batch-analytics 0.3.0__tar.gz → 0.3.2__tar.gz

{batch_analytics-0.3.0 → batch_analytics-0.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: batch-analytics
-Version: 0.3.0
+Version: 0.3.2
 Summary: PySpark batch analytics: Extract, Transform, Stage, and analytical modules (linear regression, correlation, PCA, t-test).
 Author: Litewave Analytics Team
 License: MIT

{batch_analytics-0.3.0 → batch_analytics-0.3.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "batch-analytics"
-version = "0.3.0"
+version = "0.3.2"
 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-0.3.0 → batch_analytics-0.3.2}/src/batch_analytics/job_runner.py

@@ -10,10 +10,11 @@ import sys
 import uuid
 from pathlib import Path
 from typing import Dict, List, Optional
+from urllib.parse import urlparse
 
 from pyspark.sql import SparkSession
 
-from .config import BatchAnalyticsConfig, SparkK8sConfig
+from .config import BatchAnalyticsConfig
 from .extract import extract_unified
 from .log import log_dataframe_summary, log_run
 from .modules import DEFAULT_MODULES, MODULE_REGISTRY, VALID_MODULES
@@ -27,19 +28,48 @@ logging.basicConfig(
 logger = logging.getLogger(__name__)
 
 
-def _spark_on_kubernetes_for_classpath(cfg: SparkK8sConfig) -> bool:
+def _local_jar_path_for_match(p: str) -> str:
+    """Normalize file:/path or /path for prefix checks."""
+    p = p.strip()
+    if not p:
+        return p
+    if p.startswith("file:"):
+        parsed = urlparse(p)
+        return parsed.path if parsed.path else p[5:]
+    return p
+
+
+def _omit_spark_distrib_jars(jar_list: List[str]) -> List[str]:
     """
-    True when Spark executors run on Kubernetes.
+    Drop paths under $SPARK_HOME/jars from spark.jars.
+
+    Those JARs are already on the driver and executor JVM classpath (Spark launch scripts add
+    $SPARK_HOME/jars/*). Listing them again in spark.jars makes Spark distribute them to executors
+    as ./basename.jar and breaks Kubernetes executors.
 
-    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).
+    SPARK_HOME may be unset or empty in some pods; always treat /opt/spark/jars/ as Spark distro
+    (apache/spark images).
     """
-    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"))
+    spark_home = (os.environ.get("SPARK_HOME") or "/opt/spark").rstrip("/")
+    prefixes = (f"{spark_home}/jars/", "/opt/spark/jars/")
+    out: List[str] = []
+    skipped: List[str] = []
+    for p in jar_list:
+        p = p.strip()
+        if not p:
+            continue
+        norm = _local_jar_path_for_match(p)
+        if norm.endswith(".jar") and any(norm.startswith(pref) for pref in prefixes):
+            skipped.append(p)
+        else:
+            out.append(p)
+    if skipped:
+        logger.info(
+            "Omitting spark.jars for JARs already on Spark classpath (%s): %s",
+            prefixes[0],
+            ",".join(skipped),
+        )
+    return out
 
 
 def create_spark_session(
@@ -79,36 +109,7 @@ def create_spark_session(
             else:
                 packages.append(part)
 
-    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 _spark_on_kubernetes_for_classpath(cfg) and jar_list:
-        local_cp: List[str] = []
-        remote_jars: List[str] = []
-        for p in jar_list:
-            p = p.strip()
-            if not p:
-                continue
-            if p.startswith("/") and p.endswith(".jar"):
-                local_cp.append(p)
-            else:
-                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)
-                .config("spark.kubernetes.driver.extraClassPath", joined)
-                .config("spark.kubernetes.executor.extraClassPath", joined)
-            )
-        jar_list = remote_jars
+    jar_list = _omit_spark_distrib_jars(jar_list)
 
     if jar_list:
         builder = builder.config("spark.jars", ",".join(jar_list))
@@ -217,16 +218,12 @@ def run_pipeline(
         # Native format("clickhouse") needs clickhouse-spark-runtime; JDBC needs shaded clickhouse-jdbc (*-all),
         # not the thin Maven artifact: thin JAR lacks HttpClient 5 (ClassicHttpRequest).
         # Override: BATCH_SPARK_CLICKHOUSE_PACKAGES=maven coords / https jar URLs (comma-sep) or "" for SPARK_JARS only.
+        # Empty/unset: rely on $SPARK_HOME/jars (analytics-runner image). Do not add spark.jars /
+        # spark.jars.packages for ClickHouse here — that breaks K8s executors (./basename.jar).
+        # For ad-hoc runs without the image, set e.g.
+        # BATCH_SPARK_CLICKHOUSE_PACKAGES=com.clickhouse.spark:clickhouse-spark-runtime-3.5_2.12:0.8.0,https://repo1.maven.org/maven2/com/clickhouse/clickhouse-jdbc/0.9.8/clickhouse-jdbc-0.9.8-all.jar
         _raw_ch = os.environ.get("BATCH_SPARK_CLICKHOUSE_PACKAGES")
-        if _raw_ch is None:
-            # Shaded *-all.jar; pin matches Docker image / SPARK_JARS (0.9.x; no 0.10.x on Central for this artifact).
-            _ch_jdbc = os.environ.get("BATCH_CLICKHOUSE_JDBC_VERSION", "0.9.8").strip()
-            ch_pkgs = (
-                "com.clickhouse.spark:clickhouse-spark-runtime-3.5_2.12:0.8.0,"
-                f"https://repo1.maven.org/maven2/com/clickhouse/clickhouse-jdbc/{_ch_jdbc}/"
-                f"clickhouse-jdbc-{_ch_jdbc}-all.jar"
-            )
-        elif not _raw_ch.strip():
+        if _raw_ch is None or not _raw_ch.strip():
             ch_pkgs = None
         else:
             ch_pkgs = _raw_ch.strip()

{batch_analytics-0.3.0 → batch_analytics-0.3.2}/src/batch_analytics.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: batch-analytics
-Version: 0.3.0
+Version: 0.3.2
 Summary: PySpark batch analytics: Extract, Transform, Stage, and analytical modules (linear regression, correlation, PCA, t-test).
 Author: Litewave Analytics Team
 License: MIT