From f0ea4062c5185954b9514fa1171118f00ab28724 Mon Sep 17 00:00:00 2001
From: Razvan-Daniel Mihai <84674+razvan@users.noreply.github.com>
Date: Tue, 10 Dec 2024 14:47:31 +0100
Subject: [PATCH] feat: run containerdebug in the background

---
 .../src/history/history_controller.rs                | 12 +++++++++---
 rust/operator-binary/src/spark_k8s_controller.rs     |  5 +++++
 tests/templates/kuttl/smoke/06-assert.yaml           |  7 +++++++
 tests/templates/kuttl/smoke/10-assert.yaml           |  7 +++++++
 4 files changed, 28 insertions(+), 3 deletions(-)

diff --git a/rust/operator-binary/src/history/history_controller.rs b/rust/operator-binary/src/history/history_controller.rs
index 5149bf4d..9637658b 100644
--- a/rust/operator-binary/src/history/history_controller.rs
+++ b/rust/operator-binary/src/history/history_controller.rs
@@ -527,7 +527,13 @@ fn build_stateful_set(
         .context(InvalidContainerNameSnafu)?
         .image_from_product_image(resolved_product_image)
         .resources(merged_config.resources.clone().into())
-        .command(vec!["/bin/bash".to_string()])
+        .command(vec![
+            "/bin/bash".to_string(),
+            "-x".to_string(),
+            "-euo".to_string(),
+            "pipefail".to_string(),
+            "-c".to_string(),
+        ])
         .args(command_args(log_dir))
         .add_container_port("http", 18080)
         .add_container_port("metrics", METRICS_PORT.into())
@@ -751,10 +757,10 @@ fn command_args(logdir: &ResolvedLogDir) -> Vec<String> {
     }
 
     command.extend(vec![
+        format!("CONTAINERDEBUG_LOG_DIRECTORY={VOLUME_MOUNT_PATH_LOG}/containerdebug containerdebug --output={VOLUME_MOUNT_PATH_LOG}/containerdebug-state.json --loop &"),
         format!("/stackable/spark/sbin/start-history-server.sh --properties-file {VOLUME_MOUNT_PATH_CONFIG}/{SPARK_DEFAULTS_FILE_NAME}"),
     ]);
-
-    vec![String::from("-c"), command.join(" && ")]
+    vec![command.join("\n")]
 }
 
 fn labels<'a, T>(
diff --git a/rust/operator-binary/src/spark_k8s_controller.rs b/rust/operator-binary/src/spark_k8s_controller.rs
index ccb9eb84..40020f51 100644
--- a/rust/operator-binary/src/spark_k8s_controller.rs
+++ b/rust/operator-binary/src/spark_k8s_controller.rs
@@ -594,6 +594,11 @@ fn pod_template(
         .resources(config.resources.clone().into())
         .image_from_product_image(spark_image);
 
+    cb.add_env_var(
+        "_STACKABLE_PRE_HOOK",
+        format!("CONTAINERDEBUG_LOG_DIRECTORY={VOLUME_MOUNT_PATH_LOG}/containerdebug containerdebug --output={VOLUME_MOUNT_PATH_LOG}/containerdebug-state.json --loop &"),
+    );
+
     if config.logging.enable_vector_agent {
         cb.add_env_var(
             "_STACKABLE_POST_HOOK",
diff --git a/tests/templates/kuttl/smoke/06-assert.yaml b/tests/templates/kuttl/smoke/06-assert.yaml
index 68c0e487..386ee1fb 100644
--- a/tests/templates/kuttl/smoke/06-assert.yaml
+++ b/tests/templates/kuttl/smoke/06-assert.yaml
@@ -9,3 +9,10 @@ metadata:
   name: spark-history-node-default
 status:
   readyReplicas: 1
+---
+# This test checks if the containerdebug-state.json file is present and valid
+apiVersion: kuttl.dev/v1beta1
+kind: TestAssert
+timeout: 600
+commands:
+  - script: kubectl exec -n $NAMESPACE --container airflow spark-history-node-default-0 -- cat /stackable/log/containerdebug-state.json | jq --exit-status
diff --git a/tests/templates/kuttl/smoke/10-assert.yaml b/tests/templates/kuttl/smoke/10-assert.yaml
index 38367193..9ac8d04e 100644
--- a/tests/templates/kuttl/smoke/10-assert.yaml
+++ b/tests/templates/kuttl/smoke/10-assert.yaml
@@ -9,3 +9,10 @@ metadata:
   name: spark-pi-s3-1
 status:
   phase: Succeeded
+---
+# This test checks if the containerdebug-state.json file is present and valid
+apiVersion: kuttl.dev/v1beta1
+kind: TestAssert
+timeout: 600
+commands:
+  - script: kubectl exec -n $NAMESPACE --container airflow spark-history-node-default-0 -- cat /stackable/log/containerdebug-state.json | jq --exit-status