@oneuptime/common 14.0.2 → 14.0.3

package/Tests/App/Dashboard/HostMetricTooltips.test.ts

@@ -0,0 +1,709 @@
+import { describe, expect, test } from "@jest/globals";
+import fs from "fs";
+import path from "path";
+import {
+  HOST_METRIC_DESCRIPTIONS,
+  HostMetric,
+} from "../../../../App/FeatureSet/Dashboard/src/Components/MetricDescriptions/HostMetricDescriptions";
+import {
+  activeStateMeta,
+  computeSystemdAvailability,
+  SystemdAvailability,
+  unitTypeLabel,
+} from "../../../../App/FeatureSet/Dashboard/src/Pages/Host/Utils/SystemdUnits";
+import {
+  startupModeLabel,
+  statusMeta,
+} from "../../../../App/FeatureSet/Dashboard/src/Pages/Host/Utils/WindowsServices";
+import {
+  expectReadableDescriptionRecord,
+  expectTitleExplained,
+  MAX_DESCRIPTION_LENGTH,
+} from "./MetricDescriptionRules";
+import AggregationIntervalUtil from "../../../Types/BaseDatabase/AggregationIntervalUtil";
+
+/*
+ * The (i) texts on the Host pages: the overview, one process, one Windows
+ * service, one systemd unit, and the three lists. Each text has to say what
+ * the page ACTUALLY computes, so beyond the shared readability rules these
+ * pin the facts a customer would otherwise get wrong - which window a tile
+ * averages, what a percentage is a share of, where a count is capped - and
+ * cross-check the labels a text quotes against the helpers that produce the
+ * labels on screen.
+ */
+
+const HOST_VIEW_DIR: string = path.join(
+  __dirname,
+  "..",
+  "..",
+  "..",
+  "..",
+  "App",
+  "FeatureSet",
+  "Dashboard",
+  "src",
+  "Pages",
+  "Host",
+  "View",
+);
+
+function readPage(file: string): string {
+  return fs.readFileSync(path.join(HOST_VIEW_DIR, file), "utf8");
+}
+
+const D: Record<HostMetric, string> = HOST_METRIC_DESCRIPTIONS;
+
+/*
+ * Every title a Host page shows beside an (i), and the text it explains
+ * with. The wiring suite (App/Tests/Dashboard/HostMetricTooltipWiring)
+ * pins that each page really pairs them this way.
+ */
+const TITLED: Array<[string, HostMetric]> = [
+  ["CPU", "cpu"],
+  ["Memory", "memory"],
+  ["Filesystem", "filesystem"],
+  ["Load avg (1m)", "loadAverage"],
+  ["Processes", "processes"],
+  ["Availability", "availabilityChart"],
+  ["CPU", "cpuChart"],
+  ["Memory", "memoryChart"],
+  ["Disk space", "diskSpaceChart"],
+  ["Network", "networkChart"],
+  ["Used / Total", "filesystemUsedTotal"],
+  ["Utilization", "filesystemUtilization"],
+  ["Process Count (cached)", "processCountCached"],
+  ["CPU", "processCpu"],
+  ["Memory (RSS)", "processMemoryRss"],
+  ["Virtual Memory", "processVirtualMemory"],
+  ["Threads", "processThreads"],
+  ["CPU", "processCpuChart"],
+  ["Memory (RSS)", "processMemoryRssChart"],
+  ["Disk I/O", "processDiskIoChart"],
+  ["Current Status", "serviceCurrentStatus"],
+  ["Availability", "serviceAvailability"],
+  ["Startup Mode", "serviceStartupMode"],
+  ["State Changes", "serviceStateChanges"],
+  ["Status timeline", "serviceStatusTimeline"],
+  ["Current State", "unitCurrentState"],
+  ["Availability", "unitAvailability"],
+  ["Unit Type", "unitType"],
+  ["State Changes", "unitStateChanges"],
+  ["State timeline", "unitStateTimeline"],
+  ["Resources", "hostListResources"],
+  ["CPU", "processListCpu"],
+  ["Memory", "processListMemory"],
+  ["Status", "serviceListStatus"],
+  ["Startup", "serviceListStartup"],
+  ["State", "unitListState"],
+  ["Type", "unitListType"],
+];
+
+const FIVE_MINUTE_TILES: Array<HostMetric> = [
+  "cpu",
+  "memory",
+  "processCpu",
+  "processMemoryRss",
+  "processVirtualMemory",
+  "processThreads",
+];
+
+const PER_INTERVAL_CHARTS: Array<HostMetric> = [
+  "cpuChart",
+  "memoryChart",
+  "diskSpaceChart",
+  "processCpuChart",
+  "processMemoryRssChart",
+  "processDiskIoChart",
+];
+
+const LAST_15_MINUTE_LISTS: Array<HostMetric> = [
+  "processListCpu",
+  "processListMemory",
+  "serviceListStatus",
+  "unitListState",
+];
+
+const SPLIT_CPU_READINGS: Array<HostMetric> = [
+  "processCpu",
+  "processCpuChart",
+  "processListCpu",
+];
+
+// Word-boundary matchers, hoisted: eslint's wrap-regex fights prettier.
