You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@hbase.apache.org by gi...@apache.org on 2018/03/21 14:54:07 UTC

[06/51] [partial] hbase-site git commit: Published site at 8ab7b20f48951d77945181024f5e15842bc253c4.

http://git-wip-us.apache.org/repos/asf/hbase-site/blob/6eb695c8/devapidocs/src-html/org/apache/hadoop/hbase/replication/ZKReplicationQueueStorage.html
----------------------------------------------------------------------
diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/replication/ZKReplicationQueueStorage.html b/devapidocs/src-html/org/apache/hadoop/hbase/replication/ZKReplicationQueueStorage.html
index 7556544..ce54cbb 100644
--- a/devapidocs/src-html/org/apache/hadoop/hbase/replication/ZKReplicationQueueStorage.html
+++ b/devapidocs/src-html/org/apache/hadoop/hbase/replication/ZKReplicationQueueStorage.html
@@ -41,520 +41,530 @@
 <span class="sourceLineNo">033</span>import org.apache.hadoop.fs.Path;<a name="line.33"></a>
 <span class="sourceLineNo">034</span>import org.apache.hadoop.hbase.HConstants;<a name="line.34"></a>
 <span class="sourceLineNo">035</span>import org.apache.hadoop.hbase.ServerName;<a name="line.35"></a>
-<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.exceptions.DeserializationException;<a name="line.36"></a>
-<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.37"></a>
-<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.util.CollectionUtils;<a name="line.38"></a>
-<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.util.Pair;<a name="line.39"></a>
-<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.zookeeper.ZKUtil;<a name="line.40"></a>
-<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.zookeeper.ZKUtil.ZKUtilOp;<a name="line.41"></a>
-<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.zookeeper.ZKWatcher;<a name="line.42"></a>
-<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.zookeeper.ZNodePaths;<a name="line.43"></a>
-<span class="sourceLineNo">044</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.44"></a>
-<span class="sourceLineNo">045</span>import org.apache.zookeeper.KeeperException;<a name="line.45"></a>
-<span class="sourceLineNo">046</span>import org.apache.zookeeper.KeeperException.BadVersionException;<a name="line.46"></a>
-<span class="sourceLineNo">047</span>import org.apache.zookeeper.KeeperException.NoNodeException;<a name="line.47"></a>
-<span class="sourceLineNo">048</span>import org.apache.zookeeper.KeeperException.NodeExistsException;<a name="line.48"></a>
-<span class="sourceLineNo">049</span>import org.apache.zookeeper.KeeperException.NotEmptyException;<a name="line.49"></a>
-<span class="sourceLineNo">050</span>import org.apache.zookeeper.data.Stat;<a name="line.50"></a>
-<span class="sourceLineNo">051</span>import org.slf4j.Logger;<a name="line.51"></a>
-<span class="sourceLineNo">052</span>import org.slf4j.LoggerFactory;<a name="line.52"></a>
-<span class="sourceLineNo">053</span><a name="line.53"></a>
-<span class="sourceLineNo">054</span>import org.apache.hbase.thirdparty.com.google.common.annotations.VisibleForTesting;<a name="line.54"></a>
-<span class="sourceLineNo">055</span><a name="line.55"></a>
-<span class="sourceLineNo">056</span>/**<a name="line.56"></a>
-<span class="sourceLineNo">057</span> * ZK based replication queue storage.<a name="line.57"></a>
-<span class="sourceLineNo">058</span> * &lt;p&gt;<a name="line.58"></a>
-<span class="sourceLineNo">059</span> * The base znode for each regionserver is the regionserver name. For example:<a name="line.59"></a>
-<span class="sourceLineNo">060</span> *<a name="line.60"></a>
-<span class="sourceLineNo">061</span> * &lt;pre&gt;<a name="line.61"></a>
-<span class="sourceLineNo">062</span> * /hbase/replication/rs/hostname.example.org,6020,1234<a name="line.62"></a>
-<span class="sourceLineNo">063</span> * &lt;/pre&gt;<a name="line.63"></a>
-<span class="sourceLineNo">064</span> *<a name="line.64"></a>
-<span class="sourceLineNo">065</span> * Within this znode, the region server maintains a set of WAL replication queues. These queues are<a name="line.65"></a>
-<span class="sourceLineNo">066</span> * represented by child znodes named using there give queue id. For example:<a name="line.66"></a>
-<span class="sourceLineNo">067</span> *<a name="line.67"></a>
-<span class="sourceLineNo">068</span> * &lt;pre&gt;<a name="line.68"></a>
-<span class="sourceLineNo">069</span> * /hbase/replication/rs/hostname.example.org,6020,1234/1<a name="line.69"></a>
-<span class="sourceLineNo">070</span> * /hbase/replication/rs/hostname.example.org,6020,1234/2<a name="line.70"></a>
-<span class="sourceLineNo">071</span> * &lt;/pre&gt;<a name="line.71"></a>
-<span class="sourceLineNo">072</span> *<a name="line.72"></a>
-<span class="sourceLineNo">073</span> * Each queue has one child znode for every WAL that still needs to be replicated. The value of<a name="line.73"></a>
-<span class="sourceLineNo">074</span> * these WAL child znodes is the latest position that has been replicated. This position is updated<a name="line.74"></a>
-<span class="sourceLineNo">075</span> * every time a WAL entry is replicated. For example:<a name="line.75"></a>
-<span class="sourceLineNo">076</span> *<a name="line.76"></a>
-<span class="sourceLineNo">077</span> * &lt;pre&gt;<a name="line.77"></a>
-<span class="sourceLineNo">078</span> * /hbase/replication/rs/hostname.example.org,6020,1234/1/23522342.23422 [VALUE: 254]<a name="line.78"></a>
-<span class="sourceLineNo">079</span> * &lt;/pre&gt;<a name="line.79"></a>
