Better stop behavior, support override entrypoint

[sjmiller609]

Summary

• Graceful shutdown: adds Shutdown gRPC RPC so hypeman stop sends SIGTERM to the app (via guest-agent -> init -> entrypoint), with a configurable timeout and hard-kill fallback. Replaces syscall.Exit with reboot(POWER_OFF) to eliminate the kernel panic on app exit.
• Exit info propagation: init writes a HYPEMAN-EXIT sentinel to serial console before shutdown. Host lazily parses it when the VM stops and persists exit_code + exit_message to instance metadata and the API response. Includes signal descriptions, OOM detection via /dev/kmsg, and shell convention codes (126/127).
• Entrypoint/Cmd override on create and start, matching docker run --entrypoint / docker run <image> <command>.
• All init log phases prefixed with hypeman-init: to distinguish from guest-agent and app output.

Test plan

• Unit tests for describeExitCode, formatExitSentinel, isOOMLine, parseExitSentinelLine (including carriage return handling)
• TestAppExitPropagation: alpine with /nonexistent-command -> exit 127 "command not found" in metadata
• TestOOMExitPropagation: 128MB VM with memory-filling loop -> exit 137 "killed by signal 9 (killed) - OOM" in metadata
• TestBasicEndToEnd: graceful stop -> verify ExitCode populated -> restart -> verify ExitCode cleared -> stop -> delete
• Graceful stop assertions added to TestQEMUBasicEndToEnd

---

Note

Medium Risk
Touches core instance lifecycle (start/stop/init) and introduces new gRPC/API contract changes, which could affect VM shutdown behavior and metadata persistence if edge cases/regressions exist.

Overview
Implements graceful VM stop by adding a Shutdown gRPC RPC (guest-agent signals PID 1) and updating StopInstance to attempt a timed graceful shutdown before falling back to hard-killing the hypervisor.

Adds exit info propagation: guest init now writes a machine-parseable HYPEMAN-EXIT sentinel (including OOM detection) and powers off via reboot(POWER_OFF); the host lazily parses/persists exit_code/exit_message from serial logs and returns them via the API.

Adds Docker-like entrypoint/cmd overrides on CreateInstance and StartInstance (with restart clearing stale exit info), updates OpenAPI/generated client to accept a JSON body for start, and adjusts/extends integration+unit tests plus increases test timeouts.

^{Written by Cursor Bugbot for commit 06cd0e8. This will update automatically on new commits. Configure here.}

[github-actions]

✱ Stainless preview builds

This PR will update the hypeman SDKs with the following commit message.

feat: Better stop behavior

⚠️ hypeman-openapi studio · code

There was a regression in your SDK.
generate ⚠️

⚠️ hypeman-typescript studio · code

There was a regression in your SDK.
generate ❗ → build ✅ → lint ✅ → test ✅

npm install https://pkg.stainless.com/s/hypeman-typescript/9f372846f08ec591342f525498449861990821c5/dist.tar.gz

⚠️ hypeman-go studio · code

There was a regression in your SDK.
generate ⚠️ → lint ✅ → test ✅

go get github.com/stainless-sdks/hypeman-go@12cfb4e4ddfc198c65b09efeb73c8db0fd609f5f

---

This comment is auto-generated by GitHub Actions and is automatically kept up to date as you push.
If you push custom code to the preview branch, re-run this workflow to update the comment.
Last updated: 2026-02-13 19:59:23 UTC

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

[hiroTamada]

Looks good -- clean design for graceful shutdown, exit info propagation, and entrypoint/cmd overrides. A couple nits on typos in manager.go.