+const RAW_METRIC_NAME: RegExp = /\b(system|process|windows|systemd)\.[a-z_]+/;
+const PERCENTILE_TOKEN: RegExp = /\bp\d{2}\b/i;
+const RATE_UNIT: RegExp = /bytes per second/i;
+const WHOLE_RANGE: RegExp = /whole[- ]range/;
+const LOGICAL_CORES_FIRST: RegExp = /^Logical CPU cores/;
+const RANGE_OR_AVERAGE: RegExp = /selected range|time range|averag/i;
+
+describe("HOST_METRIC_DESCRIPTIONS reads well", () => {
+  test("every text passes the shared rules and none repeats another", () => {
+    expectReadableDescriptionRecord(D, "HOST_METRIC_DESCRIPTIONS");
+  });
+
+  test("every text stays small - at most 260 characters and two sentences", () => {
+    for (const [key, text] of Object.entries(D)) {
+      expect({ key, length: text.length }).toEqual({
+        key,
+        length: Math.min(text.length, 260),
+      });
+      expect(text.length).toBeLessThanOrEqual(MAX_DESCRIPTION_LENGTH);
+
+      const sentences: number = text
+        .split(/[.!?](?:\s|$)/)
+        .filter((part: string): boolean => {
+          return part.trim().length > 0;
+        }).length;
+
+      expect({ key, sentences: sentences <= 2 }).toEqual({
+        key,
+        sentences: true,
+      });
+    }
+  });
+
+  test.each(TITLED)(
+    "%s (%s) is explained by its text",
+    (title: string, key: HostMetric) => {
+      expectTitleExplained(title, D[key]);
+    },
+  );
+
+  test("the title table covers every key exactly once", () => {
+    const keys: Array<string> = TITLED.map(
+      (entry: [string, HostMetric]): string => {
+        return entry[1];
+      },
+    );
+
+    expect([...keys].sort()).toEqual(Object.keys(D).sort());
+    expect(new Set(keys).size).toBe(keys.length);
+  });
+
+  test("no text names a percentile it would then have to explain", () => {
+    for (const text of Object.values(D)) {
+      expect(text).not.toMatch(PERCENTILE_TOKEN);
+    }
+  });
+
+  test("raw metric names appear only where the customer needs them to act", () => {
+    const naming: Array<string> = Object.entries(D)
+      .filter((entry: [string, string]): boolean => {
+        return RAW_METRIC_NAME.test(entry[1]);
+      })
+      .map((entry: [string, string]): string => {
+        return entry[0];
+      });
+
+    // Threads and open fds are off by default in the collector.
+    expect(naming).toEqual(["processThreads"]);
+    expect(D.processThreads).toContain("process.threads");
+    expect(D.processThreads).toContain("process.open_file_descriptors");
+    expect(D.processThreads).toContain("turned on");
+  });
+
+  test("the one text shown as a visible caption is short", () => {
+    /*
+     * Detail has no tooltip slot, so "Process Count (cached)" shows its
+     * text under the label - it has to stay a caption, not a paragraph.
