@oneuptime/common 12.0.14 → 12.0.15

package/Tests/Server/Utils/Monitor/Criteria/APIRequestCriteriaEvaluateOverTime.test.ts

@@ -0,0 +1,508 @@
+import APIRequestCriteria from "../../../../../Server/Utils/Monitor/Criteria/APIRequestCriteria";
+import MetricService from "../../../../../Server/Services/MetricService";
+import FindBy from "../../../../../Server/Types/AnalyticsDatabase/FindBy";
+import Metric from "../../../../../Models/AnalyticsModels/Metric";
+import {
+  CheckOn,
+  CriteriaFilter,
+  EvaluateOverTimeType,
+  FilterType,
+  NoDataPolicy,
+} from "../../../../../Types/Monitor/CriteriaFilter";
+import ProbeMonitorResponse from "../../../../../Types/Probe/ProbeMonitorResponse";
+import ObjectID from "../../../../../Types/ObjectID";
+import {
+  afterEach,
+  beforeEach,
+  describe,
+  expect,
+  jest,
+  test,
+} from "@jest/globals";
+
+/*
+ * End-to-end cover for the bug in
+ * https://github.com/OneUptime/oneuptime/issues/2321: a criteria filter set
+ * to look at "All Values over the last N minutes" fired on the first bad
+ * probe instead of waiting for the window.
+ *
+ * These tests drive the real APIRequestCriteria evaluator and stub only the
+ * metric read, so the window arithmetic, the coverage guard, the no-data
+ * policy and the comparison message are all exercised together.
+ */
+
+const NOW: Date = new Date("2026-08-20T12:00:00.000Z");
+
+const PROJECT_ID: ObjectID = new ObjectID(
+  "11111111-1111-4111-8111-111111111111",
+);
+const MONITOR_ID: ObjectID = new ObjectID(
+  "22222222-2222-4222-8222-222222222222",
+);
+
+/** Once a minute, oldest first, ending at `now`. */
+function everyMinute(values: Array<number>): Array<Metric> {
+  return values.map((value: number, index: number) => {
+    const metric: Metric = new Metric();
+    metric.value = value;
+    metric.time = new Date(
+      NOW.getTime() - (values.length - 1 - index) * 60 * 1000,
+    );
+    return metric;
+  });
+}
+
+function buildResponse(
+  overrides: Partial<ProbeMonitorResponse> = {},
+): ProbeMonitorResponse {
+  return {
+    projectId: PROJECT_ID,
+    monitorId: MONITOR_ID,
+    monitorStepId: ObjectID.generate(),
+    probeId: ObjectID.generate(),
+    failureCause: "",
+    monitoredAt: NOW,
+    isOnline: true,
+    responseCode: 200,
+    responseTimeInMs: 100,
+    ...overrides,
+  };
+}
+
+let windowSamples: Array<Metric> = [];
+
+function evaluate(input: {
+  criteriaFilter: CriteriaFilter;
+  dataToProcess: ProbeMonitorResponse;
+  monitoringInterval?: string | undefined;
+}): Promise<string | null> {
+  return APIRequestCriteria.isMonitorInstanceCriteriaFilterMet({
+    dataToProcess: input.dataToProcess,
+    criteriaFilter: input.criteriaFilter,
+    monitoringInterval: input.monitoringInterval ?? "* * * * *",
+  });
+}
+
+describe("APIRequestCriteria evaluate over time", () => {
+  beforeEach(() => {
+    jest.useFakeTimers();
+    jest.setSystemTime(NOW);
+
+    windowSamples = [];
+
+    jest
+      .spyOn(MetricService, "findBy")
+      .mockImplementation((_findBy: FindBy<Metric>): Promise<Array<Metric>> => {
+        return Promise.resolve(windowSamples);
+      });
+  });
+
+  afterEach(() => {
+    jest.useRealTimers();
+    jest.restoreAllMocks();
+  });
+
+  describe("the reported scenario: one bad status code in a healthy window", () => {
+    const offlineCriteria: CriteriaFilter = {
+      checkOn: CheckOn.ResponseStatusCode,
+      filterType: FilterType.NotEqualTo,
+      value: 200,
+      evaluateOverTime: true,
+      evaluateOverTimeOptions: {
+        timeValueInMinutes: 5,
+        evaluateOverTimeType: EvaluateOverTimeType.AllValues,
+      },
+    };
+
+    test("does not fire when only the newest sample is bad", async () => {
+      windowSamples = everyMinute([200, 200, 200, 200, 404]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(result).toBeNull();
+    });
+
+    /*
+     * The heart of the bug. Nothing had been recorded yet, so the evaluator
+     * used to compare the single status code that arrived with this check
+     * and report it as "all values over the last 5 minutes".
+     */
+    test("does not fire when the window is empty", async () => {
+      windowSamples = [];
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(result).toBeNull();
+    });
+
+    test("does not fire on a monitor that has only checked once", async () => {
+      windowSamples = everyMinute([404]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(result).toBeNull();
+    });
+
+    test("does not fire two minutes into a five minute window", async () => {
+      windowSamples = everyMinute([404, 404]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(result).toBeNull();
+    });
+
+    test("fires once the whole window has been bad", async () => {
+      windowSamples = everyMinute([404, 404, 404, 404, 404]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(result).toContain("All values of");
+      expect(result).toContain(CheckOn.ResponseStatusCode as string);
+      expect(result).toContain("over the last 5 minutes");
+      expect(result).toContain("not equal to 200");
+    });
+
+    /*
