feat: automatic cluster detection

Cargo.toml

@@ -25,6 +25,7 @@ educe =  { version = "0.6.0", default-features = false, features = ["Clone", "De
 either = "1.13.0"
 futures = "0.3.30"
 futures-util = "0.3.30"
+http = "1.3.1"
 indexmap = "2.5.0"
 indoc = "2.0.6"
 insta = { version= "1.40", features = ["glob"] }

crates/stackable-operator/CHANGELOG.md

@@ -4,6 +4,23 @@ All notable changes to this project will be documented in this file.
 
 ## [Unreleased]
 
+### Added
+
+- The default Kubernetes cluster domain name is now fetched from the kubelet API unless explicitly configured ([#1068])
+  This requires operators to have the RBAC permission to `get` `nodes/proxy` in the apiGroup "", an example RBAC rule could look like:
+
+  ```yaml
+  ---
+  apiVersion: rbac.authorization.k8s.io/v1
+  kind: ClusterRole
+  metadata:
+    name: operator-cluster-role
+  rules:
+    - apiGroups: [""]
+      resources: [nodes/proxy]
+      verbs: [get]
+  ```
+
 ### Changed
 
 - Update `kube` to `1.1.0` ([#1049]).
@@ -23,6 +40,7 @@ All notable changes to this project will be documented in this file.
 [#1058]: https://github.com/stackabletech/operator-rs/pull/1058
 [#1060]: https://github.com/stackabletech/operator-rs/pull/1060
 [#1064]: https://github.com/stackabletech/operator-rs/pull/1064
+[#1068]: https://github.com/stackabletech/operator-rs/pull/1068
 
 ## [0.93.2] - 2025-05-26
 
@@ -148,7 +166,7 @@ All notable changes to this project will be documented in this file.
 ### Added
 
 - Add Deployments to `ClusterResource`s ([#992]).
-- Add `DeploymentConditionBuilder`  ([#993]).
+- Add `DeploymentConditionBuilder` ([#993]).
 
 ### Changed
 
@@ -369,7 +387,7 @@ All notable changes to this project will be documented in this file.
 ### Fixed
 
 - BREAKING: `KeyValuePairs::insert` (as well as `Labels::`/`Annotations::` via it) now overwrites
-  the old value if the key already exists. Previously, `iter()` would return *both* values in
+  the old value if the key already exists. Previously, `iter()` would return _both_ values in
   lexicographical order (causing further conversions like `Into<BTreeMap>` to prefer the maximum
   value) ([#888]).
 
@@ -634,7 +652,7 @@ All notable changes to this project will be documented in this file.
 
 ### Changed
 
-- Implement `PartialEq` for most *Snafu* Error enums ([#757]).
+- Implement `PartialEq` for most _Snafu_ Error enums ([#757]).
 - Update Rust to 1.77 ([#759])
 
 ### Fixed
@@ -1385,7 +1403,7 @@ This is a rerelease of 0.25.1 which some last-minute incompatible API changes to
 ### Changed
 
 - Objects are now streamed rather than polled when waiting for them to be deleted ([#452]).
-- serde\_yaml 0.8.26 -> 0.9.9 ([#450])
+- serde_yaml 0.8.26 -> 0.9.9 ([#450])
 
 [#450]: https://github.com/stackabletech/operator-rs/pull/450
 [#452]: https://github.com/stackabletech/operator-rs/pull/452

crates/stackable-operator/Cargo.toml

@@ -28,6 +28,7 @@ dockerfile-parser.workspace = true
 either.workspace = true
 educe.workspace = true
 futures.workspace = true
+http.workspace = true
 indexmap.workspace = true
 json-patch.workspace = true
 k8s-openapi.workspace = true

crates/stackable-operator/src/client.rs

@@ -84,6 +84,11 @@ pub enum Error {
 
     #[snafu(display("unable to create kubernetes client"))]
     CreateKubeClient { source: kube::Error },
+
+    #[snafu(display("unable to fetch cluster information from kubelet"))]
+    NewKubeletClusterInfo {
+        source: crate::utils::cluster_info::Error,
+    },
 }
 
 /// This `Client` can be used to access Kubernetes.
@@ -651,7 +656,9 @@ pub async fn initialize_operator(
         .context(InferKubeConfigSnafu)?;
     let default_namespace = kubeconfig.default_namespace.clone();
     let client = kube::Client::try_from(kubeconfig).context(CreateKubeClientSnafu)?;
-    let cluster_info = KubernetesClusterInfo::new(cluster_info_opts);
+    let cluster_info = KubernetesClusterInfo::new(&client, cluster_info_opts)
+        .await
+        .context(NewKubeletClusterInfoSnafu)?;
 
     Ok(Client::new(
         client,

crates/stackable-operator/src/utils/cluster_info.rs

@@ -1,10 +1,14 @@
-use std::str::FromStr;
+use kube::Client;
+use snafu::{ResultExt, Snafu};
 
-use crate::commons::networking::DomainName;
+use crate::{commons::networking::DomainName, utils::kubelet};
 