+     */
+    expect(D.processCountCached.length).toBeLessThanOrEqual(200);
+  });
+});
+
+describe("time windows are the ones the pages use", () => {
+  test.each(FIVE_MINUTE_TILES)(
+    "%s says it averages the last 5 minutes of the range",
+    (key: HostMetric) => {
+      expect(D[key]).toContain("averaged over the last 5 minutes of the range");
+    },
+  );
+
+  test("overview CPU and memory admit the whole-range fallback on wide ranges", () => {
+    /*
+     * meanFromBuckets keeps buckets that START in the last 5 minutes; past
+     * 12 hours a bucket is 15 minutes or wider, so usually none does and the
+     * tile averages everything.
+     */
+    for (const key of ["cpu", "memory"] as Array<HostMetric>) {
+      expect(D[key]).toContain("12 hours");
+      expect(D[key]).toContain("whole-range average");
+    }
+  });
+
+  test.each(FIVE_MINUTE_TILES)(
+    "%s admits the whole-range fallback when nothing recent arrived",
+    (key: HostMetric) => {
+      /*
+       * The same fallback fires when the last 5 minutes hold no bucket at
+       * all - a host gone silent, a process that exited - so the tile shows
+       * an old average instead of a dash. The text has to say so.
+       */
+      expect(D[key]).toContain("12 hours");
+      expect(D[key]).toContain("no recent data");
+      expect(D[key]).toMatch(WHOLE_RANGE);
+    },
+  );
+
+  test.each(FIVE_MINUTE_TILES)(
+    "%s says the wide-range fallback is usual, not certain",
+    (key: HostMetric) => {
+      /*
+       * On a range over 12 hours a bucket STARTS inside the 5-minute tile
+       * window only when a bucket boundary falls in it: with 15-minute
+       * buckets that is a third of the time. So the tile is USUALLY - not
+       * always - the whole-range average there.
+       */
+      expect(D[key]).toContain("usually");
+    },
+  );
+
+  test("past 12 hours a bucket is at least three tile windows wide", () => {
+    const end: Date = new Date("2026-09-24T12:00:00.000Z");
+    const tileWindowMs: number = 5 * 60_000;
+
+    const bucketMsFor: (hours: number) => number = (hours: number): number => {
+      return AggregationIntervalUtil.getAggregationIntervalMs(
+        AggregationIntervalUtil.getAggregationIntervalForWindow({
+          startDate: new Date(end.getTime() - hours * 3_600_000),
+          endDate: end,
+        }),
+      );
+    };
+
+    // Up to 12 hours a bucket is no wider than the tile window...
+    expect(bucketMsFor(3)).toBeLessThanOrEqual(tileWindowMs);
+    expect(bucketMsFor(12)).toBeLessThanOrEqual(tileWindowMs);
+    // ...past it, a bucket starts in the window a third of the time at most.