+     * The recovery direction matters too: a single good sample inside an
+     * otherwise bad window must clear the criteria.
+     */
+    test("stops firing as soon as one sample recovers", async () => {
+      windowSamples = everyMinute([404, 404, 404, 404, 200]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ responseCode: 200 }),
+      });
+
+      expect(result).toBeNull();
+    });
+  });
+
+  describe("the reported scenario: a single latency spike", () => {
+    /*
+     * From the second report on the issue - one 9171 ms tick inside a two
+     * minute window of sub-second responses opened an incident that
+     * auto-resolved 72 seconds later.
+     */
+    const slowCriteria: CriteriaFilter = {
+      checkOn: CheckOn.ResponseTime,
+      filterType: FilterType.GreaterThan,
+      value: 8000,
+      evaluateOverTime: true,
+      evaluateOverTimeOptions: {
+        timeValueInMinutes: 2,
+        evaluateOverTimeType: EvaluateOverTimeType.AllValues,
+      },
+    };
+
+    test("does not fire on a lone spike among healthy samples", async () => {
+      windowSamples = everyMinute([397, 805, 9171]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: slowCriteria,
+        dataToProcess: buildResponse({ responseTimeInMs: 9171 }),
+      });
+
+      expect(result).toBeNull();
+    });
+
+    test("fires when the whole window is slow", async () => {
+      windowSamples = everyMinute([9171, 9002, 8800]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: slowCriteria,
+        dataToProcess: buildResponse({ responseTimeInMs: 8800 }),
+      });
+
+      expect(result).toContain("All values of");
+      expect(result).toContain("greater than 8000");
+    });
+
+    /*
+     * "Any Value" is the setting for "tell me the moment one check
+     * breaches", and it must keep behaving that way.
+     */
+    test("Any Value still fires on the lone spike", async () => {
+      windowSamples = everyMinute([397, 805, 9171]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: {
+          ...slowCriteria,
+          evaluateOverTimeOptions: {
+            timeValueInMinutes: 2,
+            evaluateOverTimeType: EvaluateOverTimeType.AnyValue,
+          },
+        },
+        dataToProcess: buildResponse({ responseTimeInMs: 9171 }),
+      });
+
+      expect(result).toContain("Any value of");
+    });
+
+    test("Any Value fires off the very first sample", async () => {
+      windowSamples = everyMinute([9171]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: {
+          ...slowCriteria,
+          evaluateOverTimeOptions: {
+            timeValueInMinutes: 2,
+            evaluateOverTimeType: EvaluateOverTimeType.AnyValue,
+          },
+        },
+        dataToProcess: buildResponse({ responseTimeInMs: 9171 }),
+      });
+
+      expect(result).not.toBeNull();
+    });
+  });
+
+  describe("Is Online over time", () => {
+    const offlineCriteria: CriteriaFilter = {
+      checkOn: CheckOn.IsOnline,
+      filterType: FilterType.False,
+      value: undefined,
+      evaluateOverTime: true,
+      evaluateOverTimeOptions: {
+        timeValueInMinutes: 5,
+        evaluateOverTimeType: EvaluateOverTimeType.AllValues,
+      },
+    };
+
+    test("fires only after the whole window has been offline", async () => {
+      windowSamples = everyMinute([0, 0, 0, 0, 0]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ isOnline: false }),
+      });
+
+      expect(result).toContain(CheckOn.IsOnline as string);
+    });
+
+    test("does not fire on the first failed check", async () => {
+      windowSamples = everyMinute([1, 1, 1, 1, 0]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ isOnline: false }),
+      });
+
+      expect(result).toBeNull();
+    });
+
+    test("does not fire while the window is still filling", async () => {
+      windowSamples = everyMinute([0]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: offlineCriteria,
+        dataToProcess: buildResponse({ isOnline: false }),
+      });
+
+      expect(result).toBeNull();
+    });
+  });
+
+  describe("no-data policy", () => {
+    const criteriaWith: (policy: NoDataPolicy) => CriteriaFilter = (
+      policy: NoDataPolicy,
+    ): CriteriaFilter => {
+      return {
+        checkOn: CheckOn.ResponseStatusCode,
+        filterType: FilterType.NotEqualTo,
+        value: 200,
+        evaluateOverTime: true,
+        evaluateOverTimeOptions: {
+          timeValueInMinutes: 5,
+          evaluateOverTimeType: EvaluateOverTimeType.AllValues,
+          onNoDataPolicy: policy,
+        },
+      };
+    };
+
+    test("Ignore keeps quiet when there is nothing to look at", async () => {
+      windowSamples = [];
+
+      const result: string | null = await evaluate({
+        criteriaFilter: criteriaWith(NoDataPolicy.Ignore),
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(result).toBeNull();
+    });
+
+    /*
+     * For heartbeat-style checks the absence of data is itself the problem,
+     * which is what Trigger is for.
+     */
+    test("Trigger reports the missing data as the root cause", async () => {
+      windowSamples = [];
