HBASE-22322 Use special pause for CallQueueTooBigException

Author: Apache9

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncAdminBuilder.java

@@ -36,34 +36,40 @@ public interface AsyncAdminBuilder {
    * Set timeout for a whole admin operation. Operation timeout and max attempt times(or max retry
    * times) are both limitations for retrying, we will stop retrying when we reach any of the
    * limitations.
-   * @param timeout
-   * @param unit
    * @return this for invocation chaining
    */
   AsyncAdminBuilder setOperationTimeout(long timeout, TimeUnit unit);
 
   /**
    * Set timeout for each rpc request.
-   * @param timeout
-   * @param unit
    * @return this for invocation chaining
    */
   AsyncAdminBuilder setRpcTimeout(long timeout, TimeUnit unit);
 
   /**
    * Set the base pause time for retrying. We use an exponential policy to generate sleep time when
    * retrying.
-   * @param timeout
-   * @param unit
    * @return this for invocation chaining
+   * @see #setRetryPauseForCQTBE(long, TimeUnit)
    */
   AsyncAdminBuilder setRetryPause(long timeout, TimeUnit unit);
 
+  /**
+   * Set the base pause time for retrying when we hit {@code CallQueueTooBigException}. We use an
+   * exponential policy to generate sleep time when retrying.
+   * <p/>
+   * This value should be greater than the normal pause value which could be set with the above
+   * {@link #setRetryPause(long, TimeUnit)} method, as usually {@code CallQueueTooBigException}
+   * means the server is overloaded. We just use the normal pause value for
+   * {@code CallQueueTooBigException} if here you specify a smaller value.
+   * @see #setRetryPause(long, TimeUnit)
+   */
+  AsyncAdminBuilder setRetryPauseForCQTBE(long pause, TimeUnit unit);
+
   /**
    * Set the max retry times for an admin operation. Usually it is the max attempt times minus 1.
    * Operation timeout and max attempt times(or max retry times) are both limitations for retrying,
    * we will stop retrying when we reach any of the limitations.
-   * @param maxRetries
    * @return this for invocation chaining
    */
   default AsyncAdminBuilder setMaxRetries(int maxRetries) {
@@ -74,14 +80,12 @@ default AsyncAdminBuilder setMaxRetries(int maxRetries) {
    * Set the max attempt times for an admin operation. Usually it is the max retry times plus 1.
    * Operation timeout and max attempt times(or max retry times) are both limitations for retrying,
    * we will stop retrying when we reach any of the limitations.
-   * @param maxAttempts
    * @return this for invocation chaining
    */
   AsyncAdminBuilder setMaxAttempts(int maxAttempts);
 
   /**
    * Set the number of retries that are allowed before we start to log.
-   * @param startLogErrorsCnt
    * @return this for invocation chaining
    */
   AsyncAdminBuilder setStartLogErrorsCnt(int startLogErrorsCnt);

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncAdminBuilderBase.java

@@ -33,6 +33,8 @@ abstract class AsyncAdminBuilderBase implements AsyncAdminBuilder {
 
   protected long pauseNs;
 
+  protected long pauseForCQTBENs;
+
   protected int maxAttempts;
 
   protected int startLogErrorsCnt;
@@ -41,6 +43,7 @@ abstract class AsyncAdminBuilderBase implements AsyncAdminBuilder {
     this.rpcTimeoutNs = connConf.getRpcTimeoutNs();
     this.operationTimeoutNs = connConf.getOperationTimeoutNs();
     this.pauseNs = connConf.getPauseNs();
+    this.pauseForCQTBENs = connConf.getPauseForCQTBENs();
     this.maxAttempts = connConf.getMaxRetries();
     this.startLogErrorsCnt = connConf.getStartLogErrorsCnt();
   }
@@ -63,6 +66,12 @@ public AsyncAdminBuilder setRetryPause(long timeout, TimeUnit unit) {
     return this;
   }
 
+  @Override
+  public AsyncAdminBuilder setRetryPauseForCQTBE(long timeout, TimeUnit unit) {
+    this.pauseForCQTBENs = unit.toNanos(timeout);
+    return this;
+  }
+
   @Override
   public AsyncAdminBuilder setMaxAttempts(int maxAttempts) {
     this.maxAttempts = maxAttempts;

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncAdminRequestRetryingCaller.java

@@ -44,10 +44,10 @@ public interface Callable<T> {
   private ServerName serverName;
 
   public AsyncAdminRequestRetryingCaller(Timer retryTimer, AsyncConnectionImpl conn, int priority,
-      long pauseNs, int maxAttempts, long operationTimeoutNs, long rpcTimeoutNs,
-      int startLogErrorsCnt, ServerName serverName, Callable<T> callable) {
-    super(retryTimer, conn, priority, pauseNs, maxAttempts, operationTimeoutNs, rpcTimeoutNs,
-      startLogErrorsCnt);
+      long pauseNs, long pauseForCQTBENs, int maxAttempts, long operationTimeoutNs,
+      long rpcTimeoutNs, int startLogErrorsCnt, ServerName serverName, Callable<T> callable) {
+    super(retryTimer, conn, priority, pauseNs, pauseForCQTBENs, maxAttempts, operationTimeoutNs,
+      rpcTimeoutNs, startLogErrorsCnt);
     this.serverName = serverName;
     this.callable = callable;
   }

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncBatchRpcRetryingCaller.java

@@ -45,6 +45,7 @@
 import java.util.stream.Collectors;
 import java.util.stream.Stream;
 import org.apache.commons.lang3.mutable.MutableBoolean;
+import org.apache.hadoop.hbase.CallQueueTooBigException;
 import org.apache.hadoop.hbase.CellScannable;
 import org.apache.hadoop.hbase.DoNotRetryIOException;
 import org.apache.hadoop.hbase.HConstants;
@@ -103,6 +104,8 @@ class AsyncBatchRpcRetryingCaller<T> {
 
   private final long pauseNs;
 
+  private final long pauseForCQTBENs;
+
   private final int maxAttempts;
 
   private final long operationTimeoutNs;
@@ -147,17 +150,17 @@ public int getPriority() {
   }
 
   public AsyncBatchRpcRetryingCaller(Timer retryTimer, AsyncConnectionImpl conn,
-      TableName tableName, List<? extends Row> actions, long pauseNs, int maxAttempts,
-      long operationTimeoutNs, long rpcTimeoutNs, int startLogErrorsCnt) {
+      TableName tableName, List<? extends Row> actions, long pauseNs, long pauseForCQTBENs,
+      int maxAttempts, long operationTimeoutNs, long rpcTimeoutNs, int startLogErrorsCnt) {
     this.retryTimer = retryTimer;
     this.conn = conn;
     this.tableName = tableName;
     this.pauseNs = pauseNs;
+    this.pauseForCQTBENs = pauseForCQTBENs;
     this.maxAttempts = maxAttempts;
     this.operationTimeoutNs = operationTimeoutNs;
     this.rpcTimeoutNs = rpcTimeoutNs;
     this.startLogErrorsCnt = startLogErrorsCnt;
-
     this.actions = new ArrayList<>(actions.size());
     this.futures = new ArrayList<>(actions.size());
     this.action2Future = new IdentityHashMap<>(actions.size());
@@ -337,7 +340,7 @@ private void onComplete(Map<byte[], RegionRequest> actionsByRegion, int tries,
       }
     });
     if (!failedActions.isEmpty()) {
-      tryResubmit(failedActions.stream(), tries, retryImmediately.booleanValue());
+      tryResubmit(failedActions.stream(), tries, retryImmediately.booleanValue(), false);
     }
   }
 
@@ -442,24 +445,27 @@ private void onError(Map<byte[], RegionRequest> actionsByRegion, int tries, Thro
     List<Action> copiedActions = actionsByRegion.values().stream().flatMap(r -> r.actions.stream())
       .collect(Collectors.toList());
     addError(copiedActions, error, serverName);
-    tryResubmit(copiedActions.stream(), tries, error instanceof RetryImmediatelyException);
+    tryResubmit(copiedActions.stream(), tries, error instanceof RetryImmediatelyException,
+      error instanceof CallQueueTooBigException);
   }
 
