Struct KubernetesIngressWatcher 

pub struct KubernetesIngressWatcher { /* private fields */ }

Watches a Kubernetes API server for Ingress resources and maintains a live routing table.

Implementations

impl KubernetesIngressWatcher

pub fn new(api_server: impl Into<String>, token: impl Into<String>) -> Self

Create a watcher from explicit values.

api_server should be a plain-HTTP URL such as http://localhost:8001. The default namespace is "default".

pub fn from_service_account() -> Result<Self, String>

Configure from the Kubernetes service account files mounted at /var/run/secrets/kubernetes.io/serviceaccount/ and the KUBERNETES_SERVICE_HOST/KUBERNETES_SERVICE_PORT environment variables Kubernetes injects into every pod, connecting to the in-cluster API server directly over TLS — no kubectl proxy sidecar needed.

Requires the http-client or http2 feature (both already pull in rustls); trusts only the cluster’s own CA certificate (.../ca.crt), not the public root store crate::http_client uses, since the API server’s certificate is signed by that private CA.

pub fn from_env() -> Result<Self, String>

Configure from environment variables RWS_K8S_API_SERVER, RWS_K8S_TOKEN, and (optionally) RWS_K8S_NAMESPACE.

Returns Err if RWS_K8S_API_SERVER is not set.

pub fn namespace(self, ns: impl Into<String>) -> Self

Override the namespace filter. Use "all" or empty string for all namespaces.

pub fn ingress_class(self, class: impl Into<String>) -> Self

Restrict watched Ingress objects to those whose spec.ingressClassName equals class exactly. Unset (the default) accepts every Ingress cluster- or namespace-wide regardless of class — fine for a single-controller cluster, but on a multi-controller cluster (e.g. running alongside nginx-ingress) leaves this watcher picking up Ingress objects meant for the other controller too, so set this explicitly in that case.

pub fn poll_interval_secs(self, secs: u64) -> Self

Override the polling interval in seconds (default: 30).

This remains a periodic full resync even when the watch connection (always attempted alongside it — see the module docs) is healthy, the same “trust but verify” pattern real Kubernetes controllers use: the watch stream delivers low-latency updates, and this interval is the safety net against a missed or silently swallowed event.

pub fn start(&self)

Spawn background threads that keep the rule table up to date: a periodic full resync every poll_interval_secs, and (best-effort) a long-lived watch connection that triggers an immediate resync on any change instead of waiting for the next interval. Call once at startup.

pub fn rules(&self) -> Vec<IngressRule>

Return a snapshot of the current rule list.

pub fn poll(&self) -> Result<(), String>

Perform one synchronous poll cycle.

Exported for testing without starting the background thread.