✅ Close Terracotta-OSS#191: Failover tests

management/monitoring-service/src/main/java/org/terracotta/management/service/monitoring/MonitoringServiceProvider.java

@@ -179,6 +179,8 @@ public <T> T getService(long consumerID, ServiceConfiguration<T> configuration)
     // get or creates a client-side monitoring service
     if (ClientMonitoringService.class == serviceType) {
       if (configuration instanceof ClientMonitoringServiceConfiguration) {
+        // this guard prevents an entity using a monitoring service if the platform did not previously sent us the right topology events
+        // this service cannot be used if the topology is not right
         if (!topologyService.isCurrentServerActive()) {
           throw new IllegalStateException("Server " + platformConfiguration.getServerName() + " is not active!");
         }
@@ -200,6 +202,8 @@ public <T> T getService(long consumerID, ServiceConfiguration<T> configuration)
     // get or creates a monitoring accessor service (for tms)
     if (ManagementService.class == serviceType) {
       if (configuration instanceof ManagementServiceConfiguration) {
+        // this guard prevents an entity using a monitoring service if the platform did not previously sent us the right topology events
+        // this service cannot be used if the topology is not right
         if (!topologyService.isCurrentServerActive()) {
           throw new IllegalStateException("Server " + platformConfiguration.getServerName() + " is not active!");
         }
@@ -222,6 +226,8 @@ public <T> T getService(long consumerID, ServiceConfiguration<T> configuration)
     // get or creates a monitoring service for an active entity
     if (ActiveEntityMonitoringService.class == serviceType) {
       if (configuration instanceof ActiveEntityMonitoringServiceConfiguration) {
+        // this guard prevents an entity using a monitoring service if the platform did not previously sent us the right topology events
+        // this service cannot be used if the topology is not right
         if (!topologyService.isCurrentServerActive()) {
           throw new IllegalStateException("Server " + platformConfiguration.getServerName() + " is not active!");
         }
@@ -241,6 +247,8 @@ public <T> T getService(long consumerID, ServiceConfiguration<T> configuration)
     // get or creates a monitoring service for a passive entity, bridging calls to IMonitoringProducer
     if (PassiveEntityMonitoringService.class == serviceType) {
       if (configuration instanceof PassiveEntityMonitoringServiceConfiguration) {
+        // this guard prevents an entity using a monitoring service if the platform did not previously sent us the right topology events
+        // this service cannot be used if the topology is not right
         if (topologyService.isCurrentServerActive()) {
           throw new IllegalStateException("Server " + platformConfiguration.getServerName() + " is not passive!");
         }

management/testing/integration-tests/src/test/java/org/terracotta/management/integration/tests/FailoverIT.java

@@ -15,50 +15,70 @@
  */
 package org.terracotta.management.integration.tests;
 
-import org.junit.Ignore;
+import org.junit.Before;
 import org.junit.Test;
+import org.terracotta.management.model.capabilities.descriptors.Settings;
 import org.terracotta.management.model.cluster.Cluster;
 import org.terracotta.management.model.cluster.Server;
+import org.terracotta.management.model.cluster.ServerEntity;
 import org.terracotta.management.model.message.Message;
 import org.terracotta.management.model.notification.ContextualNotification;
 
 import java.util.Arrays;
 import java.util.List;
 import java.util.stream.Collectors;
+import java.util.stream.Stream;
 
 import static org.hamcrest.CoreMatchers.equalTo;
+import static org.junit.Assert.assertEquals;
 import static org.junit.Assert.assertThat;
 
 /**
  * @author Mathieu Carbou
  */
-@Ignore("Impacted by https://github.com/Terracotta-OSS/terracotta-core/issues/405")
-//TODO: VOLTRON ISSUE ? https://github.com/Terracotta-OSS/terracotta-core/issues/405
 public class FailoverIT extends AbstractHATest {
 