-const KUBERNETES_CLUSTER_DOMAIN_DEFAULT: &str = "cluster.local";
+#[derive(Debug, Snafu)]
+pub enum Error {
+    #[snafu(display("unable to fetch kubelet config"))]
+    KubeletConfig { source: kubelet::Error },
+}
 
-/// Some information that we know about the Kubernetes cluster.
 #[derive(Debug, Clone)]
 pub struct KubernetesClusterInfo {
     /// The Kubernetes cluster domain, typically `cluster.local`.
@@ -21,25 +25,28 @@ pub struct KubernetesClusterInfoOpts {
 }
 
 impl KubernetesClusterInfo {
-    pub fn new(cluster_info_opts: &KubernetesClusterInfoOpts) -> Self {
+    pub async fn new(
+        client: &Client,
+        cluster_info_opts: &KubernetesClusterInfoOpts,
+    ) -> Result<Self, Error> {
         let cluster_domain = match &cluster_info_opts.kubernetes_cluster_domain {
             Some(cluster_domain) => {
                 tracing::info!(%cluster_domain, "Using configured Kubernetes cluster domain");
 
                 cluster_domain.clone()
             }
             None => {
-                // TODO(sbernauer): Do some sort of advanced auto-detection, see https://github.com/stackabletech/issues/issues/436.
-                // There have been attempts of parsing the `/etc/resolv.conf`, but they have been
-                // reverted. Please read on the linked issue for details.
-                let cluster_domain = DomainName::from_str(KUBERNETES_CLUSTER_DOMAIN_DEFAULT)
-                    .expect("KUBERNETES_CLUSTER_DOMAIN_DEFAULT constant must a valid domain");
-                tracing::info!(%cluster_domain, "Defaulting Kubernetes cluster domain as it has not been configured");
+                let kubelet_config = kubelet::KubeletConfig::fetch(client)
+                    .await
+                    .context(KubeletConfigSnafu)?;
+
+                let cluster_domain = kubelet_config.cluster_domain;
+                tracing::info!(%cluster_domain, "Using Kubernetes cluster domain from the kubelet config");
 
                 cluster_domain
             }
         };
 
-        Self { cluster_domain }
+        Ok(Self { cluster_domain })
     }
 }

crates/stackable-operator/src/utils/kubelet.rs

@@ -0,0 +1,74 @@
+use http;
+use k8s_openapi::api::core::v1::Node;
+use kube::{
+    Api,
+    api::{ListParams, ResourceExt},
+    client::Client,
+};
+use serde::Deserialize;
+use snafu::{OptionExt, ResultExt, Snafu};
+
+use crate::commons::networking::DomainName;
+
+#[derive(Debug, Snafu)]
+pub enum Error {
+    #[snafu(display("failed to list nodes"))]
+    ListNodes { source: kube::Error },
+
+    #[snafu(display("failed to build request for url path \"{url_path}\""))]
+    BuildConfigzRequest {
+        source: http::Error,
+        url_path: String,
+    },
+
+    #[snafu(display("failed to fetch kubelet config from node {node:?}"))]
+    FetchNodeKubeletConfig { source: kube::Error, node: String },
+
+    #[snafu(display("failed to fetch `kubeletconfig` JSON key from configz response"))]
+    KubeletConfigJsonKey,
+
+    #[snafu(display("failed to deserialize kubelet config JSON"))]
+    KubeletConfigJson { source: serde_json::Error },
+
+    #[snafu(display(
+        "empty Kubernetes nodes list. At least one node is required to fetch the cluster domain from the kubelet config"
+    ))]
+    EmptyKubernetesNodesList,
+}
+
+#[derive(Debug, Deserialize)]
+#[serde(rename_all = "camelCase")]
+struct ProxyConfigResponse {
+    kubeletconfig: KubeletConfig,
+}
+
+#[derive(Debug, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct KubeletConfig {
+    pub cluster_domain: DomainName,
+}
+
+impl KubeletConfig {
+    /// Fetches the kubelet configuration from the "first" node in the Kubernetes cluster.
+    pub async fn fetch(client: &Client) -> Result<Self, Error> {
+        let api: Api<Node> = Api::all(client.clone());
+        let nodes = api
+            .list(&ListParams::default())
+            .await
+            .context(ListNodesSnafu)?;
+        let node = nodes.iter().next().context(EmptyKubernetesNodesListSnafu)?;
+        let node_name = node.name_any();
+
+        let url_path = format!("/api/v1/nodes/{node_name}/proxy/configz");
+        let req = http::Request::get(url_path.clone())
+            .body(Default::default())
+            .context(BuildConfigzRequestSnafu { url_path })?;
+
+        let resp = client
+            .request::<ProxyConfigResponse>(req)
+            .await
+            .context(FetchNodeKubeletConfigSnafu { node: node_name })?;
+
+        Ok(resp.kubeletconfig)
+    }
+}

crates/stackable-operator/src/utils/mod.rs

@@ -1,6 +1,7 @@
 pub mod bash;
 pub mod cluster_info;
 pub mod crds;
+pub mod kubelet;
 pub mod logging;
 mod option;
 mod url;