batch-analytics 0.2.9__tar.gz → 0.3.0__tar.gz

{batch_analytics-0.2.9 → batch_analytics-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: batch-analytics
-Version: 0.2.9
+Version: 0.3.0
 Summary: PySpark batch analytics: Extract, Transform, Stage, and analytical modules (linear regression, correlation, PCA, t-test).
 Author: Litewave Analytics Team
 License: MIT

{batch_analytics-0.2.9 → batch_analytics-0.3.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "batch-analytics"
-version = "0.2.9"
+version = "0.3.0"
 description = "PySpark batch analytics: Extract, Transform, Stage, and analytical modules (linear regression, correlation, PCA, t-test)."
 readme = "README.md"
 requires-python = ">=3.8"

{batch_analytics-0.2.9 → batch_analytics-0.3.0}/src/batch_analytics/job_runner.py

@@ -13,7 +13,7 @@ from typing import Dict, List, Optional
 
 from pyspark.sql import SparkSession
 
-from .config import BatchAnalyticsConfig
+from .config import BatchAnalyticsConfig, SparkK8sConfig
 from .extract import extract_unified
 from .log import log_dataframe_summary, log_run
 from .modules import DEFAULT_MODULES, MODULE_REGISTRY, VALID_MODULES
@@ -27,6 +27,21 @@ logging.basicConfig(
 logger = logging.getLogger(__name__)
 
 
+def _spark_on_kubernetes_for_classpath(cfg: SparkK8sConfig) -> bool:
+    """
+    True when Spark executors run on Kubernetes.
+
+    Spark Operator cluster mode passes --master k8s://... to spark-submit; SPARK_MASTER is often
+    unset, so SparkK8sConfig.master defaults to local[*] while the JVM still uses Kubernetes.
+    Image-baked JARs must go on extraClassPath in that case (spark.jars breaks executors).
+    """
+    if cfg.master.startswith("k8s://"):
+        return True
+    if os.environ.get("SPARK_MASTER", "").strip().startswith("k8s://"):
+        return True
+    return bool(os.environ.get("KUBERNETES_SERVICE_HOST"))
+
+
 def create_spark_session(
     app_name: str = "BatchAnalytics",
     clickhouse_jars: Optional[str] = None,
@@ -64,13 +79,13 @@ def create_spark_session(
             else:
                 packages.append(part)
 
-    if cfg.master.startswith("k8s://"):
+    if _spark_on_kubernetes_for_classpath(cfg):
         packages.append("org.apache.hadoop:hadoop-aws:3.3.4")
 
     # Spark-on-K8s: absolute paths in spark.jars are re-sent to executors as ./basename.jar and
     # fail there ("Unable to create executor due to ./clickhouse-spark-runtime-..."). JARs baked
     # into the driver/executor image belong on the JVM classpath instead.
-    if cfg.master.startswith("k8s://") and jar_list:
+    if _spark_on_kubernetes_for_classpath(cfg) and jar_list:
         local_cp: List[str] = []
         remote_jars: List[str] = []
         for p in jar_list:
@@ -83,6 +98,10 @@ def create_spark_session(
                 remote_jars.append(p)
         if local_cp:
             joined = ":".join(local_cp)
+            logger.info(
+                "Spark on Kubernetes: image JARs on driver/executor extraClassPath (not spark.jars): %s",
+                joined,
+            )
             builder = (
                 builder.config("spark.driver.extraClassPath", joined)
                 .config("spark.executor.extraClassPath", joined)

{batch_analytics-0.2.9 → batch_analytics-0.3.0}/src/batch_analytics.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: batch-analytics
-Version: 0.2.9
+Version: 0.3.0
 Summary: PySpark batch analytics: Extract, Transform, Stage, and analytical modules (linear regression, correlation, PCA, t-test).
 Author: Litewave Analytics Team
 License: MIT