-  private void tryResubmit(Stream<Action> actions, int tries, boolean immediately) {
+  private void tryResubmit(Stream<Action> actions, int tries, boolean immediately,
+      boolean isCallQueueTooBig) {
     if (immediately) {
       groupAndSend(actions, tries);
       return;
     }
     long delayNs;
+    long pauseNsToUse = isCallQueueTooBig ? pauseForCQTBENs : pauseNs;
     if (operationTimeoutNs > 0) {
       long maxDelayNs = remainingTimeNs() - SLEEP_DELTA_NS;
       if (maxDelayNs <= 0) {
         failAll(actions, tries);
         return;
       }
-      delayNs = Math.min(maxDelayNs, getPauseTime(pauseNs, tries - 1));
+      delayNs = Math.min(maxDelayNs, getPauseTime(pauseNsToUse, tries - 1));
     } else {
-      delayNs = getPauseTime(pauseNs, tries - 1);
+      delayNs = getPauseTime(pauseNsToUse, tries - 1);
     }
     retryTimer.newTimeout(t -> groupAndSend(actions, tries + 1), delayNs, TimeUnit.NANOSECONDS);
   }
@@ -498,7 +504,7 @@ private void groupAndSend(Stream<Action> actions, int tries) {
           sendOrDelay(actionsByServer, tries);
         }
         if (!locateFailed.isEmpty()) {
-          tryResubmit(locateFailed.stream(), tries, false);
+          tryResubmit(locateFailed.stream(), tries, false, false);
         }
       });
   }

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncClientScanner.java

@@ -73,6 +73,8 @@ class AsyncClientScanner {
 
   private final long pauseNs;
 
+  private final long pauseForCQTBENs;
+
   private final int maxAttempts;
 
   private final long scanTimeoutNs;
@@ -84,8 +86,8 @@ class AsyncClientScanner {
   private final ScanResultCache resultCache;
 
   public AsyncClientScanner(Scan scan, AdvancedScanResultConsumer consumer, TableName tableName,
-      AsyncConnectionImpl conn, Timer retryTimer, long pauseNs, int maxAttempts, long scanTimeoutNs,
-      long rpcTimeoutNs, int startLogErrorsCnt) {
+      AsyncConnectionImpl conn, Timer retryTimer, long pauseNs, long pauseForCQTBENs,
+      int maxAttempts, long scanTimeoutNs, long rpcTimeoutNs, int startLogErrorsCnt) {
     if (scan.getStartRow() == null) {
       scan.withStartRow(EMPTY_START_ROW, scan.includeStartRow());
     }
@@ -98,6 +100,7 @@ public AsyncClientScanner(Scan scan, AdvancedScanResultConsumer consumer, TableN
     this.conn = conn;
     this.retryTimer = retryTimer;
     this.pauseNs = pauseNs;
+    this.pauseForCQTBENs = pauseForCQTBENs;
     this.maxAttempts = maxAttempts;
     this.scanTimeoutNs = scanTimeoutNs;
     this.rpcTimeoutNs = rpcTimeoutNs;
@@ -160,14 +163,16 @@ private CompletableFuture<OpenScannerResponse> callOpenScanner(HBaseRpcControlle
   }
 
   private void startScan(OpenScannerResponse resp) {
-    addListener(conn.callerFactory.scanSingleRegion().id(resp.resp.getScannerId())
-      .location(resp.loc).remote(resp.isRegionServerRemote)
-      .scannerLeaseTimeoutPeriod(resp.resp.getTtl(), TimeUnit.MILLISECONDS).stub(resp.stub)
-      .setScan(scan).metrics(scanMetrics).consumer(consumer).resultCache(resultCache)
-      .rpcTimeout(rpcTimeoutNs, TimeUnit.NANOSECONDS)
-      .scanTimeout(scanTimeoutNs, TimeUnit.NANOSECONDS).pause(pauseNs, TimeUnit.NANOSECONDS)
-      .maxAttempts(maxAttempts).startLogErrorsCnt(startLogErrorsCnt)
-      .start(resp.controller, resp.resp), (hasMore, error) -> {
+    addListener(
+      conn.callerFactory.scanSingleRegion().id(resp.resp.getScannerId()).location(resp.loc)
+        .remote(resp.isRegionServerRemote)
+        .scannerLeaseTimeoutPeriod(resp.resp.getTtl(), TimeUnit.MILLISECONDS).stub(resp.stub)
+        .setScan(scan).metrics(scanMetrics).consumer(consumer).resultCache(resultCache)
+        .rpcTimeout(rpcTimeoutNs, TimeUnit.NANOSECONDS)
+        .scanTimeout(scanTimeoutNs, TimeUnit.NANOSECONDS).pause(pauseNs, TimeUnit.NANOSECONDS)
+        .pauseForCQTBE(pauseForCQTBENs, TimeUnit.NANOSECONDS).maxAttempts(maxAttempts)
+        .startLogErrorsCnt(startLogErrorsCnt).start(resp.controller, resp.resp),
+      (hasMore, error) -> {
         if (error != null) {
           consumer.onError(error);
           return;
@@ -185,8 +190,8 @@ private CompletableFuture<OpenScannerResponse> openScanner(int replicaId) {
       .row(scan.getStartRow()).replicaId(replicaId).locateType(getLocateType(scan))
       .rpcTimeout(rpcTimeoutNs, TimeUnit.NANOSECONDS)
       .operationTimeout(scanTimeoutNs, TimeUnit.NANOSECONDS).pause(pauseNs, TimeUnit.NANOSECONDS)
-      .maxAttempts(maxAttempts).startLogErrorsCnt(startLogErrorsCnt).action(this::callOpenScanner)
-      .call();
+      .pauseForCQTBE(pauseForCQTBENs, TimeUnit.NANOSECONDS).maxAttempts(maxAttempts)
+      .startLogErrorsCnt(startLogErrorsCnt).action(this::callOpenScanner).call();
   }
 
