Skip to content

Add a bounded network diagnostics recipe for ephemeral runners - #26

Open
garysassano wants to merge 1 commit into
runs-on:mainfrom
garysassano:docs/network-diagnostics
Open

Add a bounded network diagnostics recipe for ephemeral runners#26
garysassano wants to merge 1 commit into
runs-on:mainfrom
garysassano:docs/network-diagnostics

Conversation

@garysassano

@garysassano garysassano commented Sep 6, 2026

Copy link
Copy Markdown

An ephemeral runner can disappear before a transient package-download slowdown is investigated. Aggregate network-byte metrics and a later roc logs --full archive cannot reconstruct destination timings or TCP counters that were never captured. This adds a Linux workflow recipe and a standalone standard-library Python probe to capture that evidence during the job.

Closes runs-on/runs-on#542

Changes

  • Read local counters only when no URLs are supplied.
  • Support optional public HTTPS transfers with time, size, and redirect limits, emitting an allowlist of curl timing fields.
  • Exclude response bodies, full URLs, environment dumps, curl stderr, and credential-bearing URLs from the output.
  • Explain host/container observations, cache API failures, remote-service delays, and evidence that still needs a live authorized session.
  • Test output filtering, URL validation, bounded invocation, and timeout handling.

Scope

This addresses the requested workflow-probe guidance. It does not establish an egress defect, add automatic agent telemetry or alerts, or reproduce the original v2.12.1-rc.4 incident on v3. No CLI commands, AWS interfaces, or flags change.

Validation

  • Python unit tests and Ruff checks/format pass.
  • A local counters-only smoke test succeeds.
  • A failed connection to the loopback interface produces valid JSONL.
  • Current public CLI documentation and source were checked.

No public-network benchmark or AWS calls were made. The public checkout does not contain the monorepo ../docs/PROJECT_CONTEXT.md referenced by AGENTS.md.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Expose diagnostics for transient public-runner egress slowdowns

1 participant