+    for (const hours of [13, 24, 72, 24 * 7, 24 * 30]) {
+      expect(bucketMsFor(hours)).toBeGreaterThanOrEqual(3 * tileWindowMs);
+    }
+  });
+
+  test("the pages really fall back to the whole range when no bucket is recent", () => {
+    for (const page of ["Overview.tsx", "ProcessView.tsx"]) {
+      const source: string = readPage(page);
+
+      expect(source).toContain("t < tileWindowStartMs");
+      expect(source).toContain("if (count === 0) {");
+    }
+  });
+
+  test("the filesystem tile says it is a whole-range average, not a 5-minute one", () => {
+    expect(D.filesystem).toContain("whole selected range");
+    expect(D.filesystem).toContain("not the last 5 minutes");
+    expect(D.filesystemUsedTotal).toContain("averaged over the selected range");
+    expect(D.filesystemUtilization).toContain(
+      "averaged over the selected range",
+    );
+  });
+
+  test("load average and processes are the newest interval, not a mean", () => {
+    expect(D.loadAverage).toContain("newest interval");
+    expect(D.processes).toContain("newest interval");
+    expect(D.loadAverage).not.toContain("5 minutes");
+  });
+
+  test.each(PER_INTERVAL_CHARTS)(
+    "%s describes one value per interval",
+    (key: HostMetric) => {
+      expect(D[key]).toContain("each interval");
+    },
+  );
+
+  test.each(LAST_15_MINUTE_LISTS)(
+    "%s says the list ignores the time picker and reads the last 15 minutes",
+    (key: HostMetric) => {
+      expect(D[key]).toContain("last 15 minutes");
+      expect(D[key]).not.toContain("selected range");
+    },
+  );
+
+  test("the lists really look back 15 minutes", () => {
+    expect(readPage("Processes.tsx")).toContain(
+      "const PROCESS_LOOKBACK_MINUTES: number = 15;",
+    );
+    expect(readPage("Services.tsx")).toContain(
+      "const SERVICE_LOOKBACK_MINUTES: number = 15;",
+    );
+    expect(readPage("SystemdUnits.tsx")).toContain(
+      "const UNIT_LOOKBACK_MINUTES: number = 15;",
+    );
+  });
+
+  test("the overview and process tiles really use a 5-minute window", () => {
+    for (const page of ["Overview.tsx", "ProcessView.tsx"]) {
+      expect(readPage(page)).toContain(
+        "const TILE_WINDOW_MINUTES: number = 5;",
+      );
+    }
+  });
+});
+
+describe("denominators, units and caps", () => {
+  test("CPU texts say 100% is every core, not one core", () => {
+    expect(D.cpu).toContain("across all cores");
+    expect(D.cpu).toContain("100% means every core was busy");
+    expect(D.cpuChart).toContain("100% means every core was busy");
+    // The tile and the chart both add user and system time - one wording.
+    for (const key of ["cpu", "cpuChart"] as Array<HostMetric>) {
+      expect(D[key]).toContain(
+        "spent running programs and the operating system",
+      );
+    }
+    expect(D.cpuChart).toContain("(user plus system time)");
+    expect(D.cpuChart).not.toContain("busy share");
+    expect(D.processCpu).toContain("share of all host CPU cores");
+    expect(D.processCpuChart).toContain("total CPU capacity");
+    expect(D.processListCpu).toContain("all cores together");
+  });
+
+  test("load average is compared with the core count and can pass 100%", () => {
+    expect(D.loadAverage).toContain("number of cores");
+    expect(D.loadAverage).toContain("over 100% means work is queuing");
+    // The tile's danger threshold is exactly that 100%.
+    expect(readPage("Overview.tsx")).toContain(
+      "thresholds={{ warn: 70, danger: 100 }}",
+    );
+  });
+
+  test("filesystem percentages count reserved space in the total", () => {
+    for (const key of [
+      "filesystem",
+      "diskSpaceChart",
+      "filesystemUsedTotal",
+      "filesystemUtilization",
+    ] as Array<HostMetric>) {
+      expect(D[key].toLowerCase()).toContain("reserved");
+    }
+    expect(D.filesystem).toContain("largest filesystem");
+    expect(D.diskSpaceChart).toContain("same mount as the Filesystem tile");
+  });
+
+  test("the filesystem utilization bar colours match the table", () => {
+    const overview: string = readPage("Overview.tsx");
+
+    expect(D.filesystemUtilization).toContain("amber at 75%");
+    expect(D.filesystemUtilization).toContain("red at 90%");
+    expect(overview).toContain("pct >= 90");
+    expect(overview).toContain("pct >= 75");
+  });
+
+  test("sizes are called out as 1024-based, as formatMemoryBytes renders them", () => {
+    expect(D.filesystemUsedTotal).toContain("1024-based");
