Halt VM is lock lost but shutdown not complete

DomGarguilo · Dec 23, 2024 · 0d5f014 · 0d5f014
1 parent c09918e
commit 0d5f014
Show file tree

Hide file tree

Showing 2 changed files with 10 additions and 2 deletions.
diff --git a/server/compactor/src/main/java/org/apache/accumulo/compactor/Compactor.java b/server/compactor/src/main/java/org/apache/accumulo/compactor/Compactor.java
@@ -170,6 +170,7 @@ public String getQueueName() {
   private ServerAddress compactorAddress = null;
 
   private final AtomicBoolean compactionRunning = new AtomicBoolean(false);
+  private volatile boolean shutdownComplete = false;
 
   protected Compactor(CompactorServerOpts opts, String[] args) {
     super("compactor", opts, args);
@@ -287,7 +288,9 @@ protected void announceExistence(HostAndPort clientAddress)
     LockWatcher lw = new LockWatcher() {
       @Override
       public void lostLock(final LockLossReason reason) {
-        if (isShutdownRequested()) {
+        // ServiceLock.unlock is called at the end of run(). We don't
+        // want to Halt when we are shutting down.
+        if (isShutdownRequested() && shutdownComplete) {
           LOG.warn("Compactor lost lock (reason = {}), not exiting because shutdown requested.",
               reason);
         } else {
@@ -905,6 +908,7 @@ public void run() {
 
       gcLogger.logGCInfo(getConfiguration());
       LOG.info("stop requested. exiting ... ");
+      shutdownComplete = true;
       try {
         if (null != compactorLock) {
           compactorLock.unlock();

diff --git a/server/tserver/src/main/java/org/apache/accumulo/tserver/TabletServer.java b/server/tserver/src/main/java/org/apache/accumulo/tserver/TabletServer.java
@@ -240,6 +240,7 @@ public TabletServerMinCMetrics getMinCMetrics() {
   private final ZooAuthenticationKeyWatcher authKeyWatcher;
   private final WalStateManager walMarker;
   private final ServerContext context;
+  private volatile boolean shutdownComplete = false;
 
   public static void main(String[] args) throws Exception {
     try (TabletServer tserver = new TabletServer(new ServerOpts(), args)) {
@@ -686,7 +687,9 @@ private void announceExistence() {
 
         @Override
         public void lostLock(final LockLossReason reason) {
-          if (isShutdownRequested()) {
+          // ServiceLock.unlock is called at the end of run(). We don't
+          // want to Halt when we are shutting down.
+          if (isShutdownRequested() && shutdownComplete) {
             LOG.warn(
                 "TabletServer lost lock (reason = {}), not exiting because shutdown requested.",
                 reason);
@@ -1027,6 +1030,7 @@ public void run() {
 
     log.info("TServerInfo: stop requested. exiting ... ");
 
+    shutdownComplete = true;
     try {
       tabletServerLock.unlock();
     } catch (Exception e) {