Suppress testDiscoveryUsesNodesFromLastClusterState again (#95414)

This test was reinstated in #95409 but in fact it sometimes fails, and
that is in fact the desired behaviour. With this commit we move it to
the suite of tests that intentionally do not work with atomic register
coordinators.

Author: DaveCTurner

server/src/test/java/org/elasticsearch/cluster/coordination/CoordinatorTests.java

@@ -58,7 +58,6 @@
 
 import static java.util.Collections.emptyMap;
 import static org.elasticsearch.cluster.coordination.AbstractCoordinatorTestCase.Cluster.DEFAULT_DELAY_VARIABILITY;
-import static org.elasticsearch.cluster.coordination.AbstractCoordinatorTestCase.Cluster.EXTREME_DELAY_VARIABILITY;
 import static org.elasticsearch.cluster.coordination.Coordinator.Mode.CANDIDATE;
 import static org.elasticsearch.cluster.coordination.Coordinator.PUBLISH_TIMEOUT_SETTING;
 import static org.elasticsearch.cluster.coordination.ElectionSchedulerFactory.ELECTION_INITIAL_TIMEOUT_SETTING;
@@ -1422,28 +1421,6 @@ public void assertMatched() {
         }
     }
 
-    public void testDiscoveryUsesNodesFromLastClusterState() {
-        try (Cluster cluster = new Cluster(randomIntBetween(3, 5))) {
-            cluster.runRandomly();
-            cluster.stabilise();
-
-            final ClusterNode partitionedNode = cluster.getAnyNode();
-            if (randomBoolean()) {
-                logger.info("--> blackholing {}", partitionedNode);
-                partitionedNode.blackhole();
-            } else {
-                logger.info("--> disconnecting {}", partitionedNode);
-                partitionedNode.disconnect();
-            }
-            cluster.setEmptySeedHostsList();
-            cluster.stabilise();
-
-            partitionedNode.heal();
-            cluster.runRandomly(false, true, EXTREME_DELAY_VARIABILITY);
-            cluster.stabilise();
-        }
-    }
-
     public void testFollowerRemovedIfUnableToSendRequestsToMaster() {
         try (Cluster cluster = new Cluster(3)) {
             cluster.runRandomly();

server/src/test/java/org/elasticsearch/cluster/coordination/CoordinatorVotingConfigurationTests.java

@@ -22,6 +22,7 @@
 import java.util.stream.Collectors;
 
 import static org.elasticsearch.cluster.coordination.AbstractCoordinatorTestCase.Cluster.DEFAULT_DELAY_VARIABILITY;
+import static org.elasticsearch.cluster.coordination.AbstractCoordinatorTestCase.Cluster.EXTREME_DELAY_VARIABILITY;
 import static org.elasticsearch.cluster.coordination.Coordinator.Mode.CANDIDATE;
 import static org.elasticsearch.cluster.coordination.ElectionSchedulerFactory.ELECTION_INITIAL_TIMEOUT_SETTING;
 import static org.elasticsearch.cluster.coordination.Reconfigurator.CLUSTER_AUTO_SHRINK_VOTING_CONFIGURATION;
@@ -488,4 +489,35 @@ public void testClusterUUIDLogging() {
             mockAppender.assertAllExpectationsMatched();
         }
     }
+
+    public void testDiscoveryUsesNodesFromLastClusterState() {
+
+        // This test only applies when using proper majority-based voting configurations. If we permit non-overlapping configurations then
+        // the isolated node can form a one-node cluster when it is healed, and the other nodes won't know be able to discover it without a
+        // proper discovery config since it wasn't in their last cluster states.
+        //
+        // (if we wanted this to work without a proper discovery config then we could e.g. record the master's publish address in the
+        // heartbeat blob, but we can assume that discovery is properly configured when using an atomic-register coordinator, and therefore
+        // there's no need for this extra complexity)
+
+        try (Cluster cluster = new Cluster(randomIntBetween(3, 5))) {
+            cluster.runRandomly();
+            cluster.stabilise();
+
+            final Cluster.ClusterNode partitionedNode = cluster.getAnyNode();
+            if (randomBoolean()) {
+                logger.info("--> blackholing {}", partitionedNode);
+                partitionedNode.blackhole();
+            } else {
+                logger.info("--> disconnecting {}", partitionedNode);
+                partitionedNode.disconnect();
+            }
+            cluster.setEmptySeedHostsList();
+            cluster.stabilise();
+
+            partitionedNode.heal();
+            cluster.runRandomly(false, true, EXTREME_DELAY_VARIABILITY);
+            cluster.stabilise();
+        }
+    }
 }