+    expect(D.filesystemUsedTotal).toContain("1 GiB is 1,024 MiB");
+    expect(readPage("Overview.tsx")).toContain(
+      'const units: Array<string> = ["B", "KiB", "MiB", "GiB", "TiB"];',
+    );
+  });
+
+  test("process memory is compared with the host's total RAM", () => {
+    expect(D.processMemoryRss).toContain("host's total RAM");
+    expect(D.processListMemory).toContain("host's total RAM");
+  });
+
+  test("the process list memory bar colours match memBarColor", () => {
+    const processes: string = readPage("Processes.tsx");
+
+    expect(D.processListMemory).toContain("amber at 10%");
+    expect(D.processListMemory).toContain("red at 20%");
+    expect(processes).toContain("if (value >= 20) {");
+    expect(processes).toContain("if (value >= 10) {");
+  });
+
+  test("network is bytes per second, summed over every interface, loopback included", () => {
+    expect(D.networkChart).toMatch(RATE_UNIT);
+    expect(D.networkChart).toContain("(In)");
+    expect(D.networkChart).toContain("(Out)");
+    expect(D.networkChart).toContain("every network interface");
+    expect(D.networkChart).toContain("loopback");
+    expect(D.processDiskIoChart).toMatch(RATE_UNIT);
+  });
+
+  test("service and unit availability name the 2,000-sample cap the pages fetch", () => {
+    for (const key of [
+      "serviceAvailability",
+      "unitAvailability",
+    ] as Array<HostMetric>) {
+      expect(D[key]).toContain("newest 2,000 samples");
+      expect(D[key]).toContain("capped");
+    }
+
+    expect(readPage("ServiceView.tsx")).toContain(
+      "const SAMPLE_FETCH_LIMIT: number = 2000;",
+    );
+    expect(readPage("SystemdUnitView.tsx")).toContain(
+      "const SAMPLE_FETCH_LIMIT: number = 2000;",
+    );
+  });
+
+  test("the cached process count says it is not live, and how often it moves", () => {
+    /*
+     * Host.processCount is written by ingest through ResourceHeartbeat, whose
+     * enrichment gate admits one write per 60-second window (plus up to 25%
+     * jitter) - "about once a minute", not "every few minutes".
+     */
+    expect(D.processCountCached).toContain("any state");
+    expect(D.processCountCached).toContain("at most about once a minute");
+    expect(D.processes).toContain("at most about once a minute");
+    expect(D.processes).not.toContain("every few minutes");
+    // The total is the tile's sublabel, printed under the number.
+    expect(D.processes).toContain("the total on the line below");
+  });
+
+  test("the cached process count's refresh rate is the host heartbeat window", () => {
+    const hostService: string = fs.readFileSync(
+      path.join(
+        __dirname,
+        "..",
+        "..",
+        "..",
+        "Server",
+        "Services",
+        "HostService.ts",
+      ),
+      "utf8",
+    );
+
+    expect(hostService).toContain(
+      "const LAST_SEEN_THROTTLE_SECONDS: number = 60;",
+    );
+    expect(hostService).toContain("metadata.processCount = extra.processCount");
+    expect(hostService).toContain(
+      "throttleInSeconds: LAST_SEEN_THROTTLE_SECONDS",
+    );
+  });
+
+  test("the Windows fallback of the Processes tile is disclosed", () => {
+    expect(D.processes).toContain("Windows");
+    expect(D.processes).toContain("every process seen in the last 5 minutes");
+  });
+
+  test("the availability chart explains the uptime badge and the unjudged edge", () => {
+    expect(D.availabilityChart).toContain("uptime badge");
+    expect(D.availabilityChart).toContain("Up if");
+    expect(D.availabilityChart).toContain("Down if");
+    expect(D.availabilityChart).toContain("newest interval is not judged");
+  });
+});
+
+describe("honest about how process CPU is read", () => {
+  test.each(SPLIT_CPU_READINGS)(
+    "%s says the collector's user, system and wait readings are not added up",
+    (key: HostMetric) => {
+      expect(D[key]).toContain("user, system and wait");
+    },
+  );
+
+  test("the tile and chart say they read about a third of the real use", () => {
+    expect(D.processCpu).toContain("about a third of the real use");
+    expect(D.processCpuChart).toContain("about a third of the real use");
+  });
+
+  test("the list says it shows only one of the readings", () => {
+    expect(D.processListCpu).toContain("only one of them");
+  });
+
+  test("the pages still read process CPU the way the texts describe", () => {
+    /*
+     * If someone fixes the aggregation (filter or sum the states), these
+     * texts must change with it - this test is the reminder.