-<span class="sourceLineNo">080</span> */<a name="line.80"></a>
-<span class="sourceLineNo">081</span>@InterfaceAudience.Private<a name="line.81"></a>
-<span class="sourceLineNo">082</span>class ZKReplicationQueueStorage extends ZKReplicationStorageBase<a name="line.82"></a>
-<span class="sourceLineNo">083</span>    implements ReplicationQueueStorage {<a name="line.83"></a>
-<span class="sourceLineNo">084</span><a name="line.84"></a>
-<span class="sourceLineNo">085</span>  private static final Logger LOG = LoggerFactory.getLogger(ZKReplicationQueueStorage.class);<a name="line.85"></a>
-<span class="sourceLineNo">086</span><a name="line.86"></a>
-<span class="sourceLineNo">087</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_KEY =<a name="line.87"></a>
-<span class="sourceLineNo">088</span>      "zookeeper.znode.replication.hfile.refs";<a name="line.88"></a>
-<span class="sourceLineNo">089</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_DEFAULT = "hfile-refs";<a name="line.89"></a>
-<span class="sourceLineNo">090</span><a name="line.90"></a>
-<span class="sourceLineNo">091</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_REGIONS_KEY =<a name="line.91"></a>
-<span class="sourceLineNo">092</span>      "zookeeper.znode.replication.regions";<a name="line.92"></a>
-<span class="sourceLineNo">093</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_REGIONS_DEFAULT = "regions";<a name="line.93"></a>
-<span class="sourceLineNo">094</span><a name="line.94"></a>
-<span class="sourceLineNo">095</span>  /**<a name="line.95"></a>
-<span class="sourceLineNo">096</span>   * The name of the znode that contains all replication queues<a name="line.96"></a>
-<span class="sourceLineNo">097</span>   */<a name="line.97"></a>
-<span class="sourceLineNo">098</span>  private final String queuesZNode;<a name="line.98"></a>
-<span class="sourceLineNo">099</span><a name="line.99"></a>
-<span class="sourceLineNo">100</span>  /**<a name="line.100"></a>
-<span class="sourceLineNo">101</span>   * The name of the znode that contains queues of hfile references to be replicated<a name="line.101"></a>
-<span class="sourceLineNo">102</span>   */<a name="line.102"></a>
-<span class="sourceLineNo">103</span>  private final String hfileRefsZNode;<a name="line.103"></a>
-<span class="sourceLineNo">104</span><a name="line.104"></a>
-<span class="sourceLineNo">105</span>  private final String regionsZNode;<a name="line.105"></a>
-<span class="sourceLineNo">106</span><a name="line.106"></a>
-<span class="sourceLineNo">107</span>  public ZKReplicationQueueStorage(ZKWatcher zookeeper, Configuration conf) {<a name="line.107"></a>
-<span class="sourceLineNo">108</span>    super(zookeeper, conf);<a name="line.108"></a>
-<span class="sourceLineNo">109</span><a name="line.109"></a>
-<span class="sourceLineNo">110</span>    String queuesZNodeName = conf.get("zookeeper.znode.replication.rs", "rs");<a name="line.110"></a>
-<span class="sourceLineNo">111</span>    String hfileRefsZNodeName = conf.get(ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_KEY,<a name="line.111"></a>
-<span class="sourceLineNo">112</span>      ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_DEFAULT);<a name="line.112"></a>
-<span class="sourceLineNo">113</span>    this.queuesZNode = ZNodePaths.joinZNode(replicationZNode, queuesZNodeName);<a name="line.113"></a>
-<span class="sourceLineNo">114</span>    this.hfileRefsZNode = ZNodePaths.joinZNode(replicationZNode, hfileRefsZNodeName);<a name="line.114"></a>
-<span class="sourceLineNo">115</span>    this.regionsZNode = ZNodePaths.joinZNode(replicationZNode, conf<a name="line.115"></a>
-<span class="sourceLineNo">116</span>        .get(ZOOKEEPER_ZNODE_REPLICATION_REGIONS_KEY, ZOOKEEPER_ZNODE_REPLICATION_REGIONS_DEFAULT));<a name="line.116"></a>
-<span class="sourceLineNo">117</span>  }<a name="line.117"></a>
-<span class="sourceLineNo">118</span><a name="line.118"></a>
-<span class="sourceLineNo">119</span>  private String getRsNode(ServerName serverName) {<a name="line.119"></a>
-<span class="sourceLineNo">120</span>    return ZNodePaths.joinZNode(queuesZNode, serverName.getServerName());<a name="line.120"></a>
-<span class="sourceLineNo">121</span>  }<a name="line.121"></a>
-<span class="sourceLineNo">122</span><a name="line.122"></a>
-<span class="sourceLineNo">123</span>  private String getQueueNode(ServerName serverName, String queueId) {<a name="line.123"></a>
-<span class="sourceLineNo">124</span>    return ZNodePaths.joinZNode(getRsNode(serverName), queueId);<a name="line.124"></a>
-<span class="sourceLineNo">125</span>  }<a name="line.125"></a>
-<span class="sourceLineNo">126</span><a name="line.126"></a>
-<span class="sourceLineNo">127</span>  private String getFileNode(String queueNode, String fileName) {<a name="line.127"></a>
-<span class="sourceLineNo">128</span>    return ZNodePaths.joinZNode(queueNode, fileName);<a name="line.128"></a>
-<span class="sourceLineNo">129</span>  }<a name="line.129"></a>
-<span class="sourceLineNo">130</span><a name="line.130"></a>
-<span class="sourceLineNo">131</span>  private String getFileNode(ServerName serverName, String queueId, String fileName) {<a name="line.131"></a>
-<span class="sourceLineNo">132</span>    return getFileNode(getQueueNode(serverName, queueId), fileName);<a name="line.132"></a>
-<span class="sourceLineNo">133</span>  }<a name="line.133"></a>
-<span class="sourceLineNo">134</span><a name="line.134"></a>
-<span class="sourceLineNo">135</span>  /**<a name="line.135"></a>
