fix(bench): bypass frontend-proxy, drive frontend directly (PR-N3.3)

[jensholdgaard]

What

Removes frontend-proxy (Envoy) from the OTel Demo capture path. Follow-up to run 26716710757, which failed the readiness gate.

Root cause (and a re-read of the first capture)

Envoy fails its bootstrap proto validation in CI:

Clusters[9] ... SocketAddress.Address: value length must be at least 1 characters

The published frontend-proxy image (demo 2.2.0) carries a baked Envoy config with a 10th cluster whose address env var the demo doesn't set — so Envoy crash-loops. The source tag's envoy.tmpl.yaml has only 9 clusters, all with vars set in .env, and the image is distroless (no shell to inspect the baked template), so the empty var can't be pinned from outside.

That crash-loop reframes the first capture: its ~34% ERR_NAME_NOT_RESOLVED flood wasn't a startup race — it was intermittent DNS for a restarting frontend-proxy. PR-N3.2's readiness probe correctly caught it instead of silently capturing a broken corpus.

Fix — bypass Envoy

Per maintainer call, drive the frontend HTTP API directly instead of debugging a distroless image's baked config:

• Compose override sets, on load-generator:

  • LOCUST_BROWSER_TRAFFIC_ENABLED=false
  • LOCUST_HOST=http://frontend:${FRONTEND_PORT}

  (environment merges by key, overriding the base's hardcoded =true and the .env LOCUST_HOST.)

• The frontend still fans out to cart / checkout / currency / product-catalog / payment / …, so business-service log diversity still flows — the corpus just becomes API-driven rather than browser-driven.

• Readiness now gates on otel-collector + frontend running. Frontend has no healthcheck and only an ephemeral host port, so "running" + the post-restart settle + the existing steady-state slice (which discards early ramp errors) is the readiness story.

• The crash-looping frontend-proxy is stopped so it doesn't burn runner CPU (nothing depends on it — it fronts the frontend, not the reverse).

Verification

• All three YAML files validate; step flow unchanged.
• No Rust changes. Real test is the next dispatch — I'll report whether the business services now dominate the corpus.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Chores
  • Updated OpenTelemetry demo corpus capture configuration to route load-generator traffic directly to the frontend service
  • Disabled browser traffic simulation in load generation
  • Simplified workflow readiness checks and service startup orchestration

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 41d7b528-e207-440e-9cdb-26ac48035099

📥 Commits

Reviewing files that changed from the base of the PR and between e3fa3ad and 8de2f3b.

📒 Files selected for processing (2)

• .github/otel-demo-capture-compose-override.yml
• .github/workflows/capture-otel-demo-corpus.yml

---

📝 Walkthrough

Walkthrough

Updated the OTel demo corpus capture by configuring load-generator to bypass the frontend-proxy/Envoy path and send traffic directly to the frontend service. The Docker Compose override now sets load-generator environment variables, and the workflow startup script no longer waits on Envoy's readiness endpoint; it instead polls service status and manages service startup order directly.

Changes

OTel Demo Corpus Capture - Bypass Envoy Path

Layer / File(s)	Summary
Compose override - load-generator direct routing configuration .github/otel-demo-capture-compose-override.yml	Documentation expanded to explain why load-generator bypasses frontend-proxy/Envoy. Added load-generator service environment overrides: disables LOCUST_BROWSER_TRAFFIC_ENABLED and sets LOCUST_HOST to route traffic directly to the frontend service using ${FRONTEND_PORT}.
Workflow startup orchestration - skip Envoy and reorder services .github/workflows/capture-otel-demo-corpus.yml	Removed ENVOY_ADMIN_PORT from workflow environment. Refactored the readiness script to use docker compose ps polling for otel-collector and frontend running status instead of curl-based Envoy /ready endpoint checks. Script now stops frontend-proxy to prevent crash-looping and restarts load-generator after frontend reaches running state.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• jensholdgaard/ourios#74: Updates the same OTel demo capture files with bind-mounted capture directory coordination for otel-collector output.
• jensholdgaard/ourios#73: Modifies the same capture workflow by changing how compose services are waited on and orchestrated during the capture run.
• jensholdgaard/ourios#75: Adds frontend-proxy readiness gating and restart logic to the workflow—directly conflicts with this PR's Envoy bypass and load-generator routing changes.

🐰 In captures old and new,
We skip the proxy's view,
Load runs to frontend, true and bright,
*No Envoy gates—just traffic's flight! 🚀

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The description provides a comprehensive explanation of the 'What' and 'Why' but is missing most of the template's checklist items (cargo fmt/clippy, tests, docs/CHANGELOG updates, RFC links).	Complete the checklist by confirming or documenting cargo fmt/clippy status, whether tests were added, and CHANGELOG.md updates.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately and specifically describes the main change: bypassing the frontend-proxy and driving the frontend directly, which is the central objective of this PR.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/otel-demo-bypass-envoy

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

This PR updates the OTel Demo corpus capture workflow to bypass the crash-looping frontend-proxy/Envoy path and drive the demo frontend directly, preserving business-service log generation for benchmark corpus capture.

Changes:

• Removes the Envoy admin readiness probe and instead gates on otel-collector and frontend reaching running.
• Stops frontend-proxy after startup to avoid CI runner churn.
• Overrides load-generator settings to disable browser traffic and target frontend directly.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
.github/workflows/capture-otel-demo-corpus.yml	Updates readiness and load-generator restart flow for direct frontend capture.
.github/otel-demo-capture-compose-override.yml	Adds load-generator environment overrides to bypass frontend-proxy.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.