@oneuptime/common 12.0.34 → 13.0.0

package/Tests/Server/Utils/Monitor/MonitorCriteriaEvaluatorMetricUnits.test.ts

@@ -0,0 +1,570 @@
+/*
+ * The evaluator's import chain pulls the native isolated-vm addon
+ * (MonitorCriteriaEvaluator → VMAPI → VMRunner). Nothing under test here
+ * touches the sandbox, and the prebuilt binary cannot always dlopen in the
+ * test environment — so stub the module out before anything imports it.
+ */
+jest.mock("isolated-vm", () => {
+  return {};
+});
+
+import MonitorCriteriaEvaluator from "../../../../Server/Utils/Monitor/MonitorCriteriaEvaluator";
+import CompareCriteria from "../../../../Server/Utils/Monitor/Criteria/CompareCriteria";
+import Monitor from "../../../../Models/DatabaseModels/Monitor";
+import MonitorStep from "../../../../Types/Monitor/MonitorStep";
+import MonitorCriteriaInstance from "../../../../Types/Monitor/MonitorCriteriaInstance";
+import MetricCriteriaContext from "../../../../Types/Monitor/MetricMonitor/MetricCriteriaContext";
+import {
+  CheckOn,
+  CriteriaFilter,
+  FilterType,
+} from "../../../../Types/Monitor/CriteriaFilter";
+import FilterCondition from "../../../../Types/Filter/FilterCondition";
+import ObjectID from "../../../../Types/ObjectID";
+import { describe, expect, test } from "@jest/globals";
+
+/*
+ * HUMAN-READABLE METRIC VALUES IN THE ROOT CAUSE.
+ *
+ * The metric root cause is what an alert or incident email shows under
+ * "Root Cause" — MonitorCriteriaEvaluator writes it, AlertService and
+ * IncidentService copy it onto the row, and the owner-notification jobs
+ * run it through Markdown.convertToHTML for the email body.
+ *
+ * Before this suite existed, nothing anywhere pinned the Breaching
+ * Samples table or the "- Unit:" line, and both rendered raw digits with
+ * the exporter's raw UCUM code glued on: a memory breach that the
+ * dashboard drew as "1.07 GB" arrived in the inbox as "1073741824 By",
+ * and a ratio metric's Value column read "0.06 1".
+ *
+ * These tests pin:
+ *   - the Value column at human scale, in the unit each row landed on,
+ *   - component columns each in their OWN unit, with no unit in the header
+ *     (auto-scaling means neighbouring rows disagree with any one header),
+ *   - the "- Unit:" line spelled out, and dropped when there is no unit,
+ *   - the formula guard: a formula expression is not a metric name,
+ *   - that the table and the "Filter Conditions Met" sentence in the same
+ *     email render the same sample identically.
+ */
+
+type EvaluatorPrivate = {
+  buildMetricRootCauseContext: (input: {
+    criteriaInstance: MonitorCriteriaInstance;
+    monitor: Monitor;
+    monitorStep?: MonitorStep | undefined;
+  }) => string | null;
+};
+
+const Evaluator: EvaluatorPrivate =
+  MonitorCriteriaEvaluator as unknown as EvaluatorPrivate;
+
+const BREACH_TIME: Date = new Date("2026-08-14T10:30:00.000Z");
+const BREACH_TIME_ISO: string = "2026-08-14T10:30:00.000Z";
+
+function makeContext(
+  overrides: Partial<MetricCriteriaContext> = {},
+): MetricCriteriaContext {
+  return {
+    metricName: "k8s.pod.memory.usage",
+    alias: "a",
+    unit: null,
+    aggregationType: null,
+    isFormula: false,
+    filterAttributes: {},
+    groupBy: [],
+    breachingSamples: [
+      {
+        value: 1073741824,
+        timestamp: BREACH_TIME,
+        attributes: {},
+      },
+    ],
+    ...overrides,
+  };
+}
+
+function makeCriteriaInstance(
+  ctx: MetricCriteriaContext,
+): MonitorCriteriaInstance {
+  const filter: CriteriaFilter = {
+    checkOn: CheckOn.MetricValue,
+    metricCriteriaContext: ctx,
+  } as CriteriaFilter;
+
+  const instance: MonitorCriteriaInstance = new MonitorCriteriaInstance();
+  instance.data = {
+    monitorStatusId: undefined,
+    filterCondition: FilterCondition.All,
+    filters: [filter],
+    incidents: [],
+    alerts: [],
+    name: "Memory breach",
+    description: "Memory breach",
+    id: ObjectID.generate().toString(),
+  };
+  return instance;
+}
+
