Hbase0.96源代码之HMaster(三)Hmaster主要循环

妖狐艹你老母 2022-08-27 07:58 177阅读 0赞

# 1，Master初始化 #

# 1.1 #

if (!this.stopped) {
        finishInitialization(startupStatus, false);
        loop();
      }

1.2 finishInitialization()，这里主要完成master组件components：filesystem manager，servermanager，assignmentmanager，regionservertracker, catalogtracker等  
1.2.1 设置为active master

isActiveMaster = true;

1.2.2 生成MasterFileSystem（fileSystemManager）,管理一些文件系统的操作，包含deleteregion,deletetable,modifyColumn,addColumn,deleteFamilyFromFS以及splitlog，splitMetaLog等  
1.2.2.1 check rootdir以及temp目录都存在，且清除temp目录

this.oldLogDir = createInitialFileSystemLayout();

1.2.2.1.1clean tempdir

// check if the root directory exists checkRootDir(this.rootdir, conf, this.fs);
    
    // check if temp directory exists and clean it checkTempDir(this.tempdir, conf, this.fs);

1.2.2.1.2在checkRootDir（）中检测meta的

// Make sure the meta region directory exists!
        if (!FSUtils.metaRegionExists(fs, rd)) {
          bootstrap(rd, c);
        } else {
          // Migrate table descriptor files if necessary
          org.apache.hadoop.hbase.util.FSTableDescriptorMigrationToSubdir
            .migrateFSTableDescriptorsIfNecessary(fs, rd);
        }

1.2.2.1.2.1 如果meta数据目录不存在，创建meta?

private static void bootstrap(final Path rd, final Configuration c)
      throws IOException {
        LOG.info("BOOTSTRAP: creating hbase:meta region");
        try {
          // Bootstrapping, make sure blockcache is off. Else, one will be
          // created here in bootstrap and it'll need to be cleaned up. Better to
          // not make it in first place. Turn off block caching for bootstrap.
          // Enable after.
          HRegionInfo metaHRI = new HRegionInfo(HRegionInfo.FIRST_META_REGIONINFO);
          setInfoFamilyCachingForMeta(false);
          HRegion meta = HRegion.createHRegion(metaHRI, rd, c,
              HTableDescriptor.META_TABLEDESC);
          setInfoFamilyCachingForMeta(true);
          HRegion.closeHRegion(meta);
        } catch (IOException e) {
          e = RemoteExceptionHandler.checkIOException(e);
          LOG.error("bootstrap", e);
          throw e;
        }
      }

1.2.2.2 splitLogManager的创建

this.splitLogManager = new SplitLogManager(master.getZooKeeper(),
          master.getConfiguration(), master, services,
          master.getServerName(), masterRecovery);

1.2.3 创建executorService与serverManager，其中executorService维护一个ExecutorMap，EventType与Executor，提交不对应的EventHandler来异步处理事件；serverManager管理regionserver

if (!masterRecovery) {
      this.executorService = new ExecutorService(getServerName().toShortString());
      this.serverManager = createServerManager(this, this);
    }

1.2.4 initializeZKBasedSystemTrackers  
1.2.4.1 create CatalogTracker 监听-ROOT-和.META.的Server地址信息变化以及事件处理  
1.2.4.2 LoadBalancerTracker  
1.2.4.3 AssignmentManager管理和分配region,监听zk上关于region的event,根据event来完成region的上下线  
1.2.4.4 RegionServerTracker watch zk中中regionSerrver的变化及事件处理，下线  
1.2.4.5 DrainingServerTracker  
1.2.4.6 SnapshotManager Snapshot快照管理祥光

void initializeZKBasedSystemTrackers() throws IOException,
          InterruptedException, KeeperException {
        this.catalogTracker = createCatalogTracker(this.zooKeeper, this.conf, this);
        this.catalogTracker.start();
    
        this.balancer = LoadBalancerFactory.getLoadBalancer(conf);
        this.loadBalancerTracker = new LoadBalancerTracker(zooKeeper, this);
        this.loadBalancerTracker.start();
        this.assignmentManager = new AssignmentManager(this, serverManager,
          this.catalogTracker, this.balancer, this.executorService, this.metricsMaster,
          this.tableLockManager);
        zooKeeper.registerListenerFirst(assignmentManager);
    
        this.regionServerTracker = new RegionServerTracker(zooKeeper, this,
            this.serverManager);
        this.regionServerTracker.start();
    
        this.drainingServerTracker = new DrainingServerTracker(zooKeeper, this,
          this.serverManager);
        this.drainingServerTracker.start();
    
        // Set the cluster as up. If new RSs, they'll be waiting on this before
        // going ahead with their startup.
        boolean wasUp = this.clusterStatusTracker.isClusterUp();
        if (!wasUp) this.clusterStatusTracker.setClusterUp();
    
        LOG.info("Server active/primary master=" + this.serverName +
            ", sessionid=0x" +
            Long.toHexString(this.zooKeeper.getRecoverableZooKeeper().getSessionId()) +
            ", setting cluster-up flag (Was=" + wasUp + ")");
    
        // create the snapshot manager
        this.snapshotManager = new SnapshotManager(this, this.metricsMaster);
      }