+     */
+    const processView: string = readPage("ProcessView.tsx");
+    const processes: string = readPage("Processes.tsx");
+
+    expect(processView).toContain('"process.cpu.utilization",');
+    expect(processView).not.toContain('state: "user"');
+    expect(processes).toContain('buildQuery("process.cpu.utilization")');
+    expect(processes).toContain(
+      "// Skip — first datapoint per key is the most recent (sorted DESC).",
+    );
+  });
+});
+
+describe("state texts quote the labels the pages render", () => {
+  test("Windows status texts use statusMeta's labels", () => {
+    for (const code of [4, 1, 7]) {
+      const label: string = statusMeta(code).label;
+
+      expect(D.serviceCurrentStatus).toContain(label);
+      expect(D.serviceListStatus).toContain(label);
+    }
+    expect(D.serviceAvailability).toContain(statusMeta(4).label);
+  });
+
+  test("Windows status texts cover all four in-progress states", () => {
+    /*
+     * statusMeta has four pending states: start (2), stop (3), continue (5,
+     * resuming a paused service) and pause (6). The texts name each one in
+     * plain words rather than quoting "Continue pending".
+     */
+    expect(
+      [2, 3, 5, 6].map((code: number): string => {
+        return statusMeta(code).label;
+      }),
+    ).toEqual([
+      "Start pending",
+      "Stop pending",
+      "Continue pending",
+      "Pause pending",
+    ]);
+
+    for (const key of [
+      "serviceCurrentStatus",
+      "serviceListStatus",
+    ] as Array<HostMetric>) {
+      expect(D[key]).toContain(
+        "a start, stop, pause or resume still in progress",
+      );
+    }
+  });
+
+  test("startup texts use startupModeLabel's labels", () => {
+    for (const mode of [
+      "auto_start",
+      "demand_start",
+      "disabled",
+      "boot_start",
+      "system_start",
+    ]) {
+      const label: string = startupModeLabel(mode);
+
+      expect(D.serviceStartupMode).toContain(label);
+      expect(D.serviceListStartup).toContain(label);
+    }
+  });
+
+  test("systemd state texts use activeStateMeta's labels", () => {
+    for (const state of ["active", "inactive", "failed", "activating"]) {
+      expect(D.unitCurrentState).toContain(activeStateMeta(state).label);
+    }
+    for (const state of ["active", "failed"]) {
+      expect(D.unitListState).toContain(activeStateMeta(state).label);
+    }
+    expect(D.unitListState).toContain(activeStateMeta(null).label);
+  });
+
+  test("unit type texts use unitTypeLabel's labels", () => {
+    expect(D.unitType).toContain(unitTypeLabel("nginx.service"));
+    expect(D.unitType).toContain(unitTypeLabel("docker.socket"));
+    expect(D.unitType).toContain(unitTypeLabel("logrotate.timer"));
+    expect(D.unitListType).toContain(unitTypeLabel("nginx.service"));
+    expect(D.unitListType).toContain(unitTypeLabel("logrotate.timer"));
+  });
+
+  test("unit availability counts Reloading as not active, as the text says", () => {
+    const availability: SystemdAvailability = computeSystemdAvailability([
+      { time: new Date(0), state: "active" },
+      { time: new Date(30_000), state: "reloading" },
+      { time: new Date(60_000), state: "active" },
+      { time: new Date(90_000), state: "refreshing" },
+    ]);
+
+    expect(availability.percent).toBe(50);
+    expect(D.unitAvailability).toContain("Reloading counts as not active");
+    expect(D.unitAvailability).toContain("Active");
+  });
+
+  test("timelines say they show the worst state, as the Min x-axis does", () => {
+    expect(D.serviceStatusTimeline).toContain("worst status");
+    expect(D.unitStateTimeline).toContain("worst state");
+    expect(readPage("ServiceView.tsx")).toContain(
+      "aggregateType: XAxisAggregateType.Min",
+    );
+    expect(readPage("SystemdUnitView.tsx")).toContain(
+      "aggregateType: XAxisAggregateType.Min",
+    );
+  });
+
+  test("state-change counts admit what sampling cannot see", () => {
+    expect(D.serviceStateChanges).toContain("between two samples is not seen");
+    expect(D.unitStateChanges).toContain("between two samples is not seen");
+  });
+});
+
+describe("the Hosts list Resources column reads cached host columns", () => {
+  /*
+   * The column shows Host.cpuCores, Host.totalMemoryBytes and
+   * Host.processCount - values ingest saves on the host record while it
+   * scans each metrics batch, not telemetry fetched for a range. Each claim
+   * of the text is pinned to the code that fills the column.
