[SPARK-23000][TEST] Keep Derby DB Location Unchanged After Session Cl…

…oning ## What changes were proposed in this pull request? After session cloning in `TestHive`, the conf of the singleton SparkContext for derby DB location is changed to a new directory. The new directory is created in `HiveUtils.newTemporaryConfiguration(useInMemoryDerby = false)`. This PR is to keep the conf value of `ConfVars.METASTORECONNECTURLKEY.varname` unchanged during the session clone. ## How was this patch tested? The issue can be reproduced by the command: > build/sbt -Phive "hive/test-only org.apache.spark.sql.hive.HiveSessionStateSuite org.apache.spark.sql.hive.DataSourceWithHiveMetastoreCatalogSuite" Also added a test case. Author: gatorsmile <[email protected]> Closes #20328 from gatorsmile/fixTestFailure.
apache · Jan 19, 2018 · 6c39654 · 6c39654
1 parent e1c33b6
commit 6c39654
Show file tree

Hide file tree

Showing 3 changed files with 21 additions and 8 deletions.
diff --git a/sql/core/src/test/scala/org/apache/spark/sql/SessionStateSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/SessionStateSuite.scala
@@ -17,8 +17,6 @@
 
 package org.apache.spark.sql
 
-import org.scalatest.BeforeAndAfterAll
-import org.scalatest.BeforeAndAfterEach
 import scala.collection.mutable.ArrayBuffer
 
 import org.apache.spark.SparkFunSuite
@@ -28,8 +26,7 @@ import org.apache.spark.sql.catalyst.rules.Rule
 import org.apache.spark.sql.execution.QueryExecution
 import org.apache.spark.sql.util.QueryExecutionListener
 
-class SessionStateSuite extends SparkFunSuite
-    with BeforeAndAfterEach with BeforeAndAfterAll {
+class SessionStateSuite extends SparkFunSuite {
 
   /**
    * A shared SparkSession for all tests in this suite. Make sure you reset any changes to this

diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala
@@ -180,7 +180,13 @@ private[hive] class TestHiveSparkSession(
       ConfVars.METASTORE_INTEGER_JDO_PUSHDOWN.varname -> "true",
       // scratch directory used by Hive's metastore client
       ConfVars.SCRATCHDIR.varname -> TestHiveContext.makeScratchDir().toURI.toString,
-      ConfVars.METASTORE_CLIENT_CONNECT_RETRY_DELAY.varname -> "1")
+      ConfVars.METASTORE_CLIENT_CONNECT_RETRY_DELAY.varname -> "1") ++
+      // After session cloning, the JDBC connect string for a JDBC metastore should not be changed.
+      existingSharedState.map { state =>
+        val connKey =
+          state.sparkContext.hadoopConfiguration.get(ConfVars.METASTORECONNECTURLKEY.varname)
+        ConfVars.METASTORECONNECTURLKEY.varname -> connKey
+      }
 
     metastoreTempConf.foreach { case (k, v) =>
       sc.hadoopConfiguration.set(k, v)

diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveSessionStateSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveSessionStateSuite.scala
@@ -17,16 +17,15 @@
 
 package org.apache.spark.sql.hive
 
-import org.scalatest.BeforeAndAfterEach
+import org.apache.hadoop.hive.conf.HiveConf.ConfVars
 
 import org.apache.spark.sql._
 import org.apache.spark.sql.hive.test.TestHiveSingleton
 
 /**
  * Run all tests from `SessionStateSuite` with a Hive based `SessionState`.
  */
-class HiveSessionStateSuite extends SessionStateSuite
-  with TestHiveSingleton with BeforeAndAfterEach {
+class HiveSessionStateSuite extends SessionStateSuite with TestHiveSingleton {
 
   override def beforeAll(): Unit = {
     // Reuse the singleton session
@@ -39,4 +38,15 @@ class HiveSessionStateSuite extends SessionStateSuite
     activeSession = null
     super.afterAll()
   }
+
+  test("Clone then newSession") {
+    val sparkSession = hiveContext.sparkSession
+    val conf = sparkSession.sparkContext.hadoopConfiguration
+    val oldValue = conf.get(ConfVars.METASTORECONNECTURLKEY.varname)
+    sparkSession.cloneSession()
+    sparkSession.sharedState.externalCatalog.client.newSession()
+    val newValue = conf.get(ConfVars.METASTORECONNECTURLKEY.varname)
+    assert(oldValue == newValue,
+      "cloneSession and then newSession should not affect the Derby directory")
+  }
 }