SOLR-12801: Harden SimSolrCloudTests.

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimSolrCloudTestCase.java

@@ -16,13 +16,10 @@
  */
 package org.apache.solr.cloud.autoscaling.sim;
 
-import static org.apache.solr.common.cloud.ZkStateReader.SOLR_AUTOSCALING_CONF_PATH;
-
 import java.io.IOException;
 import java.lang.invoke.MethodHandles;
 import java.util.ArrayList;
 import java.util.Collections;
-import java.util.HashMap;
 import java.util.List;
 import java.util.NoSuchElementException;
 import java.util.concurrent.TimeUnit;
@@ -34,14 +31,10 @@ import org.apache.solr.client.solrj.cloud.autoscaling.NotEmptyException;
 import org.apache.solr.common.cloud.DocCollection;
 import org.apache.solr.common.cloud.Replica;
 import org.apache.solr.common.cloud.Slice;
-import org.apache.solr.common.cloud.ZkNodeProps;
-import org.apache.solr.common.cloud.ZkStateReader;
 import org.apache.solr.common.util.TimeSource;
-import org.apache.solr.common.util.Utils;
 import org.apache.solr.util.TimeOut;
 import org.apache.zookeeper.KeeperException;
 import org.junit.AfterClass;
-import org.junit.Before;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
@@ -88,38 +81,6 @@ public class SimSolrCloudTestCase extends SolrTestCaseJ4 {
     super.setUp();
   }
 
-  @Before
-  public void checkClusterConfiguration() throws Exception {
-    if (cluster == null)
-      throw new RuntimeException("SimCloudManager not configured - have you called configureCluster()?");
-    // clear any persisted configuration
-    cluster.getDistribStateManager().setData(SOLR_AUTOSCALING_CONF_PATH, Utils.toJSON(new ZkNodeProps()), -1);
-    cluster.getDistribStateManager().setData(ZkStateReader.ROLES, Utils.toJSON(new HashMap<>()), -1);
-    cluster.getSimNodeStateProvider().simRemoveDeadNodes();
-    cluster.getSimClusterStateProvider().simRemoveDeadNodes();
-    // restore the expected number of nodes
-    int currentSize = cluster.getLiveNodesSet().size();
-    if (currentSize < clusterNodeCount) {
-      int addCnt = clusterNodeCount - currentSize;
-      while (addCnt-- > 0) {
-        cluster.simAddNode();
-      }
-    } else if (currentSize > clusterNodeCount) {
-      cluster.simRemoveRandomNodes(currentSize - clusterNodeCount, true, random());
-    }
-    // clean any persisted trigger state or events
-    removeChildren(ZkStateReader.SOLR_AUTOSCALING_EVENTS_PATH);
-    removeChildren(ZkStateReader.SOLR_AUTOSCALING_TRIGGER_STATE_PATH);
-    removeChildren(ZkStateReader.SOLR_AUTOSCALING_NODE_LOST_PATH);
-    removeChildren(ZkStateReader.SOLR_AUTOSCALING_NODE_ADDED_PATH);
-    cluster.getSimClusterStateProvider().simResetLeaderThrottles();
-    cluster.simRestartOverseer(null);
-    cluster.getSimClusterStateProvider().simDeleteAllCollections();
-    cluster.simClearSystemCollection();
-    cluster.getTimeSource().sleep(10000);
-    cluster.simResetOpCounts();
-  }
-
   protected void removeChildren(String path) throws Exception {
     
     TimeOut timeOut = new TimeOut(10, TimeUnit.SECONDS, TimeSource.NANO_TIME);

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimComputePlanAction.java

@@ -52,7 +52,6 @@ import org.apache.solr.core.SolrResourceLoader;
 import org.apache.solr.util.LogLevel;
 import org.junit.After;
 import org.junit.Before;
-import org.junit.BeforeClass;
 import org.junit.Test;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -73,14 +72,9 @@ public class TestSimComputePlanAction extends SimSolrCloudTestCase {
   private static final AtomicReference<Map> actionContextPropsRef = new AtomicReference<>();
   private static final AtomicReference<TriggerEvent> eventRef = new AtomicReference<>();
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
-    configureCluster(1, TimeSource.get("simTime:50"));
-  }
-
   @Before
   public void init() throws Exception {
-
+    configureCluster(1, TimeSource.get("simTime:50"));
     fired.set(false);
     triggerFiredLatch = new CountDownLatch(1);
     actionContextPropsRef.set(null);
@@ -118,7 +112,7 @@ public class TestSimComputePlanAction extends SimSolrCloudTestCase {
     for (String coll: cluster.getSimClusterStateProvider().simListCollections()) {
       log.info("* Collection " + coll + " state: " + state.getCollection(coll));
     }
-
+    shutdownCluster();
   }
 
