fix: Add startupProbe to prevent Superset startup problems

CHANGELOG.md

@@ -2,6 +2,13 @@
 
 ## [Unreleased]
 
+### Fixed
+
+- Fix container not starting because Superset was starting too slow and was killed because a failing liveness probe.
+  We now add a proper startup probe, which allows Superset to take longer to start up ([#654]).
+
+[#654]: https://github.com/stackabletech/superset-operator/pull/654
+
 ## [25.7.0] - 2025-07-23
 
 ## [25.7.0-rc1] - 2025-07-18

Cargo.toml

@@ -11,7 +11,7 @@ repository = "https://github.com/stackabletech/superset-operator"
 
 [workspace.dependencies]
 product-config = { git = "https://github.com/stackabletech/product-config.git", tag = "0.7.0" }
-stackable-operator = { git = "https://github.com/stackabletech/operator-rs.git", features = ["telemetry", "versioned"], tag = "stackable-operator-0.95.0" }
+stackable-operator = { git = "https://github.com/stackabletech/operator-rs.git", features = ["telemetry", "versioned"], tag = "stackable-operator-0.96.0" }
 
 anyhow = "1.0"
 built = { version = "0.8", features = ["chrono", "git2"] }

rust/operator-binary/src/main.rs

@@ -7,7 +7,7 @@ use clap::Parser;
 use futures::{StreamExt, pin_mut};
 use stackable_operator::{
     YamlSchema,
-    cli::{Command, ProductOperatorRun},
+    cli::{Command, CommonOptions, ProductOperatorRun},
     crd::authentication::core,
     k8s_openapi::api::{
         apps::v1::StatefulSet,
@@ -77,11 +77,14 @@ async fn main() -> anyhow::Result<()> {
                 .print_yaml_schema(built_info::PKG_VERSION, SerializeOptions::default())?;
         }
         Command::Run(ProductOperatorRun {
+            common:
+                CommonOptions {
+                    telemetry,
+                    cluster_info,
+                },
+            operator_environment: _,
             product_config,
             watch_namespace,
-            operator_environment: _,
-            telemetry,
-            cluster_info,
         }) => {
             // NOTE (@NickLarsenNZ): Before stackable-telemetry was used:
             // - The console log level was set by `SUPERSET_OPERATOR_LOG`, and is now `CONSOLE_LOG` (when using Tracing::pre_configured).

rust/operator-binary/src/superset_controller.rs

@@ -22,6 +22,7 @@ use stackable_operator::{
         pod::{
             PodBuilder,
             container::ContainerBuilder,
+            probe::ProbeBuilder,
             resources::ResourceRequirementsBuilder,
             security::PodSecurityContextBuilder,
             volume::{
@@ -40,9 +41,9 @@ use stackable_operator::{
         DeepMerge,
         api::{
             apps::v1::{StatefulSet, StatefulSetSpec},
-            core::v1::{ConfigMap, EnvVar, HTTPGetAction, Probe},
+            core::v1::{ConfigMap, EnvVar},
         },
-        apimachinery::pkg::{apis::meta::v1::LabelSelector, util::intstr::IntOrString},
+        apimachinery::pkg::apis::meta::v1::LabelSelector,
     },
     kube::{
         Resource, ResourceExt,
@@ -797,21 +798,7 @@ fn build_server_rolegroup_statefulset(
                 create_vector_shutdown_file_command(STACKABLE_LOG_DIR),
         }])
         .resources(merged_config.resources.clone().into());
-    let probe = Probe {
-        http_get: Some(HTTPGetAction {
-            port: IntOrString::Int(APP_PORT.into()),
-            path: Some("/health".to_string()),
-            ..HTTPGetAction::default()
-        }),
-        initial_delay_seconds: Some(15),
-        period_seconds: Some(15),
-        timeout_seconds: Some(1),
-        failure_threshold: Some(3),
-        success_threshold: Some(1),
-        ..Probe::default()
-    };
-    superset_cb.readiness_probe(probe.clone());
-    superset_cb.liveness_probe(probe);
+    add_superset_container_probes(&mut superset_cb);
 
     // listener endpoints will use persistent volumes
     // so that load balancers can hard-code the target addresses and
@@ -943,6 +930,36 @@ fn build_server_rolegroup_statefulset(
     })
 }
 
+fn add_superset_container_probes(superset_cb: &mut ContainerBuilder) {
+    let common =
+        ProbeBuilder::http_get_port_scheme_path(APP_PORT, None, Some("/health".to_owned()))
+            .with_period(Duration::from_secs(5));
+
+    superset_cb.startup_probe(
+        common
+            .clone()
+            .with_failure_threshold_duration(Duration::from_minutes_unchecked(10))
+            .expect("const period is non-zero")
+            .build()
+            .expect("const duration does not overflow"),
+    );
+
+    // Remove it from the Service immediately
+    superset_cb.readiness_probe(
+        common
+            .clone()
+            .build()
+            .expect("const duration does not overflow"),
+    );
+    // But only restart it after 3 failures
+    superset_cb.liveness_probe(
+        common
+            .with_failure_threshold(3)
+            .build()
+            .expect("const duration does not overflow"),
+    );
+}
+
 fn add_authentication_volumes_and_volume_mounts(
     auth_config: &SupersetClientAuthenticationDetailsResolved,
     cb: &mut ContainerBuilder,