gpu: collect reset-required/row-remap status and classify Xid codes by severity

nvidia-smi exposes reset_status.reset_required and remapped_rows.* only on newer drivers for Ampere+ GPUs; queried via a separate exec call since an unrecognized field name fails the whole --query-gpu command and would have wiped out unrelated telemetry (temp/ECC/power) on older drivers otherwise. Also refines Xid severity in both the ingest dmesg collector and the always-on kmsg watcher: Xid 64 (row-remap InfoROM write failure) now escalates to Critical instead of the generic warning every other Xid got, and fixes the SAT-window flush path which previously hardcoded "Warning" and ignored pattern severity entirely. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-07-07 19:28:59 +03:00
parent d850d4fc3a
commit 1175e6ccd8
10 changed files with 477 additions and 69 deletions
@@ -0,0 +1,53 @@
+package collector
+
+import "testing"
+
+func TestXidSeverity(t *testing.T) {
+	tests := []struct {
+		name    string
+		line    string
+		wantSev string
+		wantOK  bool
+	}{
+		{
+			name:    "xid 64 remap write failure is critical",
+			line:    "NVRM: Xid (PCI:0000:65:00): 64, pid=1234, name=python, Row Remapper: New row remapping failed",
+			wantSev: "critical",
+			wantOK:  true,
+		},
+		{
+			name:    "xid 48 uncorrectable ecc is critical",
+			line:    "NVRM: Xid (PCI:0000:65:00): 48, pid=1234, name=python",
+			wantSev: "critical",
+			wantOK:  true,
+		},
+		{
+			name:    "xid 63 remap committed is warning",
+			line:    "NVRM: Xid (PCI:0000:65:00): 63, pid=1234, Class 0x90a0",
+			wantSev: "warning",
+			wantOK:  true,
+		},
+		{
+			name:   "xid 79 unknown code falls back to caller default",
+			line:   "NVRM: Xid (PCI:0000:65:00): 79, pid=1234, GPU has fallen off the bus",
+			wantOK: false,
+		},
+		{
+			name:   "no xid in line",
+			line:   "NVRM: GPU 0000:65:00.0: RmInitAdapter failed",
+			wantOK: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			sev, ok := XidSeverity(tt.line)
+			if ok != tt.wantOK {
+				t.Fatalf("ok: got %v, want %v", ok, tt.wantOK)
+			}
+			if ok && sev != tt.wantSev {
+				t.Fatalf("severity: got %q, want %q", sev, tt.wantSev)
+			}
+		})
+	}
+}