-<span class="sourceLineNo">136</span>   * Put all regions under /hbase/replication/regions znode will lead to too many children because<a name="line.136"></a>
-<span class="sourceLineNo">137</span>   * of the huge number of regions in real production environment. So here we use hash of encoded<a name="line.137"></a>
-<span class="sourceLineNo">138</span>   * region name to distribute the znode into multiple znodes. &lt;br&gt;<a name="line.138"></a>
-<span class="sourceLineNo">139</span>   * So the final znode path will be format like this:<a name="line.139"></a>
-<span class="sourceLineNo">140</span>   *<a name="line.140"></a>
-<span class="sourceLineNo">141</span>   * &lt;pre&gt;<a name="line.141"></a>
-<span class="sourceLineNo">142</span>   * /hbase/replication/regions/254/dd04e76a6966d4ffa908ed0586764767-100<a name="line.142"></a>
-<span class="sourceLineNo">143</span>   * &lt;/pre&gt;<a name="line.143"></a>
-<span class="sourceLineNo">144</span>   *<a name="line.144"></a>
-<span class="sourceLineNo">145</span>   * The 254 indicate the hash of encoded region name, the 100 indicate the peer id.<a name="line.145"></a>
-<span class="sourceLineNo">146</span>   * @param encodedRegionName the encoded region name.<a name="line.146"></a>
-<span class="sourceLineNo">147</span>   * @param peerId peer id for replication.<a name="line.147"></a>
-<span class="sourceLineNo">148</span>   * @return ZNode path to persist the max sequence id that we've pushed for the given region and<a name="line.148"></a>
-<span class="sourceLineNo">149</span>   *         peer.<a name="line.149"></a>
-<span class="sourceLineNo">150</span>   */<a name="line.150"></a>
-<span class="sourceLineNo">151</span>  private String getSerialReplicationRegionPeerNode(String encodedRegionName, String peerId) {<a name="line.151"></a>
-<span class="sourceLineNo">152</span>    int hash = encodedRegionName.hashCode() &amp; 0x0000FFFF;<a name="line.152"></a>
-<span class="sourceLineNo">153</span>    String hashPath = ZNodePaths.joinZNode(regionsZNode, String.valueOf(hash));<a name="line.153"></a>
-<span class="sourceLineNo">154</span>    return ZNodePaths.joinZNode(hashPath, String.format("%s-%s", encodedRegionName, peerId));<a name="line.154"></a>
-<span class="sourceLineNo">155</span>  }<a name="line.155"></a>
-<span class="sourceLineNo">156</span><a name="line.156"></a>
-<span class="sourceLineNo">157</span>  @Override<a name="line.157"></a>
-<span class="sourceLineNo">158</span>  public void removeQueue(ServerName serverName, String queueId) throws ReplicationException {<a name="line.158"></a>
-<span class="sourceLineNo">159</span>    try {<a name="line.159"></a>
-<span class="sourceLineNo">160</span>      ZKUtil.deleteNodeRecursively(zookeeper, getQueueNode(serverName, queueId));<a name="line.160"></a>
-<span class="sourceLineNo">161</span>    } catch (KeeperException e) {<a name="line.161"></a>
-<span class="sourceLineNo">162</span>      throw new ReplicationException(<a name="line.162"></a>
-<span class="sourceLineNo">163</span>          "Failed to delete queue (serverName=" + serverName + ", queueId=" + queueId + ")", e);<a name="line.163"></a>
-<span class="sourceLineNo">164</span>    }<a name="line.164"></a>
+<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.client.RegionInfo;<a name="line.36"></a>
+<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.exceptions.DeserializationException;<a name="line.37"></a>
+<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.38"></a>
+<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.util.CollectionUtils;<a name="line.39"></a>
+<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.util.Pair;<a name="line.40"></a>
+<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.zookeeper.ZKUtil;<a name="line.41"></a>
+<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.zookeeper.ZKUtil.ZKUtilOp;<a name="line.42"></a>
+<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.zookeeper.ZKWatcher;<a name="line.43"></a>
+<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.zookeeper.ZNodePaths;<a name="line.44"></a>
+<span class="sourceLineNo">045</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.45"></a>
+<span class="sourceLineNo">046</span>import org.apache.zookeeper.KeeperException;<a name="line.46"></a>
+<span class="sourceLineNo">047</span>import org.apache.zookeeper.KeeperException.BadVersionException;<a name="line.47"></a>
+<span class="sourceLineNo">048</span>import org.apache.zookeeper.KeeperException.NoNodeException;<a name="line.48"></a>
+<span class="sourceLineNo">049</span>import org.apache.zookeeper.KeeperException.NodeExistsException;<a name="line.49"></a>
+<span class="sourceLineNo">050</span>import org.apache.zookeeper.KeeperException.NotEmptyException;<a name="line.50"></a>
+<span class="sourceLineNo">051</span>import org.apache.zookeeper.data.Stat;<a name="line.51"></a>
+<span class="sourceLineNo">052</span>import org.slf4j.Logger;<a name="line.52"></a>
+<span class="sourceLineNo">053</span>import org.slf4j.LoggerFactory;<a name="line.53"></a>
+<span class="sourceLineNo">054</span><a name="line.54"></a>
+<span class="sourceLineNo">055</span>import org.apache.hbase.thirdparty.com.google.common.annotations.VisibleForTesting;<a name="line.55"></a>
+<span class="sourceLineNo">056</span><a name="line.56"></a>
+<span class="sourceLineNo">057</span>/**<a name="line.57"></a>
+<span class="sourceLineNo">058</span> * ZK based replication queue storage.<a name="line.58"></a>
+<span class="sourceLineNo">059</span> * &lt;p&gt;<a name="line.59"></a>
