remove async checking of rs online in master and checkAndRecordServer in rsStartup in master

Author: caroliney14

hbase-server/src/main/java/org/apache/hadoop/hbase/executor/EventType.java

@@ -253,13 +253,6 @@ public enum EventType {
    * Master is processing log replay of failed region server
    */
   M_LOG_REPLAY              (74, ExecutorType.M_LOG_REPLAY_OPS),
-  /**
-   * Master controlled events to be executed on the master.<br>
-   *
-   * M_RS_ONLINE<br>
-   * Master has received startup request from region server and is polling for its online status
-   */
-  M_RS_ONLINE               (75, ExecutorType.MASTER_SERVER_OPERATIONS),
 
   /**
    * RS controlled events to be executed on the RS.<br>

hbase-server/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java

@@ -51,20 +51,15 @@
 import org.apache.hadoop.hbase.client.AsyncClusterConnection;
 import org.apache.hadoop.hbase.client.AsyncRegionServerAdmin;
 import org.apache.hadoop.hbase.client.RegionInfo;
-import org.apache.hadoop.hbase.executor.EventHandler;
-import org.apache.hadoop.hbase.executor.EventType;
 import org.apache.hadoop.hbase.ipc.RemoteWithExtrasException;
 import org.apache.hadoop.hbase.master.assignment.RegionStates;
 import org.apache.hadoop.hbase.master.procedure.ServerCrashProcedure;
 import org.apache.hadoop.hbase.monitoring.MonitoredTask;
 import org.apache.hadoop.hbase.procedure2.Procedure;
-import org.apache.hadoop.hbase.shaded.protobuf.generated.AdminProtos;
 import org.apache.hadoop.hbase.util.Bytes;
 import org.apache.hadoop.hbase.util.CommonFSUtils;
 import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
 import org.apache.hadoop.hbase.util.FutureUtils;
-import org.apache.hadoop.hbase.util.RetryCounter;
-import org.apache.hadoop.hbase.util.RetryCounterFactory;
 import org.apache.hadoop.hbase.zookeeper.ZKUtil;
 import org.apache.hadoop.hbase.zookeeper.ZKWatcher;
 import org.apache.hadoop.hbase.zookeeper.ZNodePaths;
@@ -177,8 +172,6 @@ public class ServerManager {
   private final long maxSkew;
   private final long warningSkew;
 
-  private final RetryCounterFactory pingRetryCounterFactory;
-
   /** Listeners that are called on server events. */
   private List<ServerListener> listeners = new CopyOnWriteArrayList<>();
 
@@ -192,11 +185,6 @@ public ServerManager(final MasterServices master) {
     warningSkew = c.getLong("hbase.master.warningclockskew", 10000);
     persistFlushedSequenceId = c.getBoolean(PERSIST_FLUSHEDSEQUENCEID,
         PERSIST_FLUSHEDSEQUENCEID_DEFAULT);
-    int pingMaxAttempts = Math.max(1, master.getConfiguration().getInt(
-      "hbase.master.maximum.ping.server.attempts", 10));
-    int pingSleepInterval = Math.max(1, master.getConfiguration().getInt(
-      "hbase.master.ping.server.retry.sleep.interval", 100));
-    this.pingRetryCounterFactory = new RetryCounterFactory(pingMaxAttempts, pingSleepInterval);
   }
 
   /**
@@ -239,43 +227,9 @@ ServerName regionServerStartup(RegionServerStartupRequest request, int versionNu
     ServerName sn = ServerName.valueOf(hostname, request.getPort(), request.getServerStartCode());
     checkClockSkew(sn, request.getServerCurrentTime());
     checkIsDead(sn, "STARTUP");
-    master.getExecutorService().submit(new RegionServerStartupHandler(sn,
-      ServerMetricsBuilder.of(sn, versionNumber, version)));
     return sn;
   }
 
-  class RegionServerStartupHandler extends EventHandler {
-    final ServerName serverName;
-    final ServerMetrics serverMetrics;
-    final int timeout;
-
-    RegionServerStartupHandler(final ServerName serverName, final ServerMetrics sm) {
-      super(master, EventType.M_RS_ONLINE);
-      this.serverName = serverName;
-      this.serverMetrics = sm;
-      this.timeout = server.getConfiguration().getInt(
-        HConstants.HBASE_RPC_READ_TIMEOUT_KEY, HConstants.DEFAULT_HBASE_RPC_TIMEOUT);
-    }
-    @Override
-    public void process() throws IOException {
-      long startTime = EnvironmentEdgeManager.currentTime();
-      while ((EnvironmentEdgeManager.currentTime() - startTime) < timeout) {
-        if (isServerReachable(this.serverName)) {
-          if (!checkAndRecordNewServer(this.serverName, this.serverMetrics)) {
-            LOG.warn("THIS SHOULD NOT HAPPEN, RegionServerStartup"
-              + " could not record the server: " + this.serverName);
-          }
-          break;
-        }
-        try {
-          Thread.sleep(100);
-        } catch (InterruptedException ie) {
-          LOG.debug("Caught interrupted exception while waiting for regionserver to online", ie);
-        }
-      }
-    }
-  }
-
   /**
    * Updates last flushed sequence Ids for the regions on server sn
    * @param sn
@@ -874,45 +828,6 @@ public void waitForRegionServers(MonitoredTask status) throws InterruptedExcepti
         " master is "+ (this.master.isStopped() ? "stopped.": "running"));
   }
 
-  /**
-   * Check if a region server is reachable and has the expected start code
-   */
-  public boolean isServerReachable(ServerName server) {
-    if (server == null) throw new NullPointerException("Passed server is null");
-
-    RetryCounter retryCounter = pingRetryCounterFactory.create();
-    while (retryCounter.shouldRetry()) {
-      synchronized (this.onlineServers) {
-        if (this.deadservers.isDeadServer(server)) {
-          return false;
-        }
-      }
-      try {
-        AsyncRegionServerAdmin admin = master.getAsyncClusterConnection()
-          .getRegionServerAdmin(server);
-        if (admin != null) {
-          AdminProtos.ServerInfo info = FutureUtils.get(
-            admin.getServerInfo(RequestConverter.buildGetServerInfoRequest())
-          ).getServerInfo();
-          return info != null && info.hasServerName()
-            && server.getStartcode() == info.getServerName().getStartCode();
-        }
-      } catch (IOException ioe) {
-        if (LOG.isDebugEnabled()) {
-          LOG.debug("Couldn't reach " + server + ", try=" + retryCounter.getAttemptTimes() + " of "
-            + retryCounter.getMaxAttempts(), ioe);
-        }
-        try {
-          retryCounter.sleepUntilNextRetry();
-        } catch(InterruptedException ie) {
-          Thread.currentThread().interrupt();
-          break;
-        }
-      }
-    }
-    return false;
-  }
-
   private String getStrForMax(final int max) {
     return max == Integer.MAX_VALUE? "NO_LIMIT": Integer.toString(max);
   }