Extend duration of fixLag() (#34364)

Today, fixLag() waits for a new cluster state to be committed. However, it does
not account for the fact that a term bump may occur, requiring a new election
to take place after the cluster state is committed. This change fixes this.

server/src/main/java/org/elasticsearch/cluster/coordination/Coordinator.java

@@ -63,7 +63,6 @@ import java.util.List;
 import java.util.Optional;
 import java.util.Random;
 import java.util.Set;
-import java.util.concurrent.atomic.AtomicLong;
 import java.util.function.Supplier;
 import java.util.stream.Collectors;
 
@@ -103,7 +102,7 @@ public class Coordinator extends AbstractLifecycleComponent implements Discovery
     private Releasable prevotingRound;
     @Nullable
     private Releasable leaderCheckScheduler;
-    private AtomicLong maxTermSeen = new AtomicLong();
+    private long maxTermSeen;
 
     private Mode mode;
     private Optional<DiscoveryNode> lastKnownLeader;
@@ -259,13 +258,19 @@ public class Coordinator extends AbstractLifecycleComponent implements Discovery
     }
 
     private void updateMaxTermSeen(final long term) {
-        final long updatedMaxTermSeen = maxTermSeen.updateAndGet(oldMaxTerm -> Math.max(oldMaxTerm, term));
         synchronized (mutex) {
-            if (mode == Mode.LEADER && publicationInProgress() == false && updatedMaxTermSeen > getCurrentTerm()) {
+            maxTermSeen = Math.max(maxTermSeen, term);
+            final long currentTerm = getCurrentTerm();
+            if (mode == Mode.LEADER && maxTermSeen > currentTerm) {
                 // Bump our term. However if there is a publication in flight then doing so would cancel the publication, so don't do that
                 // since we check whether a term bump is needed at the end of the publication too.
-                ensureTermAtLeast(getLocalNode(), updatedMaxTermSeen);
+                if (publicationInProgress()) {
-                startElection();
+                    logger.debug("updateMaxTermSeen: maxTermSeen = {} > currentTerm = {}, enqueueing term bump",
+                        maxTermSeen, currentTerm);
+                } else {
+                    ensureTermAtLeast(getLocalNode(), maxTermSeen);
+                    startElection();
+                }
             }
         }
     }
@@ -276,7 +281,7 @@ public class Coordinator extends AbstractLifecycleComponent implements Discovery
             // to check our mode again here.
             if (mode == Mode.CANDIDATE) {
                 final StartJoinRequest startJoinRequest
-                    = new StartJoinRequest(getLocalNode(), Math.max(getCurrentTerm(), maxTermSeen.get()) + 1);
+                    = new StartJoinRequest(getLocalNode(), Math.max(getCurrentTerm(), maxTermSeen) + 1);
                 logger.debug("starting election with {}", startJoinRequest);
                 getDiscoveredNodes().forEach(node -> joinHelper.sendStartJoinRequest(startJoinRequest, node));
             }

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

@@ -705,7 +705,10 @@ public class CoordinatorTests extends ESTestCase {
                         leader.submitValue(randomLong());
                     }
                 }).run();
-                runFor(DEFAULT_CLUSTER_STATE_UPDATE_DELAY, "re-stabilising after lag-fixing publication");
+                runFor(DEFAULT_CLUSTER_STATE_UPDATE_DELAY
+                    // may need to bump terms too
+                    + DEFAULT_ELECTION_DELAY,
+                    "re-stabilising after lag-fixing publication");
             } else {
                 logger.info("--> fixLag found no lag, leader={}, leaderVersion={}, minVersion={}", leader, leaderVersion, minVersion);
             }