+<span class="sourceLineNo">060</span> * The base znode for each regionserver is the regionserver name. For example:<a name="line.60"></a>
+<span class="sourceLineNo">061</span> *<a name="line.61"></a>
+<span class="sourceLineNo">062</span> * &lt;pre&gt;<a name="line.62"></a>
+<span class="sourceLineNo">063</span> * /hbase/replication/rs/hostname.example.org,6020,1234<a name="line.63"></a>
+<span class="sourceLineNo">064</span> * &lt;/pre&gt;<a name="line.64"></a>
+<span class="sourceLineNo">065</span> *<a name="line.65"></a>
+<span class="sourceLineNo">066</span> * Within this znode, the region server maintains a set of WAL replication queues. These queues are<a name="line.66"></a>
+<span class="sourceLineNo">067</span> * represented by child znodes named using there give queue id. For example:<a name="line.67"></a>
+<span class="sourceLineNo">068</span> *<a name="line.68"></a>
+<span class="sourceLineNo">069</span> * &lt;pre&gt;<a name="line.69"></a>
+<span class="sourceLineNo">070</span> * /hbase/replication/rs/hostname.example.org,6020,1234/1<a name="line.70"></a>
+<span class="sourceLineNo">071</span> * /hbase/replication/rs/hostname.example.org,6020,1234/2<a name="line.71"></a>
+<span class="sourceLineNo">072</span> * &lt;/pre&gt;<a name="line.72"></a>
+<span class="sourceLineNo">073</span> *<a name="line.73"></a>
+<span class="sourceLineNo">074</span> * Each queue has one child znode for every WAL that still needs to be replicated. The value of<a name="line.74"></a>
+<span class="sourceLineNo">075</span> * these WAL child znodes is the latest position that has been replicated. This position is updated<a name="line.75"></a>
+<span class="sourceLineNo">076</span> * every time a WAL entry is replicated. For example:<a name="line.76"></a>
+<span class="sourceLineNo">077</span> *<a name="line.77"></a>
+<span class="sourceLineNo">078</span> * &lt;pre&gt;<a name="line.78"></a>
+<span class="sourceLineNo">079</span> * /hbase/replication/rs/hostname.example.org,6020,1234/1/23522342.23422 [VALUE: 254]<a name="line.79"></a>
+<span class="sourceLineNo">080</span> * &lt;/pre&gt;<a name="line.80"></a>
+<span class="sourceLineNo">081</span> */<a name="line.81"></a>
+<span class="sourceLineNo">082</span>@InterfaceAudience.Private<a name="line.82"></a>
+<span class="sourceLineNo">083</span>class ZKReplicationQueueStorage extends ZKReplicationStorageBase<a name="line.83"></a>
+<span class="sourceLineNo">084</span>    implements ReplicationQueueStorage {<a name="line.84"></a>
+<span class="sourceLineNo">085</span><a name="line.85"></a>
+<span class="sourceLineNo">086</span>  private static final Logger LOG = LoggerFactory.getLogger(ZKReplicationQueueStorage.class);<a name="line.86"></a>
+<span class="sourceLineNo">087</span><a name="line.87"></a>
+<span class="sourceLineNo">088</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_KEY =<a name="line.88"></a>
+<span class="sourceLineNo">089</span>      "zookeeper.znode.replication.hfile.refs";<a name="line.89"></a>
+<span class="sourceLineNo">090</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_DEFAULT = "hfile-refs";<a name="line.90"></a>
+<span class="sourceLineNo">091</span><a name="line.91"></a>
+<span class="sourceLineNo">092</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_REGIONS_KEY =<a name="line.92"></a>
+<span class="sourceLineNo">093</span>      "zookeeper.znode.replication.regions";<a name="line.93"></a>
+<span class="sourceLineNo">094</span>  public static final String ZOOKEEPER_ZNODE_REPLICATION_REGIONS_DEFAULT = "regions";<a name="line.94"></a>
+<span class="sourceLineNo">095</span><a name="line.95"></a>
+<span class="sourceLineNo">096</span>  /**<a name="line.96"></a>
+<span class="sourceLineNo">097</span>   * The name of the znode that contains all replication queues<a name="line.97"></a>
+<span class="sourceLineNo">098</span>   */<a name="line.98"></a>
+<span class="sourceLineNo">099</span>  private final String queuesZNode;<a name="line.99"></a>
+<span class="sourceLineNo">100</span><a name="line.100"></a>
+<span class="sourceLineNo">101</span>  /**<a name="line.101"></a>
+<span class="sourceLineNo">102</span>   * The name of the znode that contains queues of hfile references to be replicated<a name="line.102"></a>
+<span class="sourceLineNo">103</span>   */<a name="line.103"></a>
+<span class="sourceLineNo">104</span>  private final String hfileRefsZNode;<a name="line.104"></a>
+<span class="sourceLineNo">105</span><a name="line.105"></a>
+<span class="sourceLineNo">106</span>  private final String regionsZNode;<a name="line.106"></a>
+<span class="sourceLineNo">107</span><a name="line.107"></a>
+<span class="sourceLineNo">108</span>  public ZKReplicationQueueStorage(ZKWatcher zookeeper, Configuration conf) {<a name="line.108"></a>
+<span class="sourceLineNo">109</span>    super(zookeeper, conf);<a name="line.109"></a>
+<span class="sourceLineNo">110</span><a name="line.110"></a>
+<span class="sourceLineNo">111</span>    String queuesZNodeName = conf.get("zookeeper.znode.replication.rs", "rs");<a name="line.111"></a>
+<span class="sourceLineNo">112</span>    String hfileRefsZNodeName = conf.get(ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_KEY,<a name="line.112"></a>
+<span class="sourceLineNo">113</span>      ZOOKEEPER_ZNODE_REPLICATION_HFILE_REFS_DEFAULT);<a name="line.113"></a>
+<span class="sourceLineNo">114</span>    this.queuesZNode = ZNodePaths.joinZNode(replicationZNode, queuesZNodeName);<a name="line.114"></a>
+<span class="sourceLineNo">115</span>    this.hfileRefsZNode = ZNodePaths.joinZNode(replicationZNode, hfileRefsZNodeName);<a name="line.115"></a>
