printer: bound the Win32_Printer query by the scrape timeout (#2569)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Jan-Otto Kröpke <mail@jkroepke.de>

internal/collector/printer/printer.go

@@ -184,23 +184,38 @@ type wmiPrintJob struct {
 	Status string `mi:"Status"`
 }
 
+// minQueryTimeout keeps the job query bounded once the scrape budget is spent.
+// MI counts timeouts in whole milliseconds, and a zero timeout means no timeout at all.
+const minQueryTimeout = time.Millisecond
+
 func (c *Collector) Collect(ch chan<- prometheus.Metric, maxScrapeDuration time.Duration) error {
 	var errs []error
 
-	if err := c.collectPrinterStatus(ch); err != nil {
+	deadline := time.Now().Add(maxScrapeDuration)
+
+	// Win32_Printer calls DeviceCapabilities for every printer, whichever properties are selected.
+	// For an IPP printer that is offline, this blocks until the network requests time out,
+	// which can take close to a minute. Bound the query by the scrape timeout, so that a hung call
+	// fails with the native MI timeout instead of blocking the collector for later scrapes.
+	if err := c.collectPrinterStatus(ch, maxScrapeDuration); err != nil {
 		errs = append(errs, fmt.Errorf("failed to collect printer status metrics: %w", err))
 	}
 
-	if err := c.collectPrinterJobStatus(ch, maxScrapeDuration); err != nil {
+	jobQueryTimeout := maxScrapeDuration
+	if maxScrapeDuration > 0 {
+		jobQueryTimeout = max(time.Until(deadline), minQueryTimeout)
+	}
+
+	if err := c.collectPrinterJobStatus(ch, jobQueryTimeout); err != nil {
 		errs = append(errs, fmt.Errorf("failed to collect printer job status metrics: %w", err))
 	}
 
 	return errors.Join(errs...)
 }
 
-func (c *Collector) collectPrinterStatus(ch chan<- prometheus.Metric) error {
+func (c *Collector) collectPrinterStatus(ch chan<- prometheus.Metric, maxScrapeDuration time.Duration) error {
 	var printers []wmiPrinter
-	if err := c.miSession.Query(&printers, mi.NamespaceRootCIMv2, c.miQueryPrinter, 0); err != nil {
+	if err := c.miSession.Query(&printers, mi.NamespaceRootCIMv2, c.miQueryPrinter, maxScrapeDuration); err != nil {
 		return fmt.Errorf("WMI query failed: %w", err)
 	}
 

internal/collector/printer/printer_test.go

@@ -18,12 +18,17 @@
 package printer_test
 
 import (
+	"log/slog"
 	"testing"
+	"time"
 
 	"github.com/alecthomas/kingpin/v2"
 	"github.com/prometheus-community/windows_exporter/internal/collector/printer"
+	"github.com/prometheus-community/windows_exporter/internal/mi"
 	"github.com/prometheus-community/windows_exporter/internal/utils/testutils"
 	"github.com/prometheus/client_golang/prometheus"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func BenchmarkCollector(b *testing.B) {
@@ -39,3 +44,29 @@ func TestCollector(t *testing.T) {
 	metrics := testutils.TestCollector(t, printer.New, nil)
 	testutils.RequireFixtureMetric(t, metrics, printer.Name, "windows_printer_job_count", prometheus.Labels{"printer": "CIPrinter"})
 }
+
+// TestCollectorScrapeTimeout checks that the Win32_Printer query is bounded by the scrape timeout.
+// Win32_Printer can block for close to a minute on an offline IPP printer; without a timeout,
+// that call keeps the collector busy across scrapes.
+func TestCollectorScrapeTimeout(t *testing.T) {
+	miApp, err := mi.ApplicationInitialize()
+	require.NoError(t, err)
+	t.Cleanup(func() { assert.NoError(t, miApp.Close()) })
+
+	miSession, err := miApp.NewSession(nil)
+	require.NoError(t, err)
+	t.Cleanup(func() { assert.NoError(t, miSession.Close()) })
+
+	c := printer.New(nil)
+
+	t.Cleanup(func() { assert.NoError(t, c.Close()) })
+
+	require.NoError(t, c.Build(slog.New(slog.DiscardHandler), miSession))
+
+	ch := make(chan prometheus.Metric, 1000)
+
+	// 1ms is the smallest timeout MI enforces, and well below the time Win32_Printer needs.
+	err = c.Collect(ch, time.Millisecond)
+	require.ErrorIs(t, err, mi.MI_RESULT_INVALID_OPERATION_TIMEOUT)
+	require.ErrorContains(t, err, "failed to collect printer status metrics")
+}