gg/rivet. Per project conventions the correct repo URL is https://github.com/rivet-dev/rivet.

---

Grafana dashboards

Dashboard changes look reasonable overall:

• Extending the default time window from 5 m to 30 m is a sensible operational default.
• Replacing pull-duration panels with dispatch-rate panels (rivet_gasoline_workflow_dispatched_total) gives better throughput visibility.
• The pegboard legend fix ({{workflow_name}} → {{rivet_datacenter}}) and the query join on rivet_gasoline_worker_last_metrics_publish to filter stale workers are both correct improvements.
• Verify: The pegboard dashboard default project variable was changed from prod to staging. If this dashboard is deployed to production Grafana instances, this will silently show staging data by default.

---

Summary

File	Verdict
gasoline/src/db/kv/mod.rs	✅ Good bug fix; consider a clarifying comment
util/build.rs	✅ Correct incremental build fix
docker/universal/Dockerfile	⚠️ Remove or improve the debug echo; stale rivet-gg label (pre-existing)
grafana-dashboards/gasoline.json	✅ Looks good
grafana-dashboards/pegboard.json	⚠️ Verify the staging default is intentional for production deployments

The Rust fixes are solid. The main items to address before merging are the leftover debug echo in the Dockerfile and confirming the pegboard dashboard default project.