Preserve independent ARC and stop cancelled CORE reporting before I/O

Complete the native CORE #2077 repair after its whole-suite adverse result: retain ARC without inventing free RAM, enforce live-window freshness even with coarse RRD steps, and assert bounded graph splitting including CPU temperature. Keep empty/zero, aligned ARC and transport boundaries intact.

Change-source: pulse-maintainer

docs/release-control/v6/internal/subsystems/monitoring.md

@@ -291,13 +291,16 @@ row timestamps and supported object aliases retain compatibility. A metric is
 never a timestamp, a null is never zero and a window mean is never a replacement
 for a missing current row. History retains measured buckets; native current
 readings accept only the last two steps before the requested end, preserving
-sample time rather than stamping an old value as newly measured. The five-state
+sample time rather than stamping an old value as newly measured. This allowance
+never exceeds the live query window, even with a coarse returned step. The five-state
 CORE CPU vector is normalized by its sum (RRD state rates need not sum to 100).
 All-zero CPU or complete memory-class vectors are empty buckets, not 100% usage;
 explicit zero usage, all-idle CPU, zero-free RAM with used pages and zero I/O
 remain measured zeros. Free memory comes from `memory-free_value`, not active
 pages or the sum of memory classes. ARC subtraction uses the matching free-RAM
-bucket. Reporting I/O values remain rates in the existing bytes/s contract;
+bucket when free RAM is present; independently reported ARC remains available
+without claiming measured memory usage. Pre-cancelled collection stops before
+reading the catalogue or posting a reporting query. Reporting I/O values remain rates in the existing bytes/s contract;
 step supplies time, not a second rate division; `overlap` is not extra traffic.
 
 Legacy REST reads `/reporting/graphs` once per live/History collection and sends

internal/truenas/client.go

@@ -342,6 +342,14 @@ func systemInfoFromMetricHistory(history *SystemMetricHistory) *SystemInfo {
 			latestFree = free.Timestamp
 		}
 	}
+	// ARC can be observed independently when no free-RAM sample exists;
+	// retain it without implying known memory usage. With free RAM, require
+	// cache and free to describe the same bucket before deriving usage.
+	if latestFree.IsZero() {
+		if value, ok := latestTimeSeriesValue(history.ARCSizeBytes); ok {
+			system.ARCSizeBytes = int64(value)
+		}
+	}
 	for _, arc := range history.ARCSizeBytes {
 		if arc.Timestamp.Equal(latestFree) {
 			system.ARCSizeBytes = int64(arc.Value)

internal/truenas/client_test.go

@@ -2195,8 +2195,8 @@ func TestRESTReportingGraphFailurePreservesSnapshotTelemetry(t *testing.T) {
 			t.Fatal("inventory lost")
 		}
 	}
-	if transport.calls != 12 {
-		t.Fatalf("reporting calls = %d, want bounded batch + five graphs per snapshot", transport.calls)
+	if transport.calls != 2*(1+len(legacyRESTReportingGraphs())) {
+		t.Fatalf("reporting calls = %d, want bounded batch + selected graphs per snapshot", transport.calls)
 	}
 	history, err := client.GetSystemMetricHistory(context.Background(), time.Hour)
 	if err != nil {
@@ -2464,7 +2464,7 @@ func TestRESTSnapshotRetainsSanitizedTelemetryFailure(t *testing.T) {
 			}
 			wantCalls := 1
 			if tc.status == 422 {
-				wantCalls = 6
+				wantCalls = 1 + len(legacyRESTReportingGraphs())
 			}
 			if transport.calls != wantCalls {
 				t.Fatalf("reporting calls=%d, want %d", transport.calls, wantCalls)

internal/truenas/core_reporting.go

@@ -16,6 +16,9 @@ import (
 // Older REST bridges without the catalogue retain the unscoped compatibility
 // request. Authentication, transport and malformed replies are never retried.
 func (c *Client) legacyReportingGraphs(ctx context.Context) ([]map[string]any, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
 	var catalogue []struct {
 		Name        string   `json:"name"`
 		Identifiers []string `json:"identifiers"`
@@ -97,7 +100,9 @@ func reportingRow(response trueNASReportingGetDataResponse, index int) (time.Tim
 
 // Keep native History intact, but do not label an old RRD point as a current
 // reading. Two returned steps allow the usual unfinished/null RRD bucket;
-// older points stay available only through History. Preserve row indices.
+// older points stay available only through History. Never exceed the live
+// query window even if a coarse/malformed step allows older buckets. Preserve
+// row indices.
 func liveReportingResponses(responses []trueNASReportingGetDataResponse, end int64) []trueNASReportingGetDataResponse {
 	live := append([]trueNASReportingGetDataResponse(nil), responses...)
 	for i, response := range live {
@@ -108,7 +113,7 @@ func liveReportingResponses(responses []trueNASReportingGetDataResponse, end int
 		for index := range response.Data {
 			timestamp, _, ok := reportingRow(response, index)
 			gap := end - timestamp.Unix()
-			if !ok || timestamp.Unix() > end || (gap > response.Step && gap-response.Step > response.Step) {
+			if !ok || timestamp.Unix() > end || gap > legacyRESTTelemetryWindowSeconds || (gap > response.Step && gap-response.Step > response.Step) {
 				live[i].Data[index] = nil
 			}
 		}

internal/truenas/core_reporting_test.go

@@ -417,6 +417,28 @@ func TestCORE13ARCAlignedWithFreeMemory(t *testing.T) {
 	assertCOREPoint(t, points, 0, 1789000000, 0)
 }
 
+func TestCORE13CoarseStepDoesNotMakeOldRowsCurrent(t *testing.T) {
+	response := trueNASReportingGetDataResponse{Name: "cpu", Legend: []string{"usage"}, Start: 1789000000, End: 1789000000, Step: 3600, Data: []any{[]any{12.}}}
+	if history := parseSystemMetricHistory(liveReportingResponses([]trueNASReportingGetDataResponse{response}, 1789000400)); history != nil {
+		t.Fatal("coarse step made an old CPU sample current")
+	}
+	if history := parseSystemMetricHistory([]trueNASReportingGetDataResponse{response}); history == nil {
+		t.Fatal("live age filtering destroyed native History")
+	}
+}
+
+func TestCORE13ARCOnlyDoesNotInventMemoryUsage(t *testing.T) {
+	history := &SystemMetricHistory{ARCSizeBytes: []TimeSeriesPoint{{Timestamp: time.Unix(1789000000, 0), Value: 8 << 30}}}
+	system := systemInfoFromMetricHistory(history)
+	if system.ARCSizeBytes != 8<<30 || system.Telemetry.Memory {
+		t.Fatal("ARC-only telemetry lost or claimed free RAM")
+	}
+	system.MemoryTotalBytes = 32 << 30
+	if metricsFromTrueNASSystem(*system, 0, 0).Memory != nil {
+		t.Fatal("ARC-only reading invented usage")
+	}
+}
+
 func TestCORE13DiskTemperatureHistoryExternalTiming(t *testing.T) {
 	response := trueNASReportingGetDataResponse{Name: "disktemp", Identifier: "disk-a", Legend: []string{"temperature"}, Start: 1789000000, End: 1789000020, Step: 10, Data: []any{[]any{40.}, []any{nil}, []any{42.}}}
 	points := parseReportingDiskTemperatureHistory([]trueNASReportingGetDataResponse{response})["disk-a"]