+
+      const result: string | null = await evaluate({
+        criteriaFilter: criteriaWith(NoDataPolicy.Trigger),
+        dataToProcess: buildResponse({ responseCode: 200 }),
+      });
+
+      expect(result).toContain("no data");
+      expect(result).toContain("5 minutes");
+    });
+
+    test("Treat As Zero compares the window as a single zero", async () => {
+      windowSamples = [];
+
+      const result: string | null = await evaluate({
+        criteriaFilter: criteriaWith(NoDataPolicy.TreatAsZero),
+        dataToProcess: buildResponse({ responseCode: 200 }),
+      });
+
+      // 0 is not equal to 200, so the filter is met.
+      expect(result).toContain("not equal to 200");
+    });
+  });
+
+  describe("monitors that are polled less often than the window", () => {
+    /*
+     * A monitor checked every five minutes can only ever put one sample into
+     * a five minute window, so that sample IS the whole window. Requiring
+     * more would mean it never fires at all.
+     */
+    test("a single sample covers the window at a five minute interval", async () => {
+      windowSamples = everyMinute([404]);
+
+      const result: string | null = await evaluate({
+        criteriaFilter: {
+          checkOn: CheckOn.ResponseStatusCode,
+          filterType: FilterType.NotEqualTo,
+          value: 200,
+          evaluateOverTime: true,
+          evaluateOverTimeOptions: {
+            timeValueInMinutes: 5,
+            evaluateOverTimeType: EvaluateOverTimeType.AllValues,
+          },
+        },
+        dataToProcess: buildResponse({ responseCode: 404 }),
+        monitoringInterval: "*/5 * * * *",
+      });
+
+      expect(result).not.toBeNull();
+    });
+  });
+
+  describe("attribute scoping", () => {
+    /*
+     * Only the disk-usage series carries a diskPath attribute. A stray
+     * serverMonitorOptions left on a non-disk filter used to be harmless
+     * because an empty window fell back to the live value; now it would
+     * silence the filter, so it must not reach the query.
+     */
+    test("does not scope a response-time window by a stray disk path", async () => {
+      windowSamples = everyMinute([9000, 9000, 9000]);
+
+      const findBySpy: ReturnType<typeof jest.spyOn> = jest.spyOn(
+        MetricService,
+        "findBy",
+      );
+
+      const result: string | null = await evaluate({
+        criteriaFilter: {
+          checkOn: CheckOn.ResponseTime,
+          filterType: FilterType.GreaterThan,
+          value: 8000,
+          evaluateOverTime: true,
+          evaluateOverTimeOptions: {
+            timeValueInMinutes: 2,
+            evaluateOverTimeType: EvaluateOverTimeType.AllValues,
+          },
+          serverMonitorOptions: { diskPath: "/" },
+        },
+        dataToProcess: buildResponse({ responseTimeInMs: 9000 }),
+      });
+
+      const query: Record<string, unknown> = (
+        findBySpy.mock.calls[0]![0] as { query: Record<string, unknown> }
+      ).query;
+
+      expect(query["attributes"]).toBeUndefined();
+      expect(result).not.toBeNull();
+    });
+  });
+
+  describe("filters that do not evaluate over time", () => {
+    test("still compare the value that arrived with this check", async () => {
+      windowSamples = [];
+
+      const result: string | null = await evaluate({
+        criteriaFilter: {
+          checkOn: CheckOn.ResponseStatusCode,
+          filterType: FilterType.NotEqualTo,
+          value: 200,
+          evaluateOverTime: false,
+        },
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(result).toContain("not equal to 200");
+    });
+
+    test("the metric store is never read for them", async () => {
+      const findBySpy: ReturnType<typeof jest.spyOn> = jest.spyOn(
+        MetricService,
+        "findBy",
+      );
+
+      await evaluate({
+        criteriaFilter: {
+          checkOn: CheckOn.ResponseStatusCode,
+          filterType: FilterType.NotEqualTo,
+          value: 200,
+          evaluateOverTime: false,
+        },
+        dataToProcess: buildResponse({ responseCode: 404 }),
+      });
+
+      expect(findBySpy).not.toHaveBeenCalled();
+    });
+  });
+
+  describe("checks with no metric series of their own", () => {
+    /*
+     * Response Body is not recorded as a metric, so an over-time flag on it
+     * cannot be honoured. It falls back to the body from this check rather
+     * than silently never matching.
+     */
+    test("fall back to the value from this check", async () => {
+      windowSamples = [];
+
+      const result: string | null = await evaluate({
+        criteriaFilter: {
+          checkOn: CheckOn.ResponseBody,
+          filterType: FilterType.Contains,
+          value: "healthy",
+          evaluateOverTime: true,
+          evaluateOverTimeOptions: {
+            timeValueInMinutes: 5,
+            evaluateOverTimeType: EvaluateOverTimeType.AllValues,
+          },
+        },
+        dataToProcess: buildResponse({ responseBody: "all healthy here" }),
+      });
+
+      expect(result).toContain("Response body contains healthy");
+    });
+  });
+});

