Removes HTTP error

src/routers/deriver_metrics.py

@@ -1,18 +1,15 @@
-"""Deriver work metrics as JSON. Errors rather than serving a stale number."""
+"""Deriver work metrics as JSON, with the age of the measurement alongside them."""
 
 from logging import getLogger
 
 from fastapi import APIRouter, HTTPException
 
 from src.backlog import DeriverMetricsPoller
-from src.config import settings
 
 logger = getLogger(__name__)
 
 router = APIRouter(prefix="/deriver", tags=["deriver"])
 
-MAX_SNAPSHOT_AGE_INTERVALS = 3
-
 _poller: DeriverMetricsPoller | None = None
 
 
@@ -21,13 +18,6 @@ def set_deriver_metrics_poller(poller: DeriverMetricsPoller | None) -> None:
     _poller = poller
 
 
-def max_snapshot_age_seconds() -> float:
-    return float(
-        settings.DERIVER.BACKLOG_METRICS_POLL_INTERVAL_SECONDS
-        * MAX_SNAPSHOT_AGE_INTERVALS
-    )
-
-
 @router.get("/metrics")
 async def get_deriver_metrics_response() -> dict[str, float | int]:
     """Seconds of outstanding deriver work, plus the raw counts behind it."""
@@ -37,15 +27,6 @@ async def get_deriver_metrics_response() -> dict[str, float | int]:
             status_code=503, detail="No deriver measurement available yet"
         )
 
-    age = snapshot.age_seconds
-    if age is None or age > max_snapshot_age_seconds():
-        raise HTTPException(
-            status_code=503,
-            detail=f"Deriver measurement is stale ({age:.0f}s old)"
-            if age is not None
-            else "Deriver measurement is stale",
-        )
-
     return {
         "outstanding_work_seconds": snapshot.signal_seconds,
         "eligible_work_units": snapshot.stats.eligible_work_units,
@@ -56,5 +37,5 @@ async def get_deriver_metrics_response() -> dict[str, float | int]:
         "embeddings_pending_due": snapshot.stats.embeddings_pending_due,
         "dreams_due": snapshot.dreams_due,
         "measured_at": snapshot.measured_at,
-        "measurement_age_seconds": age,
+        "measurement_age_seconds": snapshot.age_seconds or 0.0,
     }

tests/test_deriver_metrics.py

@@ -162,20 +162,21 @@ class TestDeriverMetricsRoute:
 
         assert excinfo.value.status_code == 503
 
-    async def test_errors_when_the_measurement_is_stale(self):
+    async def test_serves_an_old_snapshot_with_its_age(self):
+        """The caller decides what is too old, from measurement_age_seconds."""
         poller = DeriverMetricsPoller()
         poller._snapshot = DeriverMetricsSnapshot(  # pyright: ignore[reportPrivateUsage]
-            signal_seconds=0.0,
-            measured_at=time.time() - (deriver_metrics.max_snapshot_age_seconds() + 60),
+            signal_seconds=7.0,
+            measured_at=time.time() - 3600,
         )
         deriver_metrics.set_deriver_metrics_poller(poller)
         try:
-            with pytest.raises(HTTPException) as excinfo:
-                await deriver_metrics.get_deriver_metrics_response()
+            body = await deriver_metrics.get_deriver_metrics_response()
         finally:
             deriver_metrics.set_deriver_metrics_poller(None)
 
-        assert excinfo.value.status_code == 503
+        assert body["outstanding_work_seconds"] == 7.0
+        assert body["measurement_age_seconds"] >= 3600
 
     async def test_errors_when_no_poller_is_registered(self):
         deriver_metrics.set_deriver_metrics_poller(None)