sendnrw/windows_exporter
2
0
Fork 0
mirror of https://github.com/prometheus-community/windows_exporter.git synced 2026-10-17 01:59:04 +02:00
Code Issues Packages Projects Releases Wiki Activity

iis: add worker_responses_total as denominator for worker_request_errors_total (#2764)

Browse Source 
Signed-off-by: Jan-Otto Kröpke <mail@jkroepke.de>
This commit is contained in:
Jan-Otto Kröpke
2026-10-16 20:40:25 +02:00
committed by GitHub
parent 15316bfbd5
commit f1e6b49c54
4 changed files with 81 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/collector.iis.md
+17 -2
View File
@@ -116,7 +116,8 @@ See [#2698](https://github.com/prometheus-community/windows_exporter/issues/2698
| `windows_iis_worker_max_threads` | Maximum number of threads to which the thread pool can grow as needed | gauge | `app`, `pid` |
| `windows_iis_worker_requests_total` | Total number of HTTP requests served by the worker process | counter | `app`, `pid` |
| `windows_iis_worker_current_requests` | Current number of requests being processed by the worker process | gauge | `app`, `pid` |
| `windows_iis_worker_request_errors_total` | Total number of requests that returned an error | counter | `app`, `pid`, `status_code` |
| `windows_iis_worker_request_errors_total` | Total number of responses with the status code `401`, `403`, `404` or `500`. Divide by `windows_iis_worker_responses_total` | counter | `app`, `pid`, `status_code` |
| `windows_iis_worker_responses_total` | Total number of responses sent by the worker process. Denominator for `windows_iis_worker_request_errors_total`, read in the same sample | counter | `app`, `pid` |
| `windows_iis_worker_current_websocket_requests` | Current number of WebSocket requests being processed by the worker process | gauge | `app`, `pid` |
| `windows_iis_worker_websocket_connection_attempts_total` | Total number of WebSocket connection attempts since the worker process started | counter | `app`, `pid` |
| `windows_iis_worker_websocket_connection_accepted_total` | Total number of WebSocket connections accepted since the worker process started | counter | `app`, `pid` |
@@ -158,7 +159,21 @@ See [#2698](https://github.com/prometheus-community/windows_exporter/issues/2698
_This collector does not yet have explained examples, we would appreciate your help adding them!_
## Useful queries
_This collector does not yet have any useful queries added, we would appreciate your help adding them!_
Share of HTTP 500 responses of each application pool over the last 5 minutes:
```promql
sum by (app) (rate(windows_iis_worker_request_errors_total{status_code="500"}[5m]))
/
sum by (app) (rate(windows_iis_worker_responses_total[5m]))
```
`windows_iis_worker_request_errors_total` and `windows_iis_worker_responses_total`
are the value and the base of the `% <status> HTTP Response Sent` counters and
come from the same sample. Don't divide the errors by
`windows_iis_worker_requests_total`: that counter
(`Total HTTP Requests Served`) is read in another query at a slightly different
moment, so the ratio can exceed 1.
## Alerting examples
_This collector does not yet have alerting examples, we would appreciate your help adding them!_
internal/collector/iis/AGENTS.md
+7 -1
View File
@@ -56,7 +56,13 @@ notes were recorded on Windows Server 2025 (build 26100) with only
- `% 401/403/404/500 HTTP Response Sent` of `W3SVC_W3WP` are
`PERF_SAMPLE_FRACTION`. Raw is the cumulative count of that status, base is
the cumulative request count (after one 200 and one 404 request: 404 raw 1,
base 2).
base 2). The collector publishes the raw value as
`worker_request_errors_total` and the base of `% 500 HTTP Response Sent`
(`secondvalue`) as `worker_responses_total`, from the same V8 query, so both
come from the same sample. Both are 32-bit and wrap at 2^32. It assumes that the four counters share one base.
`worker_requests_total` (`Total HTTP Requests Served`) is read in the other
`W3SVC_W3WP` query at a different moment and is not a valid denominator
([#1538](https://github.com/prometheus-community/windows_exporter/issues/1538)).
- `APP_POOL_WAS` uptimes and `Time Since Last Worker Process Failure` are
`PERF_ELAPSED_TIME`. Raw is the start time as a FILETIME (100 ns since 1601),
the second value is the current time, and `internal/pdh` returns the elapsed
internal/collector/iis/iis_w3svc_w3wp.go
+22 -1
View File
@@ -75,6 +75,7 @@ type collectorW3SVCW3WP struct {
// IIS 8+ Only
w3SVCW3WPRequestErrorsTotal *prometheus.Desc
w3SVCW3WPResponsesTotal *prometheus.Desc
w3SVCW3WPWebSocketRequestsActive *prometheus.Desc
w3SVCW3WPWebSocketConnectionAttempts *prometheus.Desc
w3SVCW3WPWebSocketConnectionsAccepted *prometheus.Desc
@@ -139,6 +140,12 @@ type perfDataCounterValuesW3SVCW3WPV8 struct {
W3SVCW3WPRequestErrors403 float64 `perfdata:"% 403 HTTP Response Sent"`
W3SVCW3WPRequestErrors401 float64 `perfdata:"% 401 HTTP Response Sent"`
// The % ... HTTP Response Sent counters are PERF_SAMPLE_FRACTION. The second
// value is the base of the fraction, the cumulative number of responses the
// worker process sent. The four counters are assumed to share it, so one base
// is read.
