HH-635: let SSClash own Mihomo lifecycle #6

2026-08-25T10:08:12+08:00

rogee commented

2026-08-25 10:08:12 +08:00

Root cause

Bootstrap started Mihomo directly while SSClash Web UI only tracked Mihomo processes it started. This split lifecycle ownership made /api/status report stopped while Mihomo was running, so Web Start failed with “Mihomo already running”. Read-only host inspection found one SSClash and one Mihomo process and no stale PID files.

A review follow-up also found that volumes created before PR #4 retained the packaged GEOIP,CN rule and could attempt a runtime geoip.metadb download.

Fix

Make ssclash serve the only bootstrap-supervised process and let SSClash exclusively own Mihomo Start/Stop/status.
Keep the validated subscription as a managed in-memory provider link and validate updates against the persistent core/config.
Atomically migrate only the exact legacy packaged config to managed config version 1 with the local ChinaIp provider. Preserve customized legacy configs and fail with explicit remediation.
Use SSClash Web readiness for health checks and document the Web Start step.
Cover initial/same-volume Web Start, offline legacy-volume migration, and rollback failure through a fake controller.

Verification

./scripts/test.sh (coverage 75.5%, minimum 65%)
./tests/container-smoke.sh mohomo-docker:hh-635-config-migration
git diff --check

Fixes HH-635

## Root cause Bootstrap started Mihomo directly while SSClash Web UI only tracked Mihomo processes it started. This split lifecycle ownership made `/api/status` report stopped while Mihomo was running, so Web Start failed with “Mihomo already running”. Read-only host inspection found one SSClash and one Mihomo process and no stale PID files. A review follow-up also found that volumes created before PR #4 retained the packaged `GEOIP,CN` rule and could attempt a runtime `geoip.metadb` download. ## Fix - Make `ssclash serve` the only bootstrap-supervised process and let SSClash exclusively own Mihomo Start/Stop/status. - Keep the validated subscription as a managed in-memory provider link and validate updates against the persistent core/config. - Atomically migrate only the exact legacy packaged config to managed config version 1 with the local `ChinaIp` provider. Preserve customized legacy configs and fail with explicit remediation. - Use SSClash Web readiness for health checks and document the Web Start step. - Cover initial/same-volume Web Start, offline legacy-volume migration, and rollback failure through a fake controller. ## Verification - `./scripts/test.sh` (coverage 75.5%, minimum 65%) - `./tests/container-smoke.sh mohomo-docker:hh-635-config-migration` - `git diff --check` Fixes HH-635

rogee added 1 commit 2026-08-25 10:08:12 +08:00

HH-635: let SSClash own Mihomo lifecycle

Docker image / Test (pull_request) Successful in 2m29s

Details

Docker image / Build and publish (pull_request) Successful in 2m43s

Details

1c9c91f63e

Co-authored-by: multica-agent <github@multica.ai>

rogee added 1 commit 2026-08-25 10:30:02 +08:00

HH-635: migrate legacy managed configs

Docker image / Test (pull_request) Successful in 1m24s

Details

Docker image / Build and publish (pull_request) Successful in 2m20s

Details

6348abd765

Co-authored-by: multica-agent <github@multica.ai>

rogee merged commit f29a3fc855 into main

2026-08-25 10:42:08 +08:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: rogee/mohomo-docker#6