+<span class="sourceLineNo">116</span>    this.regionsZNode = ZNodePaths.joinZNode(replicationZNode, conf<a name="line.116"></a>
+<span class="sourceLineNo">117</span>        .get(ZOOKEEPER_ZNODE_REPLICATION_REGIONS_KEY, ZOOKEEPER_ZNODE_REPLICATION_REGIONS_DEFAULT));<a name="line.117"></a>
+<span class="sourceLineNo">118</span>  }<a name="line.118"></a>
+<span class="sourceLineNo">119</span><a name="line.119"></a>
+<span class="sourceLineNo">120</span>  private String getRsNode(ServerName serverName) {<a name="line.120"></a>
+<span class="sourceLineNo">121</span>    return ZNodePaths.joinZNode(queuesZNode, serverName.getServerName());<a name="line.121"></a>
+<span class="sourceLineNo">122</span>  }<a name="line.122"></a>
+<span class="sourceLineNo">123</span><a name="line.123"></a>
+<span class="sourceLineNo">124</span>  private String getQueueNode(ServerName serverName, String queueId) {<a name="line.124"></a>
+<span class="sourceLineNo">125</span>    return ZNodePaths.joinZNode(getRsNode(serverName), queueId);<a name="line.125"></a>
+<span class="sourceLineNo">126</span>  }<a name="line.126"></a>
+<span class="sourceLineNo">127</span><a name="line.127"></a>
+<span class="sourceLineNo">128</span>  private String getFileNode(String queueNode, String fileName) {<a name="line.128"></a>
+<span class="sourceLineNo">129</span>    return ZNodePaths.joinZNode(queueNode, fileName);<a name="line.129"></a>
+<span class="sourceLineNo">130</span>  }<a name="line.130"></a>
+<span class="sourceLineNo">131</span><a name="line.131"></a>
+<span class="sourceLineNo">132</span>  private String getFileNode(ServerName serverName, String queueId, String fileName) {<a name="line.132"></a>
+<span class="sourceLineNo">133</span>    return getFileNode(getQueueNode(serverName, queueId), fileName);<a name="line.133"></a>
+<span class="sourceLineNo">134</span>  }<a name="line.134"></a>
+<span class="sourceLineNo">135</span><a name="line.135"></a>
+<span class="sourceLineNo">136</span>  /**<a name="line.136"></a>
+<span class="sourceLineNo">137</span>   * Put all regions under /hbase/replication/regions znode will lead to too many children because<a name="line.137"></a>
+<span class="sourceLineNo">138</span>   * of the huge number of regions in real production environment. So here we use hash of encoded<a name="line.138"></a>
+<span class="sourceLineNo">139</span>   * region name to distribute the znode into multiple znodes. &lt;br&gt;<a name="line.139"></a>
+<span class="sourceLineNo">140</span>   * So the final znode path will be format like this:<a name="line.140"></a>
+<span class="sourceLineNo">141</span>   *<a name="line.141"></a>
+<span class="sourceLineNo">142</span>   * &lt;pre&gt;<a name="line.142"></a>
+<span class="sourceLineNo">143</span>   * /hbase/replication/regions/e1/ff/dd04e76a6966d4ffa908ed0586764767-100<a name="line.143"></a>
+<span class="sourceLineNo">144</span>   * &lt;/pre&gt;<a name="line.144"></a>
+<span class="sourceLineNo">145</span>   *<a name="line.145"></a>
+<span class="sourceLineNo">146</span>   * The e1 indicate the first level hash of encoded region name, and the ff indicate the second<a name="line.146"></a>
+<span class="sourceLineNo">147</span>   * level hash of encoded region name, the 100 indicate the peer id. &lt;br&gt;<a name="line.147"></a>
+<span class="sourceLineNo">148</span>   * Note that here we use two-level hash because if only one-level hash (such as mod 65535), it<a name="line.148"></a>
+<span class="sourceLineNo">149</span>   * will still lead to too many children under the /hbase/replication/regions znode.<a name="line.149"></a>
+<span class="sourceLineNo">150</span>   * @param encodedRegionName the encoded region name.<a name="line.150"></a>
+<span class="sourceLineNo">151</span>   * @param peerId peer id for replication.<a name="line.151"></a>
+<span class="sourceLineNo">152</span>   * @return ZNode path to persist the max sequence id that we've pushed for the given region and<a name="line.152"></a>
+<span class="sourceLineNo">153</span>   *         peer.<a name="line.153"></a>
+<span class="sourceLineNo">154</span>   */<a name="line.154"></a>
+<span class="sourceLineNo">155</span>  @VisibleForTesting<a name="line.155"></a>
+<span class="sourceLineNo">156</span>  public String getSerialReplicationRegionPeerNode(String encodedRegionName, String peerId) {<a name="line.156"></a>
+<span class="sourceLineNo">157</span>    if (encodedRegionName == null || encodedRegionName.length() != RegionInfo.MD5_HEX_LENGTH) {<a name="line.157"></a>
+<span class="sourceLineNo">158</span>      throw new IllegalArgumentException(<a name="line.158"></a>
+<span class="sourceLineNo">159</span>          "Invalid encoded region name: " + encodedRegionName + ", length should be 32.");<a name="line.159"></a>
+<span class="sourceLineNo">160</span>    }<a name="line.160"></a>
+<span class="sourceLineNo">161</span>    return new StringBuilder(regionsZNode).append(ZNodePaths.ZNODE_PATH_SEPARATOR)<a name="line.161"></a>
+<span class="sourceLineNo">162</span>        .append(encodedRegionName.substring(0, 2)).append(ZNodePaths.ZNODE_PATH_SEPARATOR)<a name="line.162"></a>
+<span class="sourceLineNo">163</span>        .append(encodedRegionName.substring(2, 4)).append(ZNodePaths.ZNODE_PATH_SEPARATOR)<a name="line.163"></a>
+<span class="sourceLineNo">164</span>        .append(encodedRegionName).append("-").append(peerId).toString();<a name="line.164"></a>
 <span class="sourceLineNo">165</span>  }<a name="line.165"></a>
 <span class="sourceLineNo">166</span><a name="line.166"></a>