   private long getPrimaryTimeoutNs() {

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncConnectionConfiguration.java

@@ -30,6 +30,7 @@
 import static org.apache.hadoop.hbase.HConstants.HBASE_CLIENT_META_REPLICA_SCAN_TIMEOUT_DEFAULT;
 import static org.apache.hadoop.hbase.HConstants.HBASE_CLIENT_OPERATION_TIMEOUT;
 import static org.apache.hadoop.hbase.HConstants.HBASE_CLIENT_PAUSE;
+import static org.apache.hadoop.hbase.HConstants.HBASE_CLIENT_PAUSE_FOR_CQTBE;
 import static org.apache.hadoop.hbase.HConstants.HBASE_CLIENT_RETRIES_NUMBER;
 import static org.apache.hadoop.hbase.HConstants.HBASE_CLIENT_SCANNER_CACHING;
 import static org.apache.hadoop.hbase.HConstants.HBASE_CLIENT_SCANNER_MAX_RESULT_SIZE_KEY;
@@ -54,13 +55,17 @@
 import java.util.concurrent.TimeUnit;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.yetus.audience.InterfaceAudience;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
 
 /**
  * Timeout configs.
  */
 @InterfaceAudience.Private
 class AsyncConnectionConfiguration {
 
+  private static final Logger LOG = LoggerFactory.getLogger(AsyncConnectionConfiguration.class);
+
   private final long metaOperationTimeoutNs;
 
   // timeout for a whole operation such as get, put or delete. Notice that scan will not be effected
@@ -79,6 +84,8 @@ class AsyncConnectionConfiguration {
 
   private final long pauseNs;
 
+  private final long pauseForCQTBENs;
+
   private final int maxRetries;
 
   /** How many retries are allowed before we start to log */
@@ -121,8 +128,16 @@ class AsyncConnectionConfiguration {
       TimeUnit.MILLISECONDS.toNanos(conf.getLong(HBASE_RPC_READ_TIMEOUT_KEY, rpcTimeoutNs));
     this.writeRpcTimeoutNs =
       TimeUnit.MILLISECONDS.toNanos(conf.getLong(HBASE_RPC_WRITE_TIMEOUT_KEY, rpcTimeoutNs));
-    this.pauseNs =
-      TimeUnit.MILLISECONDS.toNanos(conf.getLong(HBASE_CLIENT_PAUSE, DEFAULT_HBASE_CLIENT_PAUSE));
+    long pauseMs = conf.getLong(HBASE_CLIENT_PAUSE, DEFAULT_HBASE_CLIENT_PAUSE);
+    long pauseForCQTBEMs = conf.getLong(HBASE_CLIENT_PAUSE_FOR_CQTBE, pauseMs);
+    if (pauseForCQTBEMs < pauseMs) {
+      LOG.warn(
+        "The {} setting: {} ms is less than the {} setting: {} ms, use the greater one instead",
+        HBASE_CLIENT_PAUSE_FOR_CQTBE, pauseForCQTBEMs, HBASE_CLIENT_PAUSE, pauseMs);
+      pauseForCQTBEMs = pauseMs;
+    }
+    this.pauseNs = TimeUnit.MILLISECONDS.toNanos(pauseMs);
+    this.pauseForCQTBENs = TimeUnit.MILLISECONDS.toNanos(pauseForCQTBEMs);
     this.maxRetries = conf.getInt(HBASE_CLIENT_RETRIES_NUMBER, DEFAULT_HBASE_CLIENT_RETRIES_NUMBER);
     this.startLogErrorsCnt =
       conf.getInt(START_LOG_ERRORS_AFTER_COUNT_KEY, DEFAULT_START_LOG_ERRORS_AFTER_COUNT);
@@ -173,6 +188,10 @@ long getPauseNs() {
     return pauseNs;
   }
 
