HBASE-26526 Introduce a timeout to shutdown of WAL (#3297)

Signed-off-by: Andrew Purtell <apurtell@apache.org>
(cherry picked from commit ca3ba49)

Author: sunhelly

hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/AbstractFSWAL.java

@@ -37,14 +37,17 @@
 import java.util.Map;
 import java.util.OptionalLong;
 import java.util.Set;
+import java.util.concurrent.Callable;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.ConcurrentNavigableMap;
 import java.util.concurrent.ConcurrentSkipListMap;
 import java.util.concurrent.CopyOnWriteArrayList;
 import java.util.concurrent.ExecutionException;
 import java.util.concurrent.ExecutorService;
 import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
 import java.util.concurrent.TimeUnit;
+import java.util.concurrent.TimeoutException;
 import java.util.concurrent.atomic.AtomicBoolean;
 import java.util.concurrent.atomic.AtomicInteger;
 import java.util.concurrent.atomic.AtomicLong;
@@ -142,6 +145,9 @@ public abstract class AbstractFSWAL<W extends WriterBase> implements WAL {
   public static final String RING_BUFFER_SLOT_COUNT =
     "hbase.regionserver.wal.disruptor.event.count";
 
+  public static final String WAL_SHUTDOWN_WAIT_TIMEOUT_MS = "hbase.wal.shutdown.wait.timeout.ms";
+  public static final int DEFAULT_WAL_SHUTDOWN_WAIT_TIMEOUT_MS = 15 * 1000;
+
   /**
    * file system instance
    */
@@ -269,6 +275,9 @@ public abstract class AbstractFSWAL<W extends WriterBase> implements WAL {
   protected volatile boolean closed = false;
 
   protected final AtomicBoolean shutdown = new AtomicBoolean(false);
+
+  protected final long walShutdownTimeout;
+
   /**
    * WAL Comparator; it compares the timestamp (log filenum), present in the log file name. Throws
    * an IllegalArgumentException if used to compare paths from different wals.
@@ -320,8 +329,8 @@ public WalProps(Map<byte[], Long> encodedName2HighestSequenceId, long logSize) {
 
   protected final AtomicBoolean rollRequested = new AtomicBoolean(false);
 
-  private final ExecutorService logArchiveExecutor = Executors.newSingleThreadExecutor(
-    new ThreadFactoryBuilder().setDaemon(true).setNameFormat("Log-Archiver-%d").build());
+  private final ExecutorService logArchiveOrShutdownExecutor = Executors.newSingleThreadExecutor(
+    new ThreadFactoryBuilder().setDaemon(true).setNameFormat("WAL-Archive-Or-Shutdown-%d").build());
 
   private final int archiveRetries;
 
@@ -478,7 +487,9 @@ public boolean accept(final Path fileName) {
     this.syncFutureCache = new SyncFutureCache(conf);
     this.implClassName = getClass().getSimpleName();
     this.useHsync = conf.getBoolean(HRegion.WAL_HSYNC_CONF_KEY, HRegion.DEFAULT_WAL_HSYNC);
-    archiveRetries = this.conf.getInt("hbase.regionserver.logroll.archive.retries", 0);
+    archiveRetries = this.conf.getInt("hbase.regionserver.walroll.archive.retries", 0);
+    this.walShutdownTimeout =
+      conf.getLong(WAL_SHUTDOWN_WAIT_TIMEOUT_MS, DEFAULT_WAL_SHUTDOWN_WAIT_TIMEOUT_MS);
   }
 
   /**
@@ -685,7 +696,7 @@ private void cleanOldLogs() throws IOException {
       final List<Pair<Path, Long>> localLogsToArchive = logsToArchive;
       // make it async
       for (Pair<Path, Long> log : localLogsToArchive) {
-        logArchiveExecutor.execute(() -> {
+        logArchiveOrShutdownExecutor.execute(() -> {
           archive(log);
         });
         this.walFile2Props.remove(log.getFirst());
@@ -891,17 +902,42 @@ public void shutdown() throws IOException {
         i.logCloseRequested();
       }
     }
-    rollWriterLock.lock();
-    try {
-      doShutdown();
-      if (syncFutureCache != null) {
-        syncFutureCache.clear();
+
+    Future<Void> future = logArchiveOrShutdownExecutor.submit(new Callable<Void>() {
+      @Override
+      public Void call() throws Exception {
+        if (rollWriterLock.tryLock(walShutdownTimeout, TimeUnit.SECONDS)) {
+          try {
+            doShutdown();
+            if (syncFutureCache != null) {
+              syncFutureCache.clear();
+            }
+          } finally {
+            rollWriterLock.unlock();
+          }
+        } else {
+          throw new IOException("Waiting for rollWriterLock timeout");
+        }
+        return null;
       }
-      if (logArchiveExecutor != null) {
-        logArchiveExecutor.shutdownNow();
+    });
+    logArchiveOrShutdownExecutor.shutdown();
+
+    try {
+      future.get(walShutdownTimeout, TimeUnit.MILLISECONDS);
+    } catch (InterruptedException e) {
+      throw new InterruptedIOException("Interrupted when waiting for shutdown WAL");
+    } catch (TimeoutException e) {
+      throw new TimeoutIOException("We have waited " + walShutdownTimeout + "ms, but"
+        + " the shutdown of WAL doesn't complete! Please check the status of underlying "
+        + "filesystem or increase the wait time by the config \"" + WAL_SHUTDOWN_WAIT_TIMEOUT_MS
+        + "\"", e);
+    } catch (ExecutionException e) {
+      if (e.getCause() instanceof IOException) {
+        throw (IOException) e.getCause();
+      } else {
+        throw new IOException(e.getCause());
       }
-    } finally {
-      rollWriterLock.unlock();
     }
   }