 <span class="sourceLineNo">167</span>  @Override<a name="line.167"></a>
-<span class="sourceLineNo">168</span>  public void addWAL(ServerName serverName, String queueId, String fileName)<a name="line.168"></a>
-<span class="sourceLineNo">169</span>      throws ReplicationException {<a name="line.169"></a>
-<span class="sourceLineNo">170</span>    try {<a name="line.170"></a>
-<span class="sourceLineNo">171</span>      ZKUtil.createWithParents(zookeeper, getFileNode(serverName, queueId, fileName));<a name="line.171"></a>
-<span class="sourceLineNo">172</span>    } catch (KeeperException e) {<a name="line.172"></a>
-<span class="sourceLineNo">173</span>      throw new ReplicationException("Failed to add wal to queue (serverName=" + serverName<a name="line.173"></a>
-<span class="sourceLineNo">174</span>          + ", queueId=" + queueId + ", fileName=" + fileName + ")", e);<a name="line.174"></a>
-<span class="sourceLineNo">175</span>    }<a name="line.175"></a>
-<span class="sourceLineNo">176</span>  }<a name="line.176"></a>
-<span class="sourceLineNo">177</span><a name="line.177"></a>
-<span class="sourceLineNo">178</span>  @Override<a name="line.178"></a>
-<span class="sourceLineNo">179</span>  public void removeWAL(ServerName serverName, String queueId, String fileName)<a name="line.179"></a>
-<span class="sourceLineNo">180</span>      throws ReplicationException {<a name="line.180"></a>
-<span class="sourceLineNo">181</span>    String fileNode = getFileNode(serverName, queueId, fileName);<a name="line.181"></a>
-<span class="sourceLineNo">182</span>    try {<a name="line.182"></a>
-<span class="sourceLineNo">183</span>      ZKUtil.deleteNode(zookeeper, fileNode);<a name="line.183"></a>
-<span class="sourceLineNo">184</span>    } catch (NoNodeException e) {<a name="line.184"></a>
-<span class="sourceLineNo">185</span>      LOG.warn("{} already deleted when removing log", fileNode);<a name="line.185"></a>
-<span class="sourceLineNo">186</span>    } catch (KeeperException e) {<a name="line.186"></a>
-<span class="sourceLineNo">187</span>      throw new ReplicationException("Failed to remove wal from queue (serverName=" + serverName +<a name="line.187"></a>
-<span class="sourceLineNo">188</span>        ", queueId=" + queueId + ", fileName=" + fileName + ")", e);<a name="line.188"></a>
-<span class="sourceLineNo">189</span>    }<a name="line.189"></a>
-<span class="sourceLineNo">190</span>  }<a name="line.190"></a>
-<span class="sourceLineNo">191</span><a name="line.191"></a>
-<span class="sourceLineNo">192</span>  @Override<a name="line.192"></a>
-<span class="sourceLineNo">193</span>  public void setWALPosition(ServerName serverName, String queueId, String fileName, long position,<a name="line.193"></a>
-<span class="sourceLineNo">194</span>      Map&lt;String, Long&gt; lastSeqIds) throws ReplicationException {<a name="line.194"></a>
-<span class="sourceLineNo">195</span>    try {<a name="line.195"></a>
-<span class="sourceLineNo">196</span>      List&lt;ZKUtilOp&gt; listOfOps = new ArrayList&lt;&gt;();<a name="line.196"></a>
-<span class="sourceLineNo">197</span>      if (position &gt; 0) {<a name="line.197"></a>
-<span class="sourceLineNo">198</span>        listOfOps.add(ZKUtilOp.setData(getFileNode(serverName, queueId, fileName),<a name="line.198"></a>
-<span class="sourceLineNo">199</span>          ZKUtil.positionToByteArray(position)));<a name="line.199"></a>
-<span class="sourceLineNo">200</span>      }<a name="line.200"></a>
-<span class="sourceLineNo">201</span>      // Persist the max sequence id(s) of regions for serial replication atomically.<a name="line.201"></a>
-<span class="sourceLineNo">202</span>      for (Entry&lt;String, Long&gt; lastSeqEntry : lastSeqIds.entrySet()) {<a name="line.202"></a>
-<span class="sourceLineNo">203</span>        String peerId = new ReplicationQueueInfo(queueId).getPeerId();<a name="line.203"></a>
-<span class="sourceLineNo">204</span>        String path = getSerialReplicationRegionPeerNode(lastSeqEntry.getKey(), peerId);<a name="line.204"></a>
-<span class="sourceLineNo">205</span>        /*<a name="line.205"></a>
-<span class="sourceLineNo">206</span>         * Make sure the existence of path<a name="line.206"></a>
-<span class="sourceLineNo">207</span>         * /hbase/replication/regions/&lt;hash&gt;/&lt;encoded-region-name&gt;-&lt;peer-id&gt;. As the javadoc in<a name="line.207"></a>
-<span class="sourceLineNo">208</span>         * multiOrSequential() method said, if received a NodeExistsException, all operations will<a name="line.208"></a>
-<span class="sourceLineNo">209</span>         * fail. So create the path here, and in fact, no need to add this operation to listOfOps,<a name="line.209"></a>
-<span class="sourceLineNo">210</span>         * because only need to make sure that update file position and sequence id atomically.<a name="line.210"></a>
-<span class="sourceLineNo">211</span>         */<a name="line.211"></a>
-<span class="sourceLineNo">212</span>        ZKUtil.createWithParents(zookeeper, path);<a name="line.212"></a>
-<span class="sourceLineNo">213</span>        // Persist the max sequence id of region to zookeeper.<a name="line.213"></a>
-<span class="sourceLineNo">214</span>        listOfOps.add(ZKUtilOp.setData(path, ZKUtil.positionToByteArray(lastSeqEntry.getValue())));<a name="line.214"></a>
-<span class="sourceLineNo">215</span>      }<a name="line.215"></a>
-<span class="sourceLineNo">216</span>      if (!listOfOps.isEmpty()) {<a name="line.216"></a>