package/Tests/Server/Utils/Monitor/Criteria/APIRequestCriteriaPortTimings.test.ts

@@ -1,5 +1,7 @@
 import APIRequestCriteria from "../../../../../Server/Utils/Monitor/Criteria/APIRequestCriteria";
-import EvaluateOverTime from "../../../../../Server/Utils/Monitor/Criteria/EvaluateOverTime";
+import EvaluateOverTime, {
+  OverTimeCriteriaValue,
+} from "../../../../../Server/Utils/Monitor/Criteria/EvaluateOverTime";
 import {
   CheckOn,
   CriteriaFilter,
@@ -142,8 +144,11 @@ describe("APIRequestCriteria Port connection phase timings", () => {
 
   test("uses an evaluate-over-time aggregate instead of the current phase", async () => {
     const overTimeSpy: ReturnType<typeof jest.spyOn> = jest
-      .spyOn(EvaluateOverTime, "getValueOverTime")
-      .mockResolvedValue(250);
+      .spyOn(EvaluateOverTime, "getOverTimeValueForCriteriaFilter")
+      .mockResolvedValue({
+        earlyReturn: null,
+        value: 250,
+      } as OverTimeCriteriaValue);
 
     const response: ProbeMonitorResponse = buildResponse({
       dnsLookupInMs: 5,
@@ -164,13 +169,20 @@ describe("APIRequestCriteria Port connection phase timings", () => {
       expect.objectContaining({
         projectId: response.projectId,
         monitorId: response.monitorId,
-        metricType: CheckOn.PortDnsLookupTime,
+        criteriaFilter: expect.objectContaining({
+          checkOn: CheckOn.PortDnsLookupTime,
+        }),
       }),
     );
   });
 
   test("does not replace a zero evaluate-over-time aggregate with the current value", async () => {
-    jest.spyOn(EvaluateOverTime, "getValueOverTime").mockResolvedValue(0);
+    jest
+      .spyOn(EvaluateOverTime, "getOverTimeValueForCriteriaFilter")
+      .mockResolvedValue({
+        earlyReturn: null,
+        value: 0,
+      } as OverTimeCriteriaValue);
 
     const result: string | null = await evaluate(
       buildResponse({
@@ -189,10 +201,19 @@ describe("APIRequestCriteria Port connection phase timings", () => {
     expect(result).not.toBeNull();
   });
 
-  test("falls back to the current phase if the history query fails", async () => {
+  /*
+   * An over-time filter whose history could not be read says nothing about
+   * the last N minutes, so it must not fire. It used to quietly compare the
+   * single phase timing from this one check instead - the same fallback that
+   * let "all values over the last N minutes" fire off one bad probe.
+   */
+  test("does not fire off the current phase if the history query fails", async () => {
     jest
-      .spyOn(EvaluateOverTime, "getValueOverTime")
-      .mockRejectedValue(new Error("history unavailable"));
+      .spyOn(EvaluateOverTime, "getOverTimeValueForCriteriaFilter")
+      .mockResolvedValue({
+        earlyReturn: { result: null },
+        value: undefined,
+      } as OverTimeCriteriaValue);
 
     const result: string | null = await evaluate(
       buildResponse({
@@ -207,6 +228,6 @@ describe("APIRequestCriteria Port connection phase timings", () => {
       }),
     );
 
-    expect(result).not.toBeNull();
+    expect(result).toBeNull();
   });
 });