+function rootCause(ctx: MetricCriteriaContext): string {
+  const context: string | null = Evaluator.buildMetricRootCauseContext({
+    criteriaInstance: makeCriteriaInstance(ctx),
+    monitor: new Monitor(),
+  });
+
+  expect(context).not.toBeNull();
+  return context as string;
+}
+
+/** The one data row of the Breaching Samples table. */
+function sampleRow(context: string): string {
+  const row: string | undefined = context.split("\n").find((line: string) => {
+    return line.includes(BREACH_TIME_ISO);
+  });
+
+  expect(row).toBeDefined();
+  return row as string;
+}
+
+/** The header row of the Breaching Samples table. */
+function headerRow(context: string): string {
+  const row: string | undefined = context.split("\n").find((line: string) => {
+    return line.startsWith("| Timestamp |");
+  });
+
+  expect(row).toBeDefined();
+  return row as string;
+}
+
+describe("MonitorCriteriaEvaluator - Breaching Samples value column", () => {
+  test("renders bytes at human scale instead of raw digits", () => {
+    const context: string = rootCause(makeContext({ unit: "By" }));
+
+    expect(sampleRow(context)).toBe(
+      `| \`${BREACH_TIME_ISO}\` | \`k8s.pod.memory.usage\` | \`a\` | 1.07 GB |`,
+    );
+    expect(context).not.toContain("1073741824");
+  });
+
+  test("renders a duration at human scale", () => {
+    const context: string = rootCause(
+      makeContext({
+        metricName: "http.server.duration",
+        unit: "ms",
+        breachingSamples: [
+          { value: 1500, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 1.5 sec |");
+  });
+
+  test("renders a value the user typed a GB threshold against in GB terms", () => {
+    /*
+     * When the criteria carries a thresholdUnit, samples arrive already
+     * converted into it — so ctx.unit is "GB" and the values are counts
+     * of gigabytes. ValueFormatter has no ladder for "GB" on its own;
+     * MetricValueFormatter restates the value in bytes first.
+     */
+    const context: string = rootCause(
+      makeContext({
+        unit: "GB",
+        breachingSamples: [
+          { value: 2500, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 2.5 TB |");
+  });
+
+  test("each row names the unit IT landed on", () => {
+    const context: string = rootCause(
+      makeContext({
+        unit: "By",
+        breachingSamples: [
+          {
+            value: 921600,
+            timestamp: new Date("2026-08-14T10:29:00.000Z"),
+            attributes: {},
+          },
+          { value: 1258291, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    /*
+     * 922 KB and 1.26 MB in the same table. This is exactly why the unit
+     * rides each cell instead of the column header — no single header
+     * could describe both rows.
+     */
+    expect(context).toContain("| 922 KB |");
+    expect(context).toContain("| 1.26 MB |");
+  });
+
+  test("a ratio metric's fraction is shown as the percentage it means", () => {
+    const context: string = rootCause(
+      makeContext({
+        metricName: "system.cpu.utilization",
+        unit: "1",
+        breachingSamples: [
+          { value: 0.0585, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 5.85% |");
+    // The old rendering: a bare "0.06", which reads as 0.06%.
+    expect(context).not.toContain("| 0.06 |");
+    expect(context).not.toContain("0.06 1");
+  });
+
+  test("a UCUM annotation-only unit is dropped rather than printed", () => {
+    const context: string = rootCause(
+      makeContext({
+        metricName: "process.threads",
+        unit: "{thread}",
+        breachingSamples: [
+          { value: 512, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 512 |");
+    expect(context).not.toContain("{thread}");
+  });
+
+  /*
+   * BACKWARD COMPATIBILITY. A metric with no unit must keep every digit —
+   * a counter at 5000 is not "5K" in an alert, because the reader may be
+   * scanning the table for the exact value that crossed the threshold.