-<span class="sourceLineNo">217</span>        ZKUtil.multiOrSequential(zookeeper, listOfOps, false);<a name="line.217"></a>
-<span class="sourceLineNo">218</span>      }<a name="line.218"></a>
-<span class="sourceLineNo">219</span>    } catch (KeeperException e) {<a name="line.219"></a>
-<span class="sourceLineNo">220</span>      throw new ReplicationException("Failed to set log position (serverName=" + serverName<a name="line.220"></a>
-<span class="sourceLineNo">221</span>          + ", queueId=" + queueId + ", fileName=" + fileName + ", position=" + position + ")", e);<a name="line.221"></a>
-<span class="sourceLineNo">222</span>    }<a name="line.222"></a>
-<span class="sourceLineNo">223</span>  }<a name="line.223"></a>
-<span class="sourceLineNo">224</span><a name="line.224"></a>
-<span class="sourceLineNo">225</span>  @Override<a name="line.225"></a>
-<span class="sourceLineNo">226</span>  public long getLastSequenceId(String encodedRegionName, String peerId)<a name="line.226"></a>
-<span class="sourceLineNo">227</span>      throws ReplicationException {<a name="line.227"></a>
-<span class="sourceLineNo">228</span>    byte[] data;<a name="line.228"></a>
-<span class="sourceLineNo">229</span>    try {<a name="line.229"></a>
-<span class="sourceLineNo">230</span>      data =<a name="line.230"></a>
-<span class="sourceLineNo">231</span>          ZKUtil.getData(zookeeper, getSerialReplicationRegionPeerNode(encodedRegionName, peerId));<a name="line.231"></a>
-<span class="sourceLineNo">232</span>    } catch (KeeperException | InterruptedException e) {<a name="line.232"></a>
-<span class="sourceLineNo">233</span>      throw new ReplicationException("Failed to get the last sequence id(region="<a name="line.233"></a>
-<span class="sourceLineNo">234</span>          + encodedRegionName + ", peerId=" + peerId + ")");<a name="line.234"></a>
-<span class="sourceLineNo">235</span>    }<a name="line.235"></a>
-<span class="sourceLineNo">236</span>    try {<a name="line.236"></a>
-<span class="sourceLineNo">237</span>      return ZKUtil.parseWALPositionFrom(data);<a name="line.237"></a>
-<span class="sourceLineNo">238</span>    } catch (DeserializationException de) {<a name="line.238"></a>
-<span class="sourceLineNo">239</span>      LOG.warn("Failed to parse log position (region=" + encodedRegionName + ", peerId=" + peerId<a name="line.239"></a>
-<span class="sourceLineNo">240</span>          + "), data=" + Bytes.toStringBinary(data));<a name="line.240"></a>
-<span class="sourceLineNo">241</span>    }<a name="line.241"></a>
-<span class="sourceLineNo">242</span>    return HConstants.NO_SEQNUM;<a name="line.242"></a>
-<span class="sourceLineNo">243</span>  }<a name="line.243"></a>
-<span class="sourceLineNo">244</span><a name="line.244"></a>
-<span class="sourceLineNo">245</span>  @Override<a name="line.245"></a>
-<span class="sourceLineNo">246</span>  public long getWALPosition(ServerName serverName, String queueId, String fileName)<a name="line.246"></a>
-<span class="sourceLineNo">247</span>      throws ReplicationException {<a name="line.247"></a>
-<span class="sourceLineNo">248</span>    byte[] bytes;<a name="line.248"></a>
-<span class="sourceLineNo">249</span>    try {<a name="line.249"></a>
-<span class="sourceLineNo">250</span>      bytes = ZKUtil.getData(zookeeper, getFileNode(serverName, queueId, fileName));<a name="line.250"></a>
-<span class="sourceLineNo">251</span>    } catch (KeeperException | InterruptedException e) {<a name="line.251"></a>
-<span class="sourceLineNo">252</span>      throw new ReplicationException("Failed to get log position (serverName=" + serverName +<a name="line.252"></a>
-<span class="sourceLineNo">253</span>        ", queueId=" + queueId + ", fileName=" + fileName + ")", e);<a name="line.253"></a>
-<span class="sourceLineNo">254</span>    }<a name="line.254"></a>
-<span class="sourceLineNo">255</span>    try {<a name="line.255"></a>
-<span class="sourceLineNo">256</span>      return ZKUtil.parseWALPositionFrom(bytes);<a name="line.256"></a>
-<span class="sourceLineNo">257</span>    } catch (DeserializationException de) {<a name="line.257"></a>
-<span class="sourceLineNo">258</span>      LOG.warn("Failed parse log position (serverName={}, queueId={}, fileName={})",<a name="line.258"></a>
-<span class="sourceLineNo">259</span>          serverName, queueId, fileName);<a name="line.259"></a>
-<span class="sourceLineNo">260</span>    }<a name="line.260"></a>
-<span class="sourceLineNo">261</span>    // if we can not parse the position, start at the beginning of the wal file again<a name="line.261"></a>
-<span class="sourceLineNo">262</span>    return 0;<a name="line.262"></a>
-<span class="sourceLineNo">263</span>  }<a name="line.263"></a>
-<span class="sourceLineNo">264</span><a name="line.264"></a>
-<span class="sourceLineNo">265</span>  @Override<a name="line.265"></a>
-<span class="sourceLineNo">266</span>  public Pair&lt;String, SortedSet&lt;String&gt;&gt; claimQueue(ServerName sourceServerName, String queueId,<a name="line.266"></a>
-<span class="sourceLineNo">267</span>      ServerName destServerName) throws ReplicationException {<a name="line.267"></a>
-<span class="sourceLineNo">268</span>    LOG.info("Atomically moving {}/{}'s WALs to {}", sourceServerName, queueId, destServerName);<a name="line.268"></a>
-<span class="sourceLineNo">269</span>    try {<a name="line.269"></a>
-<span class="sourceLineNo">270</span>      ZKUtil.createWithParents(zookeeper, getRsNode(destServerName));<a name="line.270"></a>