   @Test

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimExecutePlanAction.java

@@ -26,7 +26,6 @@ import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicBoolean;
 import java.util.stream.Collectors;
 
-import com.google.common.collect.Lists;
 import org.apache.lucene.util.LuceneTestCase;
 import org.apache.solr.client.solrj.SolrClient;
 import org.apache.solr.client.solrj.SolrRequest;
@@ -44,15 +43,17 @@ import org.apache.solr.common.cloud.Replica;
 import org.apache.solr.common.cloud.ZkStateReader;
 import org.apache.solr.common.params.CollectionParams;
 import org.apache.solr.common.util.NamedList;
+import org.apache.solr.common.util.TimeSource;
 import org.apache.solr.common.util.Utils;
 import org.apache.solr.util.LogLevel;
-import org.apache.solr.common.util.TimeSource;
 import org.junit.After;
-import org.junit.BeforeClass;
+import org.junit.Before;
 import org.junit.Test;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import com.google.common.collect.Lists;
+
 /**
  * Test for {@link ExecutePlanAction}
  */
@@ -62,8 +63,8 @@ public class TestSimExecutePlanAction extends SimSolrCloudTestCase {
 
   private static final int NODE_COUNT = 2;
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
+  @Before
+  public void setupCluster() throws Exception {
     configureCluster(NODE_COUNT, TimeSource.get("simTime:50"));
   }
 
@@ -76,7 +77,7 @@ public class TestSimExecutePlanAction extends SimSolrCloudTestCase {
     for (String coll: cluster.getSimClusterStateProvider().simListCollections()) {
       log.info("* Collection " + coll + " state: " + state.getCollection(coll));
     }
-
+    shutdownCluster();
   }
 
   @Test

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimExtremeIndexing.java

@@ -16,13 +16,13 @@
  */
 package org.apache.solr.cloud.autoscaling.sim;
 
+import static org.apache.solr.cloud.autoscaling.AutoScalingHandlerTest.createAutoScalingRequest;
+
 import java.lang.invoke.MethodHandles;
 import java.util.Iterator;
 import java.util.Locale;
 import java.util.concurrent.TimeUnit;
 
-import com.carrotsearch.randomizedtesting.annotations.TimeoutSuite;
-
 import org.apache.solr.client.solrj.SolrClient;
 import org.apache.solr.client.solrj.SolrRequest;
 import org.apache.solr.client.solrj.request.CollectionAdminRequest;
@@ -37,13 +37,14 @@ import org.apache.solr.common.params.CommonParams;
 import org.apache.solr.common.util.NamedList;
 import org.apache.solr.common.util.TimeSource;
 import org.apache.solr.util.LogLevel;
+import org.junit.After;
 import org.junit.AfterClass;
-import org.junit.BeforeClass;
+import org.junit.Before;
 import org.junit.Test;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
-import static org.apache.solr.cloud.autoscaling.AutoScalingHandlerTest.createAutoScalingRequest;
+import com.carrotsearch.randomizedtesting.annotations.TimeoutSuite;
 
 /**
  *
@@ -75,8 +76,8 @@ public class TestSimExtremeIndexing extends SimSolrCloudTestCase {
   private static TimeSource timeSource;
   private static SolrClient solrClient;
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
+  @Before
+  public void setupCluster() throws Exception {
     configureCluster(NUM_NODES, TimeSource.get("simTime:" + SPEED));
     timeSource = cluster.getTimeSource();
     solrClient = cluster.simGetSolrClient();
@@ -87,6 +88,11 @@ public class TestSimExtremeIndexing extends SimSolrCloudTestCase {
   public static void tearDownCluster() throws Exception {
     solrClient = null;
   }
+  
+  @After
+  public void afterTest() throws Exception {
+    shutdownCluster();
+  }
 
   @Test
   public void testScaleUp() throws Exception {

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimLargeCluster.java

@@ -62,7 +62,6 @@ import org.apache.solr.util.LogLevel;
 import org.apache.solr.util.TimeOut;
 import org.junit.After;
 import org.junit.Before;
-import org.junit.BeforeClass;
 import org.junit.Test;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -74,7 +73,7 @@ import org.slf4j.LoggerFactory;
 public class TestSimLargeCluster extends SimSolrCloudTestCase {
   private static final Logger log = LoggerFactory.getLogger(MethodHandles.lookup().lookupClass());
 
-  public static final int SPEED = 50;
+  public static final int SPEED = 100;
 
   public static final int NUM_NODES = 100;
  