+   */
+  test("a unitless value keeps its exact digits", () => {
+    const context: string = rootCause(
+      makeContext({
+        metricName: "http.server.request.count",
+        unit: null,
+        breachingSamples: [
+          { value: 5000, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 5000 |");
+    expect(context).not.toContain("5K");
+  });
+
+  test("a unitless non-integer is rounded to two decimals", () => {
+    const context: string = rootCause(
+      makeContext({
+        unit: null,
+        breachingSamples: [
+          { value: 2.3333333333333335, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 2.33 |");
+  });
+
+  /*
+   * ValueFormatter renders Infinity as "InfinityP PB" — its
+   * formatLargeNumber divides by 1e15, gets Infinity back and appends the
+   * "P". A broken metric must not produce that in an on-call email.
+   */
+  test("a non-finite sample is reported verbatim", () => {
+    const context: string = rootCause(
+      makeContext({
+        unit: "By",
+        breachingSamples: [
+          { value: Infinity, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| Infinity |");
+    expect(context).not.toContain("InfinityP");
+  });
+
+  test("attribute columns are unaffected", () => {
+    const context: string = rootCause(
+      makeContext({
+        unit: "By",
+        groupBy: ["k8s.pod.name"],
+        breachingSamples: [
+          {
+            value: 1073741824,
+            timestamp: BREACH_TIME,
+            attributes: { "k8s.pod.name": "web-1" },
+          },
+        ],
+      }),
+    );
+
+    expect(headerRow(context)).toBe(
+      "| Timestamp | Metric | Alias | Value | k8s.pod.name |",
+    );
+    expect(sampleRow(context)).toContain("| web-1 |");
+  });
+});
+
+describe("MonitorCriteriaEvaluator - Breaching Samples component columns", () => {
+  function formulaContext(): MetricCriteriaContext {
+    return makeContext({
+      metricName: "a + b",
+      alias: "c",
+      isFormula: true,
+      formulaExpression: "a + b",
+      unit: "By",
+      components: [
+        {
+          alias: "a",
+          name: "container.memory.usage",
+          unit: "By",
+          isFormula: false,
+        },
+        {
+          alias: "b",
+          name: "container.memory.cache",
+          unit: "ms",
+          isFormula: false,
+        },
+      ],
+      breachingSamples: [
+        {
+          value: 1073741824,
+          timestamp: BREACH_TIME,
+          attributes: {},
+          componentValues: [
+            { alias: "a", value: 536870912 },
+            { alias: "b", value: 1500 },
+          ],
+        },
+      ],
+    });
+  }
+
+  test("each component cell is formatted in its own unit", () => {
+    const context: string = rootCause(formulaContext());
+
+    expect(sampleRow(context)).toBe(
+      `| \`${BREACH_TIME_ISO}\` | \`a + b\` | \`c\` | 1.07 GB | 537 MB | 1.5 sec |`,
+    );
+  });
+
+  /*
+   * The header used to read "a (By)". Now that a cell says which scale it
+   * landed on, a header unit is a claim two rows of the same column can
+   * contradict — and the Components bullet list above the table still
+   * records the configured unit, so nothing is lost.
+   */
+  test("component headers carry the alias only, not the configured unit", () => {
+    const context: string = rootCause(formulaContext());
+
+    expect(headerRow(context)).toBe(
+      "| Timestamp | Metric | Alias | Value | a | b |",
+    );
+    expect(context).not.toContain("| a (By) |");
+  });
+
+  test("the Components bullet list spells each component's unit out", () => {
+    const context: string = rootCause(formulaContext());
+
+    expect(context).toContain("- `a` = `container.memory.usage` — unit: Bytes");
+    expect(context).toContain(
+      "- `b` = `container.memory.cache` — unit: Milliseconds",
+    );
+  });
+
+  test("a component with no value for this sample still renders a dash", () => {
+    const ctx: MetricCriteriaContext = formulaContext();
+    ctx.breachingSamples![0]!.componentValues = [
+      { alias: "a", value: 536870912 },
+      { alias: "b", value: null },
+    ];
+
+    expect(sampleRow(rootCause(ctx))).toContain("| 537 MB | - |");
+  });
+});
+
+describe("MonitorCriteriaEvaluator - the Unit line", () => {
+  test("spells the UCUM code out", () => {
+    expect(rootCause(makeContext({ unit: "By" }))).toContain("- Unit: Bytes");
+  });
+
+  test("names the percent a fraction metric carries", () => {
+    const context: string = rootCause(
+      makeContext({ metricName: "system.cpu.utilization", unit: "1" }),
+    );
+
+    expect(context).toContain("- Unit: Percent");
+  });
+
+  /*
+   * The line used to read "- Unit: 1" for every ratio metric — a unit
+   * that names nothing, printed as if it did.