W3SVCW3WPResponsesTotal float64 `perfdata:"% 500 HTTP Response Sent,secondvalue"`
W3SVCW3WPWebSocketRequestsActive float64 `perfdata:"WebSocket Active Requests"`
W3SVCW3WPWebSocketConnectionAttempts float64 `perfdata:"WebSocket Connection Attempts / Sec"`
W3SVCW3WPWebSocketConnectionsAccepted float64 `perfdata:"WebSocket Connections Accepted / Sec"`
@@ -366,10 +373,16 @@ func (c *Collector) buildW3SVCW3WPDescriptors() {
// W3SVC_W3WP_IIS8
c.w3SVCW3WPRequestErrorsTotal = prometheus.NewDesc(
prometheus.BuildFQName(types.Namespace, Name, "worker_request_errors_total"),
"Total number of requests that returned an error",
"Total number of responses with the status code 401, 403, 404 or 500. Divide by windows_iis_worker_responses_total",
[]string{"app", "pid", "status_code"},
nil,
)
c.w3SVCW3WPResponsesTotal = prometheus.NewDesc(
prometheus.BuildFQName(types.Namespace, Name, "worker_responses_total"),
"Total number of responses sent by the worker process. Denominator for windows_iis_worker_request_errors_total, read in the same sample",
[]string{"app", "pid"},
nil,
)
c.w3SVCW3WPWebSocketRequestsActive = prometheus.NewDesc(
prometheus.BuildFQName(types.Namespace, Name, "worker_current_websocket_requests"),
"Current number of WebSocket requests being processed by the worker process",
@@ -468,6 +481,14 @@ func (c *Collector) collectW3SVCW3WPv8(ch chan<- prometheus.Metric) error {
"500",
)
ch <- prometheus.MustNewConstMetric(
c.w3SVCW3WPResponsesTotal,
prometheus.CounterValue,
data.W3SVCW3WPResponsesTotal,
name,
pid,
)
ch <- prometheus.MustNewConstMetric(
c.w3SVCW3WPWebSocketRequestsActive,
prometheus.GaugeValue,
internal/collector/iis/metric_types_test.go
+35
View File
@@ -197,3 +197,38 @@ func TestW3SVCW3WPCurrentValuesAreGauges(t *testing.T) {
requireMetric(t, metrics, "windows_iis_worker_current_websocket_requests", dto.MetricType_GAUGE, 2, labels)
requireMetric(t, metrics, "windows_iis_worker_requests_total", dto.MetricType_COUNTER, 10, labels)
}
func TestW3SVCW3WPResponsesTotalIsErrorDenominator(t *testing.T) {
t.Parallel()
c := New(nil)
c.buildW3SVCW3WPDescriptors()
// PERF_SAMPLE_FRACTION: raw is the count per status code, the second value
// is the cumulative response count of the same sample.
c.w3SVCW3WPPerfDataCollectorV8 = cacheFixture[perfDataCounterValuesW3SVCW3WPV8]{sources: map[string]float64{
"% 401 HTTP Response Sent": 1,
"% 403 HTTP Response Sent": 2,
"% 404 HTTP Response Sent": 3,
"% 500 HTTP Response Sent": 4,
"% 500 HTTP Response Sent,secondvalue": 20,
}}
metrics := gatherCollect(t, c.collectW3SVCW3WPv8)
labels := map[string]string{"app": "Example", "pid": "123"}
requireMetric(t, metrics, "windows_iis_worker_responses_total", dto.MetricType_COUNTER, 20, labels)
require.NotContains(t, metrics["windows_iis_worker_responses_total"][0].labels, "status_code")
errorMetrics := metrics["windows_iis_worker_request_errors_total"]
require.Len(t, errorMetrics, 4)
want := map[string]float64{"401": 1, "403": 2, "404": 3, "500": 4}
for _, metric := range errorMetrics {
require.Equal(t, dto.MetricType_COUNTER, metric.kind)
require.Equal(t, labels["app"], metric.labels["app"])
require.Equal(t, labels["pid"], metric.labels["pid"])
require.InDelta(t, want[metric.labels["status_code"]], metric.value, 1e-9, metric.labels["status_code"])
}
}