From 6008a8f94150c30a7029f4eb86ef17bf6e256a9f Mon Sep 17 00:00:00 2001
From: Razvan-Daniel Mihai <84674+razvan@users.noreply.github.com>
Date: Sat, 14 Dec 2024 14:46:20 +0100
Subject: [PATCH] reintroduce the CONTAINERDEBUG_LOG_DIRECTORY env var

---
 rust/crd/src/history.rs                       | 10 +++++++
 rust/crd/src/lib.rs                           |  9 +++++++
 .../src/spark_k8s_controller.rs               | 26 ++++++++++++++++---
 3 files changed, 41 insertions(+), 4 deletions(-)

diff --git a/rust/crd/src/history.rs b/rust/crd/src/history.rs
index 1ca93081..55b9f4be 100644
--- a/rust/crd/src/history.rs
+++ b/rust/crd/src/history.rs
@@ -242,6 +242,16 @@ impl SparkHistoryServer {
         let mut vars: BTreeMap<String, EnvVar> = BTreeMap::new();
         let role_env_overrides = &self.role().config.env_overrides;
 
+        // Needed by the `containerdebug` running in the background of the history container
+        // to log it's tracing information to.
+        vars.insert(
+            "CONTAINERDEBUG_LOG_DIRECTORY".to_string(),
+            EnvVar {
+                name: "CONTAINERDEBUG_LOG_DIRECTORY".to_string(),
+                value: Some(format!("{VOLUME_MOUNT_PATH_LOG}/containerdebug")),
+                value_from: None,
+            },
+        );
         // This env var prevents the history server from detaching itself from the
         // start script because this leads to the Pod terminating immediately.
         vars.insert(
diff --git a/rust/crd/src/lib.rs b/rust/crd/src/lib.rs
index 3f85f2d8..9a8bb704 100644
--- a/rust/crd/src/lib.rs
+++ b/rust/crd/src/lib.rs
@@ -692,6 +692,15 @@ impl SparkApplication {
         logdir: &Option<ResolvedLogDir>,
     ) -> Vec<EnvVar> {
         let mut e: Vec<EnvVar> = self.spec.env.clone();
+
+        // Needed by the `containerdebug` process running in the background of the `spark-submit`
+        // container to log it's tracing information to.
+        e.push(EnvVar {
+            name: "CONTAINERDEBUG_LOG_DIRECTORY".to_string(),
+            value: Some(format!("{VOLUME_MOUNT_PATH_LOG}/containerdebug")),
+            value_from: None,
+        });
+
         if self.requirements().is_some() {
             e.push(EnvVar {
                 name: "PYTHONPATH".to_string(),
diff --git a/rust/operator-binary/src/spark_k8s_controller.rs b/rust/operator-binary/src/spark_k8s_controller.rs
index d08f0142..7bcef777 100644
--- a/rust/operator-binary/src/spark_k8s_controller.rs
+++ b/rust/operator-binary/src/spark_k8s_controller.rs
@@ -612,11 +612,29 @@ fn pod_template(
         .resources(config.resources.clone().into())
         .image_from_product_image(spark_image);
 
-    cb.add_env_var(
-        "_STACKABLE_PRE_HOOK",
-        format!(
+    // These env variables enable the `containerdebug` process in driver and executor pods.
+    // More precisely, this process runs in the background of every `spark` container.
+    // - `CONTAINERDEBUG_LOG_DIRECTORY` - is the location where tracing information from the process
+    // is written. This directory is created by the process it's self.
+    // - `_STACKABLE_PRE_HOOK` - is evaluated by the entrypoint script (run-spark.sh) in the Spark images
+    // before the actual JVM process is started. The result of this evaluation is that the
+    // `containerdebug` process is executed in the background.
+    cb.add_env_vars(
+        [
+            EnvVar {
+                name: "CONTAINERDEBUG_LOG_DIRECTORY".into(),
+                value: Some(format!("{VOLUME_MOUNT_PATH_LOG}/containerdebug")),
+                value_from: None,
+            },
+            EnvVar {
+                name: "_STACKABLE_PRE_HOOK".into(),
+                value: Some(format!(
             "containerdebug --output={VOLUME_MOUNT_PATH_LOG}/containerdebug-state.json --loop &"
-        ),
+        )),
+                value_from: None,
+            },
+        ]
+        .into(),
     );
 
     if config.logging.enable_vector_agent {