+  long getPauseForCQTBENs() {
+    return pauseForCQTBENs;
+  }
+
   int getMaxRetries() {
     return maxRetries;
   }

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncMasterRequestRpcRetryingCaller.java

@@ -44,10 +44,10 @@ public interface Callable<T> {
   private final Callable<T> callable;
 
   public AsyncMasterRequestRpcRetryingCaller(Timer retryTimer, AsyncConnectionImpl conn,
-      Callable<T> callable, int priority, long pauseNs, int maxRetries, long operationTimeoutNs,
-      long rpcTimeoutNs, int startLogErrorsCnt) {
-    super(retryTimer, conn, priority, pauseNs, maxRetries, operationTimeoutNs, rpcTimeoutNs,
-      startLogErrorsCnt);
+      Callable<T> callable, int priority, long pauseNs, long pauseForCQTBENs, int maxRetries,
+      long operationTimeoutNs, long rpcTimeoutNs, int startLogErrorsCnt) {
+    super(retryTimer, conn, priority, pauseNs, pauseForCQTBENs, maxRetries, operationTimeoutNs,
+      rpcTimeoutNs, startLogErrorsCnt);
     this.callable = callable;
   }
 

hbase-client/src/main/java/org/apache/hadoop/hbase/client/AsyncRpcRetryingCaller.java

@@ -31,6 +31,7 @@
 import java.util.concurrent.TimeUnit;
 import java.util.function.Consumer;
 import java.util.function.Supplier;
+import org.apache.hadoop.hbase.CallQueueTooBigException;
 import org.apache.hadoop.hbase.DoNotRetryIOException;
 import org.apache.hadoop.hbase.NotServingRegionException;
 import org.apache.hadoop.hbase.TableName;
@@ -59,6 +60,8 @@ public abstract class AsyncRpcRetryingCaller<T> {
 
   private final long pauseNs;
 
+  private final long pauseForCQTBENs;
+
   private int tries = 1;
 
   private final int maxAttempts;
@@ -78,12 +81,13 @@ public abstract class AsyncRpcRetryingCaller<T> {
   protected final HBaseRpcController controller;
 
   public AsyncRpcRetryingCaller(Timer retryTimer, AsyncConnectionImpl conn, int priority,
-      long pauseNs, int maxAttempts, long operationTimeoutNs, long rpcTimeoutNs,
-      int startLogErrorsCnt) {
+      long pauseNs, long pauseForCQTBENs, int maxAttempts, long operationTimeoutNs,
+      long rpcTimeoutNs, int startLogErrorsCnt) {
     this.retryTimer = retryTimer;
     this.conn = conn;
     this.priority = priority;
     this.pauseNs = pauseNs;
+    this.pauseForCQTBENs = pauseForCQTBENs;
     this.maxAttempts = maxAttempts;
     this.operationTimeoutNs = operationTimeoutNs;
     this.rpcTimeoutNs = rpcTimeoutNs;
@@ -123,16 +127,17 @@ protected final void resetCallTimeout() {
   }
 
   private void tryScheduleRetry(Throwable error, Consumer<Throwable> updateCachedLocation) {
+    long pauseNsToUse = error instanceof CallQueueTooBigException ? pauseForCQTBENs : pauseNs;
     long delayNs;
     if (operationTimeoutNs > 0) {
       long maxDelayNs = remainingTimeNs() - SLEEP_DELTA_NS;
       if (maxDelayNs <= 0) {
         completeExceptionally();
         return;
       }
-      delayNs = Math.min(maxDelayNs, getPauseTime(pauseNs, tries - 1));
+      delayNs = Math.min(maxDelayNs, getPauseTime(pauseNsToUse, tries - 1));
     } else {
-      delayNs = getPauseTime(pauseNs, tries - 1);
+      delayNs = getPauseTime(pauseNsToUse, tries - 1);
     }
     tries++;
     retryTimer.newTimeout(t -> doCall(), delayNs, TimeUnit.NANOSECONDS);