[SPARK-31957][SQL] Cleanup hive scratch dir for the developer api sta…

…rtWithContext
apache · Jun 10, 2020 · d055d60 · d055d60
1 parent 43063e2
commit d055d60
Show file tree

Hide file tree

Showing 3 changed files with 25 additions and 13 deletions.
diff --git a/...hriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2.scala b/...hriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2.scala
@@ -52,14 +52,17 @@ object HiveThriftServer2 extends Logging {
    */
   @DeveloperApi
   def startWithContext(sqlContext: SQLContext): HiveThriftServer2 = {
-    val server = new HiveThriftServer2(sqlContext)
-
     val executionHive = HiveUtils.newClientForExecution(
       sqlContext.sparkContext.conf,
       sqlContext.sessionState.newHadoopConf())
 
+    // Cleanup the scratch dir before starting
+    ServerUtils.cleanUpScratchDir(executionHive.conf)
+    val server = new HiveThriftServer2(sqlContext)
+
     server.init(executionHive.conf)
     server.start()
+    logInfo("HiveThriftServer2 started")
     createListenerAndUI(server, sqlContext.sparkContext)
     server
   }
@@ -97,18 +100,8 @@ object HiveThriftServer2 extends Logging {
       uiTab.foreach(_.detach())
     }
 
-    val executionHive = HiveUtils.newClientForExecution(
-      SparkSQLEnv.sqlContext.sparkContext.conf,
-      SparkSQLEnv.sqlContext.sessionState.newHadoopConf())
-
     try {
-      // Cleanup the scratch dir before starting
-      ServerUtils.cleanUpScratchDir(executionHive.conf)
-      val server = new HiveThriftServer2(SparkSQLEnv.sqlContext)
-      server.init(executionHive.conf)
-      server.start()
-      logInfo("HiveThriftServer2 started")
-      createListenerAndUI(server, SparkSQLEnv.sparkContext)
+      startWithContext(SparkSQLEnv.sqlContext)
       // If application was killed before HiveThriftServer2 start successfully then SparkSubmit
       // process can not exit, so check whether if SparkContext was stopped.
       if (SparkSQLEnv.sparkContext.stopped.get()) {

diff --git a/...riftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/SharedThriftServer.scala b/...riftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/SharedThriftServer.scala
@@ -17,6 +17,7 @@
 
 package org.apache.spark.sql.hive.thriftserver
 
+import java.io.File
 import java.sql.{DriverManager, Statement}
 
 import scala.collection.JavaConverters._
@@ -27,12 +28,20 @@ import org.apache.hadoop.hive.conf.HiveConf.ConfVars
 import org.apache.hive.service.cli.thrift.ThriftCLIService
 
 import org.apache.spark.sql.test.SharedSparkSession
+import org.apache.spark.util.Utils
 
 trait SharedThriftServer extends SharedSparkSession {
 
   private var hiveServer2: HiveThriftServer2 = _
   private var serverPort: Int = 0
 
+  protected val tempScratchDir: File = {
+    val dir = Utils.createTempDir()
+    dir.setWritable(true, false)
+    Utils.createTempDir(dir.getAbsolutePath)
+    dir
+  }
+
   def mode: ServerMode.Value
 
   override def beforeAll(): Unit = {
@@ -85,6 +94,9 @@ trait SharedThriftServer extends SharedSparkSession {
     sqlContext.setConf(ConfVars.HIVE_SERVER2_THRIFT_PORT.varname, "0")
     sqlContext.setConf(ConfVars.HIVE_SERVER2_THRIFT_HTTP_PORT.varname, "0")
     sqlContext.setConf(ConfVars.HIVE_SERVER2_TRANSPORT_MODE.varname, mode.toString)
+    sqlContext.setConf(ConfVars.SCRATCHDIR.varname, tempScratchDir.getAbsolutePath)
+    sqlContext.setConf(ConfVars.HIVE_START_CLEANUP_SCRATCHDIR.varname, "true")
+    assert(tempScratchDir.exists())
 
     try {
       hiveServer2 = HiveThriftServer2.startWithContext(sqlContext)
@@ -94,6 +106,9 @@ trait SharedThriftServer extends SharedSparkSession {
           logInfo(s"Started HiveThriftServer2: port=$serverPort, attempt=$attempt")
         case _ =>
       }
+      // the scratch dir will be recreated after the probe sql `SELECT 1` executed, so we
+      // check it here first.
+      assert(!tempScratchDir.exists())
 
       // Wait for thrift server to be ready to serve the query, via executing simple query
       // till the query succeeds. See SPARK-30345 for more details.

diff --git a/...test/scala/org/apache/spark/sql/hive/thriftserver/ThriftServerWithSparkContextSuite.scala b/...test/scala/org/apache/spark/sql/hive/thriftserver/ThriftServerWithSparkContextSuite.scala
@@ -19,6 +19,10 @@ package org.apache.spark.sql.hive.thriftserver
 
 trait ThriftServerWithSparkContextSuite extends SharedThriftServer {
 
+  test("the scratch dir will be deleted during server start but recreated with new operation") {
+    assert(tempScratchDir.exists())
+  }
+
   test("SPARK-29911: Uncache cached tables when session closed") {
     val cacheManager = spark.sharedState.cacheManager
     val globalTempDB = spark.sharedState.globalTempViewManager.database