HDFS-16902. Add Namenode status to BPServiceActor metrics and improve logging in offerservice (#5334)

Reviewed-by: Mingliang Liu <liuml07@apache.org>
Reviewed-by: Shilun Fan <slfan1989@apache.org>
Signed-off-by: Tao Li <tomscut@apache.org>

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/BPServiceActor.java

@@ -200,6 +200,7 @@ class BPServiceActor implements Runnable {
   Map<String, String> getActorInfoMap() {
     final Map<String, String> info = new HashMap<String, String>();
     info.put("NamenodeAddress", getNameNodeAddress());
+    info.put("NamenodeHaState", state != null ? state.toString() : "Unknown");
     info.put("BlockPoolID", bpos.getBlockPoolId());
     info.put("ActorState", getRunningState());
     info.put("LastHeartbeat",
@@ -685,6 +686,8 @@ class BPServiceActor implements Runnable {
         // Every so often, send heartbeat or block-report
         //
         final boolean sendHeartbeat = scheduler.isHeartbeatDue(startTime);
+        LOG.debug("BP offer service run start time: {}, sendHeartbeat: {}", startTime,
+            sendHeartbeat);
         HeartbeatResponse resp = null;
         if (sendHeartbeat) {
           //
@@ -697,6 +700,8 @@ class BPServiceActor implements Runnable {
           boolean requestBlockReportLease = (fullBlockReportLeaseId == 0) &&
                   scheduler.isBlockReportDue(startTime);
           if (!dn.areHeartbeatsDisabledForTests()) {
+            LOG.debug("Before sending heartbeat to namenode {}, the state of the namenode known"
+                + " to datanode so far is {}", this.getNameNodeAddress(), state);
             resp = sendHeartBeat(requestBlockReportLease);
             assert resp != null;
             if (resp.getFullBlockReportLeaseId() != 0) {
@@ -721,7 +726,12 @@ class BPServiceActor implements Runnable {
             // that we should actually process.
             bpos.updateActorStatesFromHeartbeat(
                 this, resp.getNameNodeHaState());
-            state = resp.getNameNodeHaState().getState();
+            HAServiceState stateFromResp = resp.getNameNodeHaState().getState();
+            if (state != stateFromResp) {
+              LOG.info("After receiving heartbeat response, updating state of namenode {} to {}",
+                  this.getNameNodeAddress(), stateFromResp);
+            }
+            state = stateFromResp;
 
             if (state == HAServiceState.ACTIVE) {
               handleRollingUpgradeStatus(resp);
@@ -781,6 +791,7 @@ class BPServiceActor implements Runnable {
       long sleepTime = Math.min(1000, dnConf.heartBeatInterval);
       Thread.sleep(sleepTime);
     } catch (InterruptedException ie) {
+      LOG.info("BPServiceActor {} is interrupted", this);
       Thread.currentThread().interrupt();
     }
   }
@@ -983,6 +994,8 @@ class BPServiceActor implements Runnable {
     while (!duplicateQueue.isEmpty()) {
       BPServiceActorAction actionItem = duplicateQueue.remove();
       try {
+        LOG.debug("BPServiceActor ( {} ) processing queued messages. Action item: {}", this,
+            actionItem);
         actionItem.reportTo(bpNamenode, bpRegistration);
       } catch (BPServiceActorActionException baae) {
         LOG.warn(baae.getMessage() + nnAddr , baae);

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/ErrorReportAction.java

@@ -20,6 +20,7 @@ package org.apache.hadoop.hdfs.server.datanode;
 
 import java.io.IOException;
 
+import org.apache.commons.lang3.builder.ToStringBuilder;
 import org.apache.hadoop.hdfs.protocolPB.DatanodeProtocolClientSideTranslatorPB;
 import org.apache.hadoop.hdfs.server.protocol.DatanodeRegistration;
 import org.apache.hadoop.ipc.RemoteException;
@@ -84,4 +85,12 @@ public class ErrorReportAction implements BPServiceActorAction {
     }
     return true;
   }
+
+  @Override
+  public String toString() {
+    return new ToStringBuilder(this)
+        .append("errorCode", errorCode)
+        .append("errorMessage", errorMessage)
+        .toString();
+  }
 }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/ReportBadBlockAction.java

@@ -20,6 +20,7 @@ package org.apache.hadoop.hdfs.server.datanode;
 
 import java.io.IOException;
 
+import org.apache.commons.lang3.builder.ToStringBuilder;
 import org.apache.hadoop.fs.StorageType;
 import org.apache.hadoop.hdfs.protocol.DatanodeInfo;
 import org.apache.hadoop.hdfs.protocol.DatanodeInfo.DatanodeInfoBuilder;
@@ -111,4 +112,13 @@ public class ReportBadBlockAction implements BPServiceActorAction {
     }
     return true;
   }
+
+  @Override
+  public String toString() {
+    return new ToStringBuilder(this)
+        .append("block", block)
+        .append("storageUuid", storageUuid)
+        .append("storageType", storageType)
+        .toString();
+  }
 }

hadoop-hdfs-project/hadoop-hdfs/src/main/webapps/datanode/datanode.html

@@ -81,6 +81,7 @@
   <thead>
     <tr>
       <th>Namenode Address</th>
+      <th>Namenode HA State</th>
       <th>Block Pool ID</th>
       <th>Actor State</th>
       <th>Last Heartbeat</th>
@@ -91,6 +92,7 @@
   {#dn.BPServiceActorInfo}
     <tr>
       <td>{NamenodeAddress}</td>
+      <td>{NamenodeHaState}</td>
       <td>{BlockPoolID}</td>
       <td>{ActorState}</td>
       <td>{LastHeartbeat}s</td>