feat(health): add endpoint TCP reachability telemetry

Signed-off-by: Jay Zhu <jayzhu@nvidia.com>

Author: jayzhudev

crates/health/example/config.example.toml

@@ -201,6 +201,16 @@ max_jitter = "40ms"
 refresh_interval = "60m"
 discovery_concurrency = 1
 
+# Optional direct TCP checks for ports used by eligible enabled collectors.
+# Every check emits a metric; "unreachable" logs failures and "all" logs every result.
+# Checks bypass BMC proxies and do not verify TLS, authentication, or protocol readiness.
+# Results never create health reports or alerts.
+[collectors.reachability]
+enabled = true
+interval = "30s"
+timeout = "3s"
+log_mode = "unreachable"
+
 [collectors.sensors]
 sensor_fetch_interval = "1m"
 sensor_fetch_concurrency = 10

crates/health/src/collectors/mod.rs

@@ -27,6 +27,7 @@ mod nmxt;
 mod nvue;
 #[cfg(test)]
 pub(in crate::collectors) mod projection_test_support;
+mod reachability;
 mod runtime;
 mod sensors;
 
@@ -42,9 +43,14 @@ pub use logs::{
     SseLogCollector, SseLogCollectorConfig,
 };
 pub use nmxc::{NmxcCollector, NmxcCollectorConfig};
+pub(crate) use nmxt::NMXT_PORT;
 pub use nmxt::{NmxtCollector, NmxtCollectorConfig};
 pub(crate) use nvue::gnmi::subscriber::spawn_gnmi_collector;
 pub use nvue::rest::collector::{NvueRestCollector, NvueRestCollectorConfig};
+pub(crate) use reachability::{
+    COLLECTOR_TYPE as REACHABILITY_COLLECTOR_TYPE, ReachabilityCollector,
+    ReachabilityCollectorStartConfig, ReachabilityService, ReachabilityTarget,
+};
 pub use runtime::{
     BackoffConfig, Collector, CollectorStartContext, EventStream, ExponentialBackoff,
     IterationResult, PeriodicCollector, StreamMetrics, StreamingCollector,

crates/health/src/collectors/nmxt.rs

@@ -41,8 +41,8 @@ use crate::endpoint::BmcEndpoint;
 use crate::sink::{CollectorEvent, DataSink, EventContext, MetricSample};
 use crate::tls::MtlsHttpClientProvider;
 
-/// default NMX-T port
-const NMXT_PORT: u16 = 9352;
+/// Default TCP port for NMX-T.
+pub(crate) const NMXT_PORT: u16 = 9352;
 
 /// NMX-T endpoint
 const NMXT_ENDPOINT: &str = "/xcset/nvlink_domain_telemetry";