+   */
+  const REPO_PACKAGES: string = path.join(__dirname, "..", "..", "..", "..");
+  const ingestBase: string = fs.readFileSync(
+    path.join(
+      REPO_PACKAGES,
+      "App",
+      "FeatureSet",
+      "Telemetry",
+      "Services",
+      "OtelIngestBaseService.ts",
+    ),
+    "utf8",
+  );
+  const hostService: string = fs.readFileSync(
+    path.join(REPO_PACKAGES, "Common", "Server", "Services", "HostService.ts"),
+    "utf8",
+  );
+  const hostDocs: string = fs.readFileSync(
+    path.join(HOST_VIEW_DIR, "..", "Utils", "DocumentationMarkdown.ts"),
+    "utf8",
+  );
+
+  // The branch of scanHostInfraStatsFromMetrics that handles one metric.
+  function scanBranch(metricName: string): string {
+    const start: number = ingestBase.indexOf(`name === "${metricName}"`);
+
+    expect(start).toBeGreaterThan(-1);
+
+    return ingestBase.slice(start, ingestBase.indexOf("continue;", start));
+  }
+
+  test("cores are the logical CPU count, so hyperthreads count", () => {
+    const cores: string = scanBranch("system.cpu.logical.count");
+
+    expect(cores).toContain("this.firstDatapointNumber(m)");
+    expect(cores).toContain("result.cpuCores = Math.round(v);");
+    // The setup guide turns the (off-by-default) logical count on.
+    expect(hostDocs).toContain("system.cpu.logical.count:");
+    expect(D.hostListResources).toMatch(LOGICAL_CORES_FIRST);
+    expect(D.hostListResources).toContain("each hyperthread counts as one");
+  });
+
+  test("total RAM is every memory state added up, and the text says so", () => {
+    const memory: string = scanBranch("system.memory.usage");
+
+    expect(memory).toContain("this.sumDatapointNumbers(m)");
+    expect(memory).toContain("result.totalMemoryBytes = Math.round(v);");
+    expect(D.hostListResources).toContain(
+      "the sum of every memory state the collector reports",
+    );
+  });
+
+  test("processes are summed over every status, as 'in any state' says", () => {
+    const processes: string = scanBranch("system.processes.count");
+
+    expect(processes).toContain("this.sumDatapointNumbers(m)");
+    expect(processes).toContain("result.processCount = Math.round(v);");
+    expect(D.hostListResources).toContain("processes in any state");
+    // The same number as the overview's cached count, in the same words.
+    expect(D.processCountCached).toContain("any state");
+  });
+
+  test("all three ride the host heartbeat, so 'at most about once a minute'", () => {
+    expect(hostService).toContain(
+      "const LAST_SEEN_THROTTLE_SECONDS: number = 60;",
+    );
+    for (const field of ["cpuCores", "totalMemoryBytes", "processCount"]) {
+      expect(hostService).toContain(`metadata.${field} = extra.${field};`);
+    }
+    expect(D.hostListResources).toContain(
+      "latest values saved on the host, updated at most about once a minute",
+    );
+  });
+
+  test("the text claims no time range or average - the list has no time picker", () => {
+    expect(D.hostListResources).not.toMatch(RANGE_OR_AVERAGE);
+  });
+});