-  @Test
-  public void failover_management() throws Exception {
+  Server oldActive;
+  Server oldPassive;
+
+  @Override
+  @Before
+  public void setUp() throws Exception {
+    super.setUp();
+
     Cluster cluster = tmsAgentService.readTopology();
-    Server active = cluster.serverStream().filter(Server::isActive).findFirst().get();
-    Server passive = cluster.serverStream().filter(server -> !server.isActive()).findFirst().get();
-    assertThat(active.getState(), equalTo(Server.State.ACTIVE));
-    assertThat(passive.getState(), equalTo(Server.State.PASSIVE));
+    oldActive = cluster.serverStream().filter(Server::isActive).findFirst().get();
+    oldPassive = cluster.serverStream().filter(server -> !server.isActive()).findFirst().get();
+    assertThat(oldActive.getState(), equalTo(Server.State.ACTIVE));
+    assertThat(oldPassive.getState(), equalTo(Server.State.PASSIVE));
 
     // clear buffer
     tmsAgentService.readMessages();
 
     // kill active - passive should take the active role
-    System.out.printf("==> terminateActive()");
     voltron.getClusterControl().terminateActive();
-    System.out.printf("==> waitForActive()");
     voltron.getClusterControl().waitForActive();
+  }
 
-    System.out.printf("==> readTopology()");
-    cluster = tmsAgentService.readTopology();
+  @Test
+  public void topology_recovery_after_failover() throws Exception {
+    Cluster cluster = tmsAgentService.readTopology();
+
+    // verify new server
     Server newActive = cluster.serverStream().filter(Server::isActive).findFirst().get();
     assertThat(newActive.getState(), equalTo(Server.State.ACTIVE));
-    assertThat(newActive.getServerName(), equalTo(passive.getServerName()));
+    assertThat(newActive.getServerName(), equalTo(oldPassive.getServerName()));
+
+    // removes all random values
+    String currentTopo = toJson(cluster.toMap()).toString();
+    String actual = removeRandomValues(currentTopo);
 
+    String expected = readJson("topology.json").toString();
+    assertEquals(expected, actual);
+  }
+
+  @Test
+  public void notifications_after_failover() throws Exception {
     // read messages
     List<Message> messages = tmsAgentService.readMessages();
 
@@ -74,7 +94,7 @@ public void failover_management() throws Exception {
 
     assertThat(
         notifs.stream().map(notif -> notif.getContext().get(Server.NAME_KEY)).collect(Collectors.toList()),
-        equalTo(Arrays.asList(newActive.getServerName(), newActive.getServerName())));
+        equalTo(Arrays.asList(oldPassive.getServerName(), oldPassive.getServerName())));
 
     assertThat(
         notifs.stream().map(notif -> notif.getAttributes().get("state")).collect(Collectors.toList()),

management/testing/integration-tests/src/test/java/org/terracotta/management/integration/tests/TopologyIT.java

@@ -42,13 +42,7 @@ public void can_read_topology() throws Exception {
     Cluster cluster = tmsAgentService.readTopology();
     String currentTopo = toJson(cluster.toMap()).toString();
     String actual = removeRandomValues(currentTopo);
-
-
     String expected = readJson("topology.json").toString();
-
-    System.out.println("This is the actual topology : " + actual);
-    System.out.println("This is the expected topology : " + expected);
-    // and compare
     assertEquals(expected, actual);
   }
 

pom.xml

@@ -31,8 +31,8 @@
     <terracotta-apis.version>1.1.0</terracotta-apis.version>
     <tcconfig.version>10.1.0</tcconfig.version>
     <passthrough-testing.version>1.1.1-pre2</passthrough-testing.version>
-    <terracotta-core.version>5.1.1-pre2</terracotta-core.version>
-    <galvan.version>1.1.1-pre2</galvan.version>
+    <terracotta-core.version>5.2-SNAPSHOT</terracotta-core.version>
+    <galvan.version>1.2-SNAPSHOT</galvan.version>
     <statistics.version>1.4.1</statistics.version>
     <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
     <project.reporting.outputEncoding>UTF-8</project.reporting.outputEncoding>