fix(server): service launcher starts its server again once another server lets go of the T3 home#24230

External image Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: main
• File: .coderabbit.config.ts
• Commit: b24e8ce
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Advanced
• Run ID: d3057371-f27a-489f-87e4-a5aadb315ad8

External image Commits

Reviewing files that changed from the base of the PR and between e4440594601605ff89ec7339feba8155fd5341b5 and 5fe142bf5ec2d3a05d34f5e3005d875fa6d3d71e.

External image Files selected for processing: 1

• apps/server/src/serviceLauncher.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 8 remain after this review.

The launcher now checks state-directory ownership after an ownership refusal. It polls for ownership to clear and starts the active version when the state directory is no longer owned. Tests cover ownership checks and recovery after a lock is released.

Changes: State directory ownership recovery

Layer / File(s)	Summary
Detect state-directory ownership apps/server/src/serviceLauncher.ts, apps/server/src/serviceLauncher.test.ts	stateDirOwned checks for a busy SQLite lock. If the lock is not busy, it checks for a runtime record without ownerId and a live positive-integer PID. The launcher accepts a configurable polling interval, which defaults to 30 seconds. Tests cover these ownership checks.
Poll for lock release and restart apps/server/src/serviceLauncher.ts, apps/server/src/serviceLauncher.test.ts	After an ownership refusal, the launcher polls for ownership to clear and starts the active version when it clears. A pending update remains failed without restoring its backup. The integration test checks that the launcher starts again after another process releases the lock.

Priority: External image Normal

Estimated code review effort: 3 (Moderate) | ~20 minutes

Change: Bug fix · Severity of issue fixed: Medium

Sequence Diagram(s)

mermaid

1	sequenceDiagram
2	participant Launcher
3	participant stateDirOwned
4	participant SQLite
5	Launcher->>stateDirOwned: Check state-directory ownership
6	stateDirOwned->>SQLite: Attempt exclusive lock
7	SQLite-->>stateDirOwned: Return busy status or allow access
8	stateDirOwned-->>Launcher: Return ownership result
9	loop At the configured polling interval
10	Launcher->>stateDirOwned: Check ownership again
11	stateDirOwned->>SQLite: Attempt exclusive lock
12	SQLite-->>stateDirOwned: Return busy status or allow access
13	stateDirOwned-->>Launcher: Return ownership result
14	end
15	Launcher->>Launcher: Start active version when ownership clears

Suggested labels: bugfix, pr:needs-verification

Suggested reviewers: maria-rcks

No material issue remains identified in the ownership-recovery change; it is ready to merge after normal checks.

Check name	Status	Explanation
Description check	Passed	The description includes the required Problem, Change, Scope and approval, and Verification sections. It explains the failure, implementation, tests, manual validation, trade-offs, and untested scenar…
Title check	Passed	The title is concise, specific, and accurately describes the primary change: the service launcher restarts after another server releases the T3 home.
Linked Issues check	Passed	Issue #24225 requires host-side recovery after a child exits with code 78. The launcher now polls every 30 seconds inside the launcher process. stateDirOwned checks the non-blocking SQLite owner loc…
Out of Scope Changes check	Passed	The reported changes are limited to the service launcher and its tests. The ownership probe, polling, recovery flow, log update, and tests directly support issue #24225. No unrelated implementation ch…

Full details: Description check

Explanation

The description includes the required Problem, Change, Scope and approval, and Verification sections. It explains the failure, implementation, tests, manual validation, trade-offs, and untested scenarios.

Full details: Linked Issues check

Explanation

Issue #24225 requires host-side recovery after a child exits with code 78. The launcher now polls every 30 seconds inside the launcher process. stateDirOwned checks the non-blocking SQLite owner lock and a live pre-lock server-runtime.json PID. The launcher starts the active version when both checks are clear. The idle path still marks a pending update failed. Tests cover lock ownership, pre-lock records, continued idle operation, and recovery after lock release. The optional SSH launch-script change is separate and is not required for the host-side fix.

Full details: Out of Scope Changes check

Explanation

The reported changes are limited to the service launcher and its tests. The ownership probe, polling, recovery flow, log update, and tests directly support issue #24225. No unrelated implementation change is shown. The SSH launch-script change is explicitly excluded.

🧪 Generate unit tests (beta)

• Create a new PR

🏁 Script executed:

Length of output: 14854