Cmk 36785

charts/cmk-rustik/templates/metrics-cache/deployment.yaml

@@ -68,6 +68,10 @@ spec:
             - --push-registration-insecure-skip-site-ca-verification
             {{- end }}
             {{- end }}
+            {{- if .Values.otel.enabled }}
+            - --otel-endpoint={{ required "otel.endpoint is required when otel is enabled" .Values.otel.endpoint }}
+            - --otel-push-interval={{ .Values.otel.interval }}
+            {{- end }}
             {{- range $k, $v := .Values.emitAll }}
             {{- if $v }}
             - --all-{{ $k }}

charts/cmk-rustik/values.yaml

@@ -178,6 +178,20 @@ pull:
     # certificate, in days. It is ignored when existingSecret is set.
     longevity: 3650
 
+# otel allows you to enable, disable, and configure the OpenTelemetry export.
+# Unlike push mode, this does not send Checkmk sections: it exports pod- and
+# container-level metrics derived from the kubelet data, using OTLP over HTTP.
+otel:
+  # enabled controls whether the OpenTelemetry export is activated or not.
+  enabled: false
+  # endpoint is the base URL of an OpenTelemetry collector's OTLP/HTTP receiver.
+  # Do *not* include the signal path: "/v1/metrics" is appended for you.
+  # Required when otel.enabled is true.
+  endpoint: ""
+  # interval is how often, in seconds, metrics are collected and exported to
+  # the collector. Must be greater than zero.
+  interval: 60
+
 # hostLabels defines how annotations get imported as labels in Checkmk. By
 # default annotations do not become host labels.
 hostLabels:

metrics-cache/src/cli_args.rs

@@ -224,6 +224,14 @@ pub struct CliArgs {
     #[arg(long = "otel-endpoint")]
     pub otel_endpoint: Option<String>,
 
+    /// Push interval in seconds for sending otel metrics.
+    #[arg(
+        long = "otel-push-interval",
+        value_parser = parse_push_interval,
+        default_value = "60"
+    )]
+    pub otel_push_interval: Duration,
+
     /// Emit all Pod resources rather than only annotated ones
     #[arg(long = "all-pods")]
     pub all_pods: bool,
@@ -406,12 +414,19 @@ mod tests {
     }
 
     #[test]
-    fn push_interval_cannot_be_zero() {
+    fn otel_push_interval_cannot_be_less_than_one() {
         assert_eq!(
             parse(&["--push-interval", "0"])
                 .expect_err("push interval of zero should fail")
                 .kind(),
             ErrorKind::ValueValidation
         );
+
+        assert_eq!(
+            parse(&["--push-interval=-1"])
+                .expect_err("push interval of -1 should fail")
+                .kind(),
+            ErrorKind::ValueValidation
+        );
     }
 }

metrics-cache/src/main.rs

@@ -165,7 +165,7 @@ async fn main() -> anyhow::Result<()> {
         // Push to OpenTelemetry collector
         () = async {
             match otel_client {
-                Some(client) => otel_loop(client, state.clone()).await,
+                Some(client) => otel_loop(client, state.clone(), args.otel_push_interval).await,
                 None => std::future::pending().await,
             }
         } => {

metrics-cache/src/otel/mod.rs

@@ -66,8 +66,12 @@ use crate::state::AppState;
 ///
 /// Depends entirely on the kubelet stats cache in `state` and does effectively
 /// nothing if it is not yet populated.
-pub async fn otel_loop(client: OtelClient, state: AppState<impl TokenValidator>) {
-    let mut interval = time::interval(Duration::from_secs(60)); // TODO: Unhardcode
+pub async fn otel_loop(
+    client: OtelClient,
+    state: AppState<impl TokenValidator>,
+    otel_push_interval: Duration,
+) {
+    let mut interval = time::interval(otel_push_interval);
     loop {
         interval.tick().await; // note: The very first tick() is no-op
         let stat_summaries = state.kubelet_stats_summary_cache.iter().map(|(_, v)| v);