@@ -85,19 +84,15 @@ public class TestSimLargeCluster extends SimSolrCloudTestCase {
   static CountDownLatch triggerFinishedLatch;
   static int waitForSeconds;
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
-    configureCluster(NUM_NODES, TimeSource.get("simTime:" + SPEED));
-  }
-
   @After
   public void tearDownTest() throws Exception {
     shutdownCluster();
-    configureCluster(NUM_NODES, TimeSource.get("simTime:" + SPEED));
   }
   
   @Before
   public void setupTest() throws Exception {
+    configureCluster(NUM_NODES, TimeSource.get("simTime:" + SPEED));
+    
     waitForSeconds = 5;
     triggerStartedCount.set(0);
     triggerFinishedCount.set(0);
@@ -351,11 +346,16 @@ public class TestSimLargeCluster extends SimSolrCloudTestCase {
     log.info("Ready after " + CloudTestUtils.waitForState(cluster, collectionName, 20 * NUM_NODES, TimeUnit.SECONDS,
         CloudTestUtils.clusterShape(NUM_NODES / 10, NUM_NODES / 8 * 3, false, true)) + " ms");
 
-    int count = 50;
+    int count = 1000;
     SolrInputDocument finishedEvent = null;
     long lastNumOps = cluster.simGetOpCount("MOVEREPLICA");
     while (count-- > 0) {
       cluster.getTimeSource().sleep(10000);
+      
+      if (cluster.simGetOpCount("MOVEREPLICA") < 2) {
+        continue;
+      }
+      
       long currentNumOps = cluster.simGetOpCount("MOVEREPLICA");
       if (currentNumOps == lastNumOps) {
         int size = systemColl.size() - 1;
@@ -432,7 +432,6 @@ public class TestSimLargeCluster extends SimSolrCloudTestCase {
           if (cluster != null) {
             cluster.close();
           }
-          setupCluster();
           setUp();
           setupTest();
           long total = doTestNodeLost(wait, delay * 1000, 0);
@@ -699,9 +698,9 @@ public class TestSimLargeCluster extends SimSolrCloudTestCase {
     // wait for listener to capture the SUCCEEDED stage
     cluster.getTimeSource().sleep(25000);
     
-    assertNotNull(listenerEvents.entrySet().toString(), listenerEvents.get("srt"));
+    assertNotNull(listenerEvents.entrySet().iterator().toString(), listenerEvents.get("srt"));
 
-    assertTrue(listenerEvents.entrySet().toString(), listenerEvents.get("srt").size() >= 1);
+    assertTrue(listenerEvents.entrySet().iterator().toString(), listenerEvents.get("srt").size() >= 1);
 
     CapturedEvent ev = listenerEvents.get("srt").get(0);
     assertEquals(TriggerEventType.SEARCHRATE, ev.event.getEventType());

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimNodeAddedTrigger.java

@@ -36,8 +36,8 @@ import org.apache.solr.cloud.autoscaling.TriggerEvent;
 import org.apache.solr.cloud.autoscaling.TriggerValidationException;
 import org.apache.solr.common.util.TimeSource;
 import org.apache.solr.core.SolrResourceLoader;
+import org.junit.After;
 import org.junit.Before;
-import org.junit.BeforeClass;
 import org.junit.Test;
 
 /**
@@ -60,18 +60,20 @@ public class TestSimNodeAddedTrigger extends SimSolrCloudTestCase {
 
   private static TimeSource timeSource;
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
-    configureCluster(1, TimeSource.get("simTime:" + SPEED));
-    timeSource = cluster.getTimeSource();
-  }
-
   @Before
   public void beforeTest() throws Exception {
+    configureCluster(1, TimeSource.get("simTime:" + SPEED));
+    timeSource = cluster.getTimeSource();
+    
     actionConstructorCalled = new AtomicBoolean(false);
     actionInitCalled = new AtomicBoolean(false);
     actionCloseCalled = new AtomicBoolean(false);
   }
+  
+  @After
+  public void afterTest() throws Exception {
+    shutdownCluster();
+  }
 
   @Test
   public void testTrigger() throws Exception {

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimNodeLostTrigger.java

@@ -37,8 +37,8 @@ import org.apache.solr.cloud.autoscaling.TriggerEvent;
 import org.apache.solr.cloud.autoscaling.TriggerValidationException;
 import org.apache.solr.common.util.TimeSource;
 import org.apache.solr.core.SolrResourceLoader;
+import org.junit.After;
 import org.junit.Before;
-import org.junit.BeforeClass;
 import org.junit.Test;
 
