Enforce log limit per monitor

src/critic/libs/ddb.py

@@ -143,7 +143,7 @@ def get(cls, partition_value: Any, sort_value: Any | None = None) -> BaseModel |
         return cls.ddb_to_model(response['Item']) if 'Item' in response else None
 
     @classmethod
-    def query(cls, partition_value: Any) -> list[BaseModel]:
+    def query(cls, partition_value: Any, **kwargs) -> list[BaseModel]:
         """Query for all items with the given partition key."""
         names, values, clauses = cls.alias({cls.partition_key: partition_value})
 
@@ -152,6 +152,7 @@ def query(cls, partition_value: Any) -> list[BaseModel]:
             KeyConditionExpression=clauses[0],
             ExpressionAttributeNames=names,
             ExpressionAttributeValues=values,
+            **kwargs,
         )
 
         items = response.get('Items', [])

src/critic/libs/uptime.py

@@ -129,6 +129,7 @@ def put_log(
         status_code: int,
         latency: float,
         error_messages: list[str],
+        log_counter: int,
     ):
         """
         Puts a log for the check. This method should only be called once per monitor check.
@@ -143,6 +144,11 @@ def put_log(
             latency_secs=latency,
             error_message=error_messages if error_messages else None,
         )
+        if log_counter >= UptimeLogTable.retention_limit:
+            # The "-1" is to make room for the log we're about to put.
+            UptimeLogTable.prune(
+                self.monitor.id, log_counter - (UptimeLogTable.retention_limit - 1)
+            )
         UptimeLogTable.put(uptime_log)
         self._put_log = True
 
@@ -171,7 +177,13 @@ def run(self):
         state, consecutive_fails, error_messages = self.check_resp(resp)
 
         # Update the monitor
-        updated = self.update_monitor({'state': state, 'consecutive_fails': consecutive_fails})
+        updated = self.update_monitor(
+            {
+                'state': state,
+                'consecutive_fails': consecutive_fails,
+                'log_counter': min(self.monitor.log_counter + 1, UptimeLogTable.retention_limit),
+            }
+        )
 
         # Save a log
         if updated:
@@ -187,4 +199,5 @@ def run(self):
                 resp.status_code if resp else 0,
                 latency,
                 error_messages,
+                self.monitor.log_counter,
             )

src/critic/models.py

@@ -31,6 +31,7 @@ class UptimeMonitorModel(BaseModel):
     state: MonitorState = Field(default=MonitorState.new)
     frequency_mins: int = Field(ge=1, default=1)
     consecutive_fails: int = Field(ge=0, default=0)
+    log_counter: int = Field(ge=0, default=0)
     next_due_at: AwareDatetime = Field(
         default_factory=lambda: datetime.now(UTC).replace(second=0, microsecond=0)
     )

src/critic/tables.py

@@ -66,3 +66,10 @@ class UptimeLogTable(Table):
     model = UptimeLogModel
     partition_key = 'monitor_id'
     sort_key = 'timestamp'
+    retention_limit = 1440  # 24 hours if logging once a minute
+
+    @classmethod
+    def prune(cls, monitor_id: str, n: int):
+        """Prune the n oldest logs"""
+        for log in cls.query(monitor_id, ScanIndexForward=True, Limit=n):
+            cls.delete(monitor_id, getattr(log, cls.sort_key))

tests/critic_tests/test_libs/test_uptime.py

@@ -27,9 +27,9 @@ def test_duplicate_update(self):
     def test_duplicate_log(self):
         monitor = UptimeMonitorFactory.put()
         check = UptimeCheck(str(monitor.project_id), monitor.slug)
-        check.put_log(MonitorState.up, 200, 0.1, None)
+        check.put_log(MonitorState.up, 200, 0.1, None, 0)
         with pytest.raises(Exception, match='Log already put'):
-            check.put_log(MonitorState.up, 200, 0.1, None)
+            check.put_log(MonitorState.up, 200, 0.1, None, 0)
 
     def test_race_condition(self, httpx_mock):
         monitor = UptimeMonitorFactory.put(next_due_at='2026-02-10 11:50:00Z')
@@ -74,6 +74,7 @@ def test_run_up(self, caplog, httpx_mock):
         assert response.state == MonitorState.up
         assert response.next_due_at > time_to_check
         assert response.consecutive_fails == 0
+        assert response.log_counter == 1
 
         response: UptimeLogModel = UptimeLogTable.query(monitor.id)[-1]
 
@@ -97,6 +98,7 @@ def test_down_with_consec_fails_above_threshold(self, httpx_mock):
         # Monitor should be down with 2 consec fails
         assert response.state == MonitorState.down
         assert response.consecutive_fails == 2
+        assert response.log_counter == 1
 
         response: UptimeLogModel = UptimeLogTable.query(monitor.id)[-1]
         # log should have resp of 0 since there was a timeout
@@ -117,6 +119,7 @@ def test_down_with_consec_fails_below_threshold(self, httpx_mock):
         response: UptimeMonitorModel = UptimeMonitorTable.get(monitor.project_id, monitor.slug)
         assert response.state == MonitorState.down
         assert response.consecutive_fails == 1
+        assert response.log_counter == 1
 
         response: UptimeLogModel = UptimeLogTable.query(monitor.id)[-1]
         # log should have resp of 0 since there was a timeout
@@ -193,3 +196,26 @@ def test_assertion_fails_with_multiple_errors(self, httpx_mock):
         response: UptimeLogModel = UptimeLogTable.query(monitor_id)[-1]
 
         assert '404' in response.error_message[1]
+
+    def test_log_retention_limit(self, monkeypatch, httpx_mock):
+        monkeypatch.setattr(UptimeLogTable, 'retention_limit', 3)
+        monitor: UptimeMonitorModel = UptimeMonitorFactory.put(
+            next_due_at='2026-02-01 12:00:00Z',
+            frequency_mins=1,
+        )
+
+        for minute in range(4):
+            current = datetime(2026, 2, 1, 12, minute, 0, tzinfo=UTC)
+            httpx_mock.add_response()
+            with freeze_time(current):
+                UptimeCheck(str(monitor.project_id), monitor.slug).run()
+
+        logs = UptimeLogTable.query(monitor.id)
+        assert len(logs) == 3
+        monitor = UptimeMonitorTable.get(monitor.project_id, monitor.slug)
+        assert monitor.log_counter == 3
+        assert [log.timestamp for log in logs] == [
+            datetime(2026, 2, 1, 12, 1, 0, tzinfo=UTC),
+            datetime(2026, 2, 1, 12, 2, 0, tzinfo=UTC),
+            datetime(2026, 2, 1, 12, 3, 0, tzinfo=UTC),
+        ]