From 8aaa42f1019db54ee79c2a18276e31c86becc611 Mon Sep 17 00:00:00 2001
From: bot-backend <bot-backend@gendsgn.local>
Date: Sun, 28 Jun 2026 19:30:14 +0300
Subject: [PATCH] chore(tradein/deploy): don't recreate scraper on infra-only
 deploys + 120s graceful stop (phase 0+1)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The tradein-scraper container runs the same image as backend
(python -m app.scheduler_main, in-app scheduler) and owns the in-flight
scrape sweeps (a browser card takes ~15-27s). Two problems compounded to
kill running jobs on every deploy:

Phase 0 (deploy-tradein.yml): SCRAPER_CHANGED also fired on the `infra`
paths-filter (compose / workflow / deploy/**), so any generic infra edit
recreated the scraper container and SIGKILLed the running job. Dropped the
`|| infra == 'true'` term — only real scraper-code paths (already covered
by the `scraper` paths-filter) or a manual workflow_dispatch recreate it.

Phase 1 (docker-compose.prod.yml): the scraper had no stop_grace_period,
so Docker's default 10s window SIGKILLed an in-flight card (15-27s) before
it could finish. Added stop_grace_period: 120s + explicit stop_signal:
SIGTERM so a running unit can finish + checkpoint. The cooperative-drain
handler that consumes SIGTERM lands in a later phase; this is the
foundation.
---
 .forgejo/workflows/deploy-tradein.yml | 8 +++++++-
 tradein-mvp/docker-compose.prod.yml   | 5 +++++
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/.forgejo/workflows/deploy-tradein.yml b/.forgejo/workflows/deploy-tradein.yml
index eafe2980..6ae7a76c 100644
--- a/.forgejo/workflows/deploy-tradein.yml
+++ b/.forgejo/workflows/deploy-tradein.yml
@@ -276,7 +276,13 @@ jobs:
         env:
           IMAGE_TAG: latest
           GHCR_PAT: ${{ secrets.GHCR_PAT }}
-          SCRAPER_CHANGED: ${{ needs.changes.outputs.scraper == 'true' || needs.changes.outputs.infra == 'true' || github.event_name == 'workflow_dispatch' }}
+          # Phase 0: generic infra edits (compose / workflow / deploy/**) must NOT
+          # recreate the scraper and SIGKILL a running multi-hour job. Only genuine
+          # scraper-code paths (the `scraper` paths-filter already covers
+          # app/services/scrapers/**, scrape_pipeline.py, scheduler.py,
+          # scheduler_main.py, app/tasks/**) — or a manual workflow_dispatch —
+          # should trigger a scraper recreate. (infra term intentionally dropped.)
+          SCRAPER_CHANGED: ${{ needs.changes.outputs.scraper == 'true' || github.event_name == 'workflow_dispatch' }}
           GITHUB_SHA: ${{ github.sha }}
         with:
           host: ${{ secrets.DEPLOY_HOST }}
diff --git a/tradein-mvp/docker-compose.prod.yml b/tradein-mvp/docker-compose.prod.yml
index a349595b..d3a3d12f 100644
--- a/tradein-mvp/docker-compose.prod.yml
+++ b/tradein-mvp/docker-compose.prod.yml
@@ -129,6 +129,11 @@ services:
       browser:
         condition: service_started
     restart: unless-stopped
+    # Phase 1: give the in-flight scrape unit (browser card ~15-27s) time to finish
+    # + checkpoint on SIGTERM before Docker SIGKILLs. Default grace is only 10s, which
+    # SIGKILLs a running card mid-flight. (Cooperative-drain handler lands in a later phase.)
+    stop_grace_period: 120s
+    stop_signal: SIGTERM
     networks:
       - tradein-net