 /**
@@ -60,18 +60,19 @@ public class TestSimNodeLostTrigger extends SimSolrCloudTestCase {
   // currentTimeMillis is not as precise so to avoid false positives while comparing time of fire, we add some delta
   private static final long WAIT_FOR_DELTA_NANOS = TimeUnit.MILLISECONDS.toNanos(5);
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
-    configureCluster(5, TimeSource.get("simTime:" + SPEED));
-    timeSource = cluster.getTimeSource();
-  }
-
   @Before
   public void beforeTest() throws Exception {
+    configureCluster(5, TimeSource.get("simTime:" + SPEED));
+    timeSource = cluster.getTimeSource();
     actionConstructorCalled = new AtomicBoolean(false);
     actionInitCalled = new AtomicBoolean(false);
     actionCloseCalled = new AtomicBoolean(false);
   }
+  
+  @After
+  public void afterTest() throws Exception {
+    shutdownCluster();
+  }
 
   @Test
   public void testTrigger() throws Exception {

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimPolicyCloud.java

@@ -16,6 +16,8 @@
  */
 package org.apache.solr.cloud.autoscaling.sim;
 
+import static org.apache.solr.cloud.autoscaling.AutoScalingHandlerTest.createAutoScalingRequest;
+
 import java.io.IOException;
 import java.lang.invoke.MethodHandles;
 import java.util.ArrayList;
@@ -47,13 +49,12 @@ import org.apache.solr.common.util.TimeSource;
 import org.apache.solr.common.util.Utils;
 import org.apache.solr.util.LogLevel;
 import org.apache.zookeeper.KeeperException;
-import org.junit.BeforeClass;
+import org.junit.After;
+import org.junit.Before;
 import org.junit.rules.ExpectedException;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
-import static org.apache.solr.cloud.autoscaling.AutoScalingHandlerTest.createAutoScalingRequest;
-
 @LogLevel("org.apache.solr.cloud.autoscaling=DEBUG")
 public class TestSimPolicyCloud extends SimSolrCloudTestCase {
 
@@ -61,10 +62,15 @@ public class TestSimPolicyCloud extends SimSolrCloudTestCase {
   @org.junit.Rule
   public ExpectedException expectedException = ExpectedException.none();
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
+  @Before
+  public void setupCluster() throws Exception {
     configureCluster(5, TimeSource.get("simTime:50"));
   }
+  
+  @After
+  public void afterTest() throws Exception {
+    shutdownCluster();
+  }
 
   public void testDataProviderPerReplicaDetails() throws Exception {
     SolrClient solrClient = cluster.simGetSolrClient();

solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/TestSimTriggerIntegration.java

@@ -68,9 +68,8 @@ import org.apache.solr.core.SolrResourceLoader;
 import org.apache.solr.util.LogLevel;
 import org.apache.solr.util.TimeOut;
 import org.apache.zookeeper.KeeperException;
-import org.junit.AfterClass;
+import org.junit.After;
 import org.junit.Before;
-import org.junit.BeforeClass;
 import org.junit.Test;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -103,14 +102,9 @@ public class TestSimTriggerIntegration extends SimSolrCloudTestCase {
   private static final long WAIT_FOR_DELTA_NANOS = TimeUnit.MILLISECONDS.toNanos(5);
 
 
-  @BeforeClass
-  public static void setupCluster() throws Exception {
-    configureCluster(2, TimeSource.get("simTime:" + SPEED));
-  }
-  
-  @AfterClass
-  public static void teardownCluster() {
-    cluster.simClearSystemCollection();
+  @After
+  public void afterTest() throws Exception {
+    shutdownCluster();
   }
 
   private static CountDownLatch getTriggerFiredLatch() {
@@ -131,6 +125,8 @@ public class TestSimTriggerIntegration extends SimSolrCloudTestCase {
 
   @Before
   public void setupTest() throws Exception {
+    configureCluster(2, TimeSource.get("simTime:" + SPEED));
+    
     // disable .scheduled_maintenance
     String suspendTriggerCommand = "{" +
         "'suspend-trigger' : {'name' : '.scheduled_maintenance'}" +
@@ -157,14 +153,6 @@ public class TestSimTriggerIntegration extends SimSolrCloudTestCase {
     triggerFinishedCount = new AtomicInteger();
     events.clear();
     listenerEvents.clear();
-    cluster.getLiveNodesSet().removeAllLiveNodesListeners();
-    while (cluster.getClusterStateProvider().getLiveNodes().size() < 2) {
-      // perhaps a test stopped a node but didn't start it back
-      // lets start a node
-      cluster.simAddNode();
-      cluster.getTimeSource().sleep(1000);
-    }
-    cluster.getTimeSource().sleep(10000);
   }
 
   @Test