From b4d67e48372ea670d31170908ca59e6b9c7b7bc6 Mon Sep 17 00:00:00 2001
From: huangzhir <306824224@qq.com>
Date: Thu, 20 Apr 2023 17:54:06 +0800
Subject: [PATCH] [KYUUBI #4713][TEST] Fix false positive result in
 SchedulerPoolSuite (#4714)

### _Why are the changes needed?_

fix issuse https://github.com/apache/kyuubi/issues/4713


### _How was this patch tested?_
- [ ] Add some test cases that check the changes thoroughly including negative and positive cases if possible

- [ ] Add screenshots for manual tests if appropriate

- [X] [Run test](https://kyuubi.readthedocs.io/en/master/develop_tools/testing.html#running-tests) locally before make a pull request

Closes #4714 from huangzhir/fixtest-schedulerpool.

Closes #4713

Authored-by: huangzhir <306824224@qq.com>
Signed-off-by: Cheng Pan <chengpan@apache.org>
---
 .../engine/spark/SchedulerPoolSuite.scala     | 53 ++++++++++---------
 1 file changed, 28 insertions(+), 25 deletions(-)

diff --git a/externals/kyuubi-spark-sql-engine/src/test/scala/org/apache/kyuubi/engine/spark/SchedulerPoolSuite.scala b/externals/kyuubi-spark-sql-engine/src/test/scala/org/apache/kyuubi/engine/spark/SchedulerPoolSuite.scala
index af8c90cf2..43bd3f4db 100644
--- a/externals/kyuubi-spark-sql-engine/src/test/scala/org/apache/kyuubi/engine/spark/SchedulerPoolSuite.scala
+++ b/externals/kyuubi-spark-sql-engine/src/test/scala/org/apache/kyuubi/engine/spark/SchedulerPoolSuite.scala
@@ -19,6 +19,8 @@ package org.apache.kyuubi.engine.spark
 
 import java.util.concurrent.Executors
 
+import scala.concurrent.duration.SECONDS
+
 import org.apache.spark.scheduler.{SparkListener, SparkListenerJobEnd, SparkListenerJobStart}
 import org.scalatest.concurrent.PatienceConfiguration.Timeout
 import org.scalatest.time.SpanSugar.convertIntToGrainOfTime
@@ -76,33 +78,34 @@ class SchedulerPoolSuite extends WithSparkSQLEngine with HiveJDBCTestHelper {
       eventually(Timeout(3.seconds)) {
         assert(job0Started)
       }
-      Seq(1, 0).foreach { priority =>
-        threads.execute(() => {
-          priority match {
-            case 0 =>
-              withJdbcStatement() { statement =>
-                statement.execute("SET kyuubi.operation.scheduler.pool=p0")
-                statement.execute("SELECT java_method('java.lang.Thread', 'sleep', 1500l)" +
-                  "FROM range(1, 3, 1, 2)")
-              }
+      threads.execute(() => {
+        // job name job1
+        withJdbcStatement() { statement =>
+          statement.execute("SET kyuubi.operation.scheduler.pool=p1")
+          statement.execute("SELECT java_method('java.lang.Thread', 'sleep', 1500l)" +
+            " FROM range(1, 3, 1, 2)")
+        }
+      })
+      // make sure job1 started before job2
+      eventually(Timeout(2.seconds)) {
+        assert(job1StartTime > 0)
+      }
 
-            case 1 =>
-              withJdbcStatement() { statement =>
-                statement.execute("SET kyuubi.operation.scheduler.pool=p1")
-                statement.execute("SELECT java_method('java.lang.Thread', 'sleep', 1500l)" +
-                  " FROM range(1, 3, 1, 2)")
-              }
-          }
-        })
-      }
+      threads.execute(() => {
+        // job name job2
+        withJdbcStatement() { statement =>
+          statement.execute("SET kyuubi.operation.scheduler.pool=p0")
+          statement.execute("SELECT java_method('java.lang.Thread', 'sleep', 1500l)" +
+            "FROM range(1, 3, 1, 2)")
+        }
+      })
       threads.shutdown()
-      eventually(Timeout(20.seconds)) {
-        // We can not ensure that job1 is started before job2 so here using abs.
-        assert(Math.abs(job1StartTime - job2StartTime) < 1000)
-        // Job1 minShare is 2(total resource) so that job2 should be allocated tasks after
-        // job1 finished.
-        assert(job2FinishTime - job1FinishTime >= 1000)
-      }
+      threads.awaitTermination(20, SECONDS)
+      // job1 should be started before job2
+      assert(job1StartTime < job2StartTime)
+      // job2 minShare is 2(total resource) so that job1 should be allocated tasks after
+      // job2 finished.
+      assert(job2FinishTime < job1FinishTime)
     } finally {
       spark.sparkContext.removeSparkListener(listener)
     }