+   */
+  test("is dropped entirely when the unit names no dimension", () => {
+    expect(rootCause(makeContext({ unit: "1" }))).not.toContain("- Unit:");
+    expect(rootCause(makeContext({ unit: "{thread}" }))).not.toContain(
+      "- Unit:",
+    );
+    expect(rootCause(makeContext({ unit: null }))).not.toContain("- Unit:");
+  });
+});
+
+/*
+ * THE FORMULA GUARD.
+ *
+ * MetricCriteriaContext.metricName is the FORMULA EXPRESSION when the
+ * criteria targets a formula. The only use a name has downstream is
+ * ValueFormatter.isFractionMetric, whose regex asks whether the name ends
+ * in `.utilization` / `_ratio` / `_percent`. A formula written as
+ * `a / b_ratio` ends in `_ratio`, so feeding it through would silently
+ * report every value at 100× its real magnitude.
+ */
+describe("MonitorCriteriaEvaluator - formula names are not metric names", () => {
+  test("a formula ending in _ratio is not multiplied by 100", () => {
+    const context: string = rootCause(
+      makeContext({
+        metricName: "a / b_ratio",
+        alias: "c",
+        isFormula: true,
+        formulaExpression: "a / b_ratio",
+        unit: "1",
+        breachingSamples: [
+          { value: 0.42, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 0.42 |");
+    expect(context).not.toContain("42.00%");
+    expect(context).not.toContain("- Unit:");
+  });
+
+  test("a formula COMPONENT ending in _ratio is not multiplied by 100 either", () => {
+    const context: string = rootCause(
+      makeContext({
+        metricName: "c",
+        alias: "c",
+        isFormula: true,
+        formulaExpression: "x",
+        unit: null,
+        components: [
+          {
+            alias: "x",
+            name: "p / q_utilization",
+            unit: "1",
+            isFormula: true,
+          },
+        ],
+        breachingSamples: [
+          {
+            value: 1,
+            timestamp: BREACH_TIME,
+            attributes: {},
+            componentValues: [{ alias: "x", value: 0.42 }],
+          },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 0.42 |");
+    expect(context).not.toContain("42.00%");
+  });
+
+  /*
+   * A plain metric criteria DOES get the heuristic — this is the control
+   * that shows the guard is narrow rather than a blanket opt-out.
+   */
+  test("a plain metric criteria still gets the fraction heuristic", () => {
+    const context: string = rootCause(
+      makeContext({
+        metricName: "system.memory.utilization",
+        isFormula: false,
+        unit: "1",
+        breachingSamples: [
+          { value: 0.42, timestamp: BREACH_TIME, attributes: {} },
+        ],
+      }),
+    );
+
+    expect(sampleRow(context)).toContain("| 42.00% |");
+  });
+});
+
+/*
+ * ONE EMAIL, ONE ANSWER.
+ *
+ * The unit-suppression rule was hand-rolled three separate times before
+ * this change — in CompareCriteria's "Filter Conditions Met" sentence, in
+ * MonitorCriteriaObservationBuilder, and nowhere at all in the Breaching
+ * Samples table — with the result that a single evaluation read
+ * differently in its own two halves: "0.31" in the sentence and "0.31 1"
+ * in the table directly beneath it.
+ *
+ * Both sections of one email now go through MetricValueFormatter. These
+ * tests assert the two strings are IDENTICAL rather than merely both
+ * plausible, so a fourth copy of the rule cannot quietly appear.
+ */
+describe("MonitorCriteriaEvaluator - the table and the sentence agree", () => {
+  function bothRenderings(input: {
+    value: number;
+    unit: string | null;
+    metricName: string;
+  }): { cell: string; sentence: string } {
+    const ctx: MetricCriteriaContext = makeContext({
+      metricName: input.metricName,
+      unit: input.unit,
+      breachingSamples: [
+        { value: input.value, timestamp: BREACH_TIME, attributes: {} },
+      ],
+    });
+
+    const row: string = sampleRow(rootCause(ctx));
+    const cells: Array<string> = row.split("|").map((cell: string) => {
+      return cell.trim();
+    });
+
+    return {
+      // | ts | metric | alias | VALUE | → index 4 once the leading "" is counted.
