chore: increase workspace readiness timeout to 60s and add startup log

[jamesc]

Summary

Fixes intermittent MCP integration test failures in CI (run 22522874219) where all 18 MCP tests failed because the 30-second workspace readiness budget was exhausted.

Root cause

READINESS_PROBE_MAX_RETRIES = 150 × 200ms = 30 seconds — too tight when the workspace integration tests (which start/kill 12 BEAM nodes in ~10s) run immediately before MCP tests, leaving the CI runner under load.

The failure signature: ECONNREFUSED on port 46141 for the full 30s after PID + port file were written, suggesting the BEAM node either took >30s to bind or crashed and rolled back before any probe landed.

Changes

• READINESS_PROBE_MAX_RETRIES 150 → 300 — doubles the readiness budget to 60 seconds to handle loaded CI runners
• BEAM node stderr → startup.log — redirects OTP crash reports / error_logger output to {workspace_dir}/startup.log instead of /dev/null, giving actionable diagnostics on future failures
• Better timeout error — checks PID liveness via sysinfo and emits a distinct message for crash (PID gone) vs slow startup (PID alive), including the log file path in both cases

Test plan

• just build && just clippy && just fmt-check — passes
• just test — 1898 Rust + 237 stdlib + 645 BUnit + 2208 runtime tests pass
• CI MCP tests should no longer time out on loaded runners

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Increased startup timeout window from 30 to 60 seconds for BEAM node initialization.
  • Redirected node stderr to a startup log (with a non-fatal fallback) to capture crash diagnostics.
  • Added diagnostic-backed messaging to distinguish slow startups from process crashes, including optional log file hints for troubleshooting.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Disabled knowledge base sources:

• Linear integration is disabled

You can enable these sources in your CodeRabbit configuration.

📥 Commits

Reviewing files that changed from the base of the PR and between f6b0696 and 8e37dbe.

📒 Files selected for processing (1)

• crates/beamtalk-cli/src/commands/workspace/process.rs

---

📝 Walkthrough

Walkthrough

Doubles the BEAM readiness probe retries (150→300), redirects BEAM stderr to a startup log (with /dev/null fallback), passes the log path into the TCP readiness waiter, and augments readiness-timeout errors to distinguish slow startup vs. crashed process with optional log-path hints.

Changes

Cohort / File(s)	Summary
BEAM Startup Diagnostics & Readiness Enhancement crates/beamtalk-cli/src/commands/workspace/process.rs	READINESS_PROBE_MAX_RETRIES increased from 150→300; BEAM stderr redirected to startup.log (fallback to /dev/null on failure); wait_for_tcp_ready signature extended to accept an optional log path and all call sites updated; readiness failure handling now reports alive vs. crashed process and includes log-path diagnostic hints.

Sequence Diagram(s)

mermaid
sequenceDiagram
participant CLI as CLI
participant BEAM as BEAM process
participant Waiter as wait_for_tcp_ready
participant Log as startup.log
CLI->>BEAM: spawn BEAM, redirect stderr -> startup.log (or /dev/null)
CLI->>Waiter: wait_for_tcp_ready(host, port, retries, optional log path)
loop retry until timeout or ready
Waiter->>BEAM: check TCP port
alt TCP open
Waiter-->>CLI: ready
else still closed
Waiter->>BEAM: check process liveness
end
end
alt timeout reached
alt BEAM alive
Waiter-->>CLI: error("slow startup", log path?)
else BEAM crashed
Waiter-->>CLI: error("crashed", include log path)
end
end

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• Use --timeout 300 in MCP test REPL to prevent orphaned workspaces (BT-789) #793: Modifies the same readiness probing logic in crates/beamtalk-cli/src/commands/workspace/process.rs; earlier increase of retry budget and related changes.
• Fix flaky workspace integration tests (BT-662) #646: Introduced wait_for_tcp_ready usage in node startup and TCP readiness/retry behavior that this change extends with log-path diagnostics.

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately reflects the main changes: increasing the workspace readiness timeout to 60s and adding startup log functionality for diagnostics.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

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

✨ Finishing Touches

• 📝 Generate docstrings (stacked PR)
• 📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch worktree-main-mcp-fails

---

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

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@crates/beamtalk-cli/src/commands/workspace/process.rs`:
- Around line 331-340: The startup-log attempt currently falls back silently to
/dev/null; modify the code around the OpenOptions block that tries to open
startup_log_path so that on Err(_) you print a warning with eprintln! (including
the path and the error) and set a flag (e.g., startup_log_enabled = true/false)
when you successfully call cmd.stderr(Stdio::from(log_file)). Then update the
timeout/error messages that currently unconditionally tell users to inspect
startup_log_path (the messages produced later where you reference the startup
log on failures) to only mention the startup log path when startup_log_enabled
is true; otherwise omit that suggestion. Apply the same pattern to the other
similar blocks referenced in the review (the other startup/log-related message
sites), using the unique names startup_log_path, cmd.stderr(...), and the
timeout/error message locations to find and update the code.

---

ℹ️ Review info

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Disabled knowledge base sources:

• Linear integration is disabled

You can enable these sources in your CodeRabbit configuration.

📥 Commits

Reviewing files that changed from the base of the PR and between afab809 and f6b0696.

📒 Files selected for processing (1)

• crates/beamtalk-cli/src/commands/workspace/process.rs