Pin image versions, move deploy files out, restore failure streak aft…

.github/workflows/docker.yml

@@ -31,7 +31,6 @@ jobs:
           images: ghcr.io/${{ github.repository_owner }}/openedx-status-api
           tags: |
             type=semver,pattern={{version}}
-            type=raw,value=latest,enable=${{ !contains(github.ref_name, '-') }}
       - uses: docker/build-push-action@v6
         with:
           context: .

README.md

@@ -6,23 +6,9 @@ Django, so it keeps working when Open edX is down.
 
 ## Quick start
 
-This runs the API, the web page ([openedx-status-web](https://github.com/Abstract-Tech/openedx-status-web))
-and a Caddy proxy with Docker Compose:
-
-    cd deploy
-    cp .env.example .env              # set STATUS_DOMAIN, and point its DNS at this machine
-    cp checks.example.yml checks.yml  # put your own addresses in
-    docker compose up -d
-
-Then open `https://<STATUS_DOMAIN>`. To try it locally without HTTPS, set `STATUS_DOMAIN=:80`,
-`STATUS_BIND=127.0.0.1` and `STATUS_HTTP_PORT=8088` in `.env`.
-
-Day to day:
-
-- Change the checks: edit `checks.yml`, then `docker compose up -d --force-recreate status-api` (a plain restart can keep a stale copy of the file).
-- Logs: `docker compose logs -f status-api`.
-- Backup: the history is the SQLite file in the `status-data` volume.
-- Already have a reverse proxy? Remove the `caddy` service and point your proxy at `status-web:8080`.
+To run the API, the web page ([openedx-status-web](https://github.com/Abstract-Tech/openedx-status-web))
+and a Caddy proxy together, use [openedx-status-deploy](https://github.com/Abstract-Tech/openedx-status-deploy).
+This repo only contains the API and its Docker image (`ghcr.io/abstract-tech/openedx-status-api`).
 
 To run the API alone without Docker:
 
@@ -44,7 +30,7 @@ If you mount a host folder as `/data`, run `chown 10001` on it, because the cont
 - `params`: extra options for some kinds
 
 To keep a secret out of the file, write `<name>_env: SOME_VARIABLE` instead of the value, for example
-`password_env: DB_PASSWORD`. With the compose file in `deploy/`, put `DB_PASSWORD=...` in `deploy/.env`
+`password_env: DB_PASSWORD`. With [openedx-status-deploy](https://github.com/Abstract-Tech/openedx-status-deploy), put `DB_PASSWORD=...` in its `.env`
 (it is passed to the API container) and recreate the container. This works for the http `expected_status` and `verify_tls`, the mysql `user`,
 `password` and `database`, the mongo `username` and `password`, the redis `password`, and the ssl `warn_days`.
 

openedx_status_api/incidents.py

@@ -10,6 +10,11 @@ def __init__(self, store: Store, threshold: int):
         self._fails: dict[str, int] = {}
         self._first_fail: dict[str, int] = {}  # ts of the first failure in the current streak
 
+    def restore(self, check: Check):
+        n, first = self.store.failure_streak(check.id, self.threshold)
+        if n:
+            self._fails[check.id], self._first_fail[check.id] = n, first
+
     def observe(self, check: Check, r: Result):
         open_id = self.store.open_incident_for(check.id)
         if r.ok:

openedx_status_api/runner.py

@@ -22,6 +22,8 @@ def start(self):
             n = self.store.resolve_orphaned_incidents([c.id for c in self.checks], int(time.time()))
             if n:
                 log.info("resolved %d incident(s) for checks no longer configured", n)
+        for c in self.checks:
+            self.tracker.restore(c)
         self._tasks = [asyncio.create_task(self._loop(c), name=f"check-{c.id}") for c in self.checks]
         self._tasks.append(asyncio.create_task(self._prune_loop(), name="prune"))
 

openedx_status_api/store.py

@@ -55,6 +55,18 @@ def is_failing(self, check_id: str, n: int) -> bool:
             ).fetchall()
         return len(rows) == n and not any(r["ok"] for r in rows)
 
+    def failure_streak(self, check_id: str, limit: int) -> tuple[int, int | None]:
+        with self._lock:
+            rows = self._db.execute(
+                "SELECT ts, ok FROM results WHERE check_id=? ORDER BY ts DESC, rowid DESC LIMIT ?", (check_id, limit)
+            ).fetchall()
+        streak = []
+        for r in rows:
+            if r["ok"]:
+                break
+            streak.append(r["ts"])
+        return len(streak), (streak[-1] if streak else None)
+
     def days(self, check_id: str, retention: int = 90, now: int | None = None) -> list[dict]:
         """Exactly `retention` entries oldest->newest, zero-filled."""
         today = datetime.fromtimestamp(now or time.time(), timezone.utc).date()

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "openedx-status-api"
-version = "0.1.0"
+version = "0.1.2"
 description = "Standalone status-page backend for Open edX"
 readme = "README.md"
 license = {text = "MIT"}

tests/test_core.py

@@ -154,3 +154,15 @@ def test_never_reported_check_waits_but_stale_one_degrades(store, settings):
     assert build_status(settings, cs, store, NOW)["groups"][0]["status"] == "operational"  # b has not reported yet
     store.record("b", Result(NOW - 1000, True, "ok"))
     assert build_status(settings, cs, store, NOW)["groups"][0]["status"] == "degraded"  # b went stale
+
+
+def test_failure_streak_survives_restart(store):
+    c = Check("a", "A", "G", "tcp", "x:1")
+    for i in range(2):
+        store.record("a", Result(NOW + i, False, ""))
+    t = IncidentTracker(store, 3)
+    t.restore(c)
+    store.record("a", Result(NOW + 2, False, ""))
+    t.observe(c, Result(NOW + 2, False, ""))
+    assert store.open_incident_for("a") is not None
+    assert store.incidents(0)[0]["started"] == NOW