+      cell: cells[4] as string,
+      sentence: CompareCriteria.getCompareMessage({
+        values: input.value,
+        threshold: input.value,
+        criteriaFilter: {
+          checkOn: CheckOn.MetricValue,
+          filterType: FilterType.GreaterThan,
+        } as CriteriaFilter,
+        metricDisplayName: input.metricName,
+        metricName: input.metricName,
+        unit: input.unit || undefined,
+      }),
+    };
+  }
+
+  const cases: Array<{
+    value: number;
+    unit: string | null;
+    metricName: string;
+  }> = [
+    { value: 1073741824, unit: "By", metricName: "k8s.pod.memory.usage" },
+    { value: 1500, unit: "ms", metricName: "http.server.duration" },
+    { value: 0.0585, unit: "1", metricName: "system.cpu.utilization" },
+    { value: 0.31, unit: "1", metricName: "browser.cls" },
+    { value: 5000, unit: null, metricName: "http.server.request.count" },
+    { value: 2500, unit: "GB", metricName: "system.filesystem.usage" },
+  ];
+
+  for (const testCase of cases) {
+    test(`${testCase.metricName} @ ${testCase.unit ?? "no unit"} reads the same in both`, () => {
+      const rendered: { cell: string; sentence: string } =
+        bothRenderings(testCase);
+
+      expect(rendered.cell.length).toBeGreaterThan(0);
+      expect(rendered.sentence).toContain(`is ${rendered.cell} which is`);
+      expect(rendered.sentence).toContain(`greater than ${rendered.cell}.`);
+    });
+  }
+});

package/Tests/Server/Utils/Monitor/MonitorCriteriaExpectationBuilderUnits.test.ts

@@ -517,7 +517,18 @@ describe("MonitorCriteriaExpectationBuilder.describeCriteriaExpectation", () =>
    * template string, so a negative sign and any decimals survive verbatim
    * (no toFixed / number formatting), and the unit suffix still follows.
    */
-  describe("negative thresholds with a unit are interpolated raw", () => {
+  /*
+   * A threshold carrying a unit is rendered by the same formatter as the
+   * observed value beside it, so both halves of one sentence describe the
+   * quantity the same way. That means it can be RESCALED — "-0.5 sec" is
+   * written "-500 ms", the same number at the scale a reader reads it at,
+   * and the observation in that sentence uses milliseconds too.
+   *
+   * What must never happen is decimal PADDING: a threshold the user typed
+   * as "-5" must not come back as "-5.00", which is what the old raw
+   * interpolation was protecting against and what these still assert.
+   */
+  describe("negative thresholds with a unit are formatted, never padded", () => {
     test("GreaterThan '-5' with { unit: 'sec' } => 'to be greater than -5 sec'", () => {
       const filter: CriteriaFilter = makeFilter({
         filterType: FilterType.GreaterThan,
@@ -530,11 +541,11 @@ describe("MonitorCriteriaExpectationBuilder.describeCriteriaExpectation", () =>
         });
 
       expect(result).toBe("to be greater than -5 sec");
-      // Raw interpolation — never number-formatted.
+      // Integers keep their digits — never padded to "-5.00".
       expect(result).not.toContain("-5.00");
     });
 
-    test("LessThan '-0.5' with { unit: 'sec' } => 'to be less than -0.5 sec'", () => {
+    test("LessThan '-0.5' with { unit: 'sec' } rescales to milliseconds", () => {
       const filter: CriteriaFilter = makeFilter({
         filterType: FilterType.LessThan,
         value: "-0.5",
@@ -545,9 +556,23 @@ describe("MonitorCriteriaExpectationBuilder.describeCriteriaExpectation", () =>
           unit: "sec",
         });
 
-      expect(result).toBe("to be less than -0.5 sec");
-      // Raw interpolation — the decimal is kept as-is, not padded to "-0.50".
+      expect(result).toBe("to be less than -500 ms");
       expect(result).not.toContain("-0.50");
     });
+
+    /*
+     * The boundary the formatter is gated on: with no unit resolved — every
+     * non-metric checkOn — the threshold is interpolated exactly as before.
+     */
+    test("a threshold with no unit is still interpolated raw", () => {
+      const filter: CriteriaFilter = makeFilter({
+        filterType: FilterType.LessThan,
+        value: "-0.5",
+      });
+
+      expect(
+        MonitorCriteriaExpectationBuilder.describeCriteriaExpectation(filter),
+      ).toBe("to be less than -0.5");
+    });
   });
 });