-<span class="sourceLineNo">271</span>    } catch (KeeperException e) {<a name="line.271"></a>
-<span class="sourceLineNo">272</span>      throw new ReplicationException(<a name="line.272"></a>
-<span class="sourceLineNo">273</span>          "Claim queue queueId=" + queueId + " from " + sourceServerName + " to " + destServerName +<a name="line.273"></a>
-<span class="sourceLineNo">274</span>            " failed when creating the node for " + destServerName,<a name="line.274"></a>
-<span class="sourceLineNo">275</span>          e);<a name="line.275"></a>
-<span class="sourceLineNo">276</span>    }<a name="line.276"></a>
-<span class="sourceLineNo">277</span>    try {<a name="line.277"></a>
-<span class="sourceLineNo">278</span>      String oldQueueNode = getQueueNode(sourceServerName, queueId);<a name="line.278"></a>
-<span class="sourceLineNo">279</span>      List&lt;String&gt; wals = ZKUtil.listChildrenNoWatch(zookeeper, oldQueueNode);<a name="line.279"></a>
-<span class="sourceLineNo">280</span>      String newQueueId = queueId + "-" + sourceServerName;<a name="line.280"></a>
-<span class="sourceLineNo">281</span>      if (CollectionUtils.isEmpty(wals)) {<a name="line.281"></a>
-<span class="sourceLineNo">282</span>        ZKUtil.deleteNodeFailSilent(zookeeper, oldQueueNode);<a name="line.282"></a>
-<span class="sourceLineNo">283</span>        LOG.info("Removed empty {}/{}", sourceServerName, queueId);<a name="line.283"></a>
-<span class="sourceLineNo">284</span>        return new Pair&lt;&gt;(newQueueId, Collections.emptySortedSet());<a name="line.284"></a>
-<span class="sourceLineNo">285</span>      }<a name="line.285"></a>
-<span class="sourceLineNo">286</span>      String newQueueNode = getQueueNode(destServerName, newQueueId);<a name="line.286"></a>
-<span class="sourceLineNo">287</span>      List&lt;ZKUtilOp&gt; listOfOps = new ArrayList&lt;&gt;();<a name="line.287"></a>
-<span class="sourceLineNo">288</span>      SortedSet&lt;String&gt; logQueue = new TreeSet&lt;&gt;();<a name="line.288"></a>
-<span class="sourceLineNo">289</span>      // create the new cluster znode<a name="line.289"></a>
-<span class="sourceLineNo">290</span>      listOfOps.add(ZKUtilOp.createAndFailSilent(newQueueNode, HConstants.EMPTY_BYTE_ARRAY));<a name="line.290"></a>
-<span class="sourceLineNo">291</span>      // get the offset of the logs and set it to new znodes<a name="line.291"></a>
-<span class="sourceLineNo">292</span>      for (String wal : wals) {<a name="line.292"></a>
-<span class="sourceLineNo">293</span>        String oldWalNode = getFileNode(oldQueueNode, wal);<a name="line.293"></a>
-<span class="sourceLineNo">294</span>        byte[] logOffset = ZKUtil.getData(this.zookeeper, oldWalNode);<a name="line.294"></a>
-<span class="sourceLineNo">295</span>        LOG.debug("Creating {} with data {}", wal, Bytes.toStringBinary(logOffset));<a name="line.295"></a>
-<span class="sourceLineNo">296</span>        String newWalNode = getFileNode(newQueueNode, wal);<a name="line.296"></a>
-<span class="sourceLineNo">297</span>        listOfOps.add(ZKUtilOp.createAndFailSilent(newWalNode, logOffset));<a name="line.297"></a>
-<span class="sourceLineNo">298</span>        listOfOps.add(ZKUtilOp.deleteNodeFailSilent(oldWalNode));<a name="line.298"></a>
-<span class="sourceLineNo">299</span>        logQueue.add(wal);<a name="line.299"></a>
-<span class="sourceLineNo">300</span>      }<a name="line.300"></a>
-<span class="sourceLineNo">301</span>      // add delete op for peer<a name="line.301"></a>
-<span class="sourceLineNo">302</span>      listOfOps.add(ZKUtilOp.deleteNodeFailSilent(oldQueueNode));<a name="line.302"></a>
-<span class="sourceLineNo">303</span><a name="line.303"></a>
-<span class="sourceLineNo">304</span>      LOG.trace("The multi list size is {}", listOfOps.size());<a name="line.304"></a>
-<span class="sourceLineNo">305</span>      ZKUtil.multiOrSequential(zookeeper, listOfOps, false);<a name="line.305"></a>
-<span class="sourceLineNo">306</span><a name="line.306"></a>
-<span class="sourceLineNo">307</span>      LOG.info("Atomically moved {}/{}'s WALs to {}", sourceServerName, queueId, destServerName);<a name="line.307"></a>
-<span class="sourceLineNo">308</span>      return new Pair&lt;&gt;(newQueueId, logQueue);<a name="line.308"></a>
-<span class="sourceLineNo">309</span>    } catch (NoNodeException | NodeExistsException | NotEmptyException | BadVersionException e) {<a name="line.309"></a>
-<span class="sourceLineNo">310</span>      // Multi call failed; it looks like some other regionserver took away the logs.<a name="line.310"></a>
-<span class="sourceLineNo">311</span>      // These exceptions mean that zk tells us the request can not be execute so it is safe to just<a name="line.311"></a>
-<span class="sourceLineNo">312</span>      // return a null. For other types of exception should be thrown out to notify the upper layer.<a name="line.312"></a>
-<span class="sourceLineNo">313</span>      LOG.info("Claim queue queueId={} from {} to {} failed with {}, someone else took the log?",<a name="line.313"></a>
-<span class="sourceLineNo">314</span>          queueId,sourceServerName, destServerName, e.toString());<a name="line.314"></a>
-<span class="sourceLineNo">315</span>      return null;<a name="line.315"></a>
-<span class="sourceLineNo">316</span>    } catch (KeeperException | InterruptedException e) {<a name="line.316"></a>
-<span class="sourceLineNo">317</span>      throw new ReplicationException("Claim queue queueId=" + queueId + " from " +<a name="line.317"></a>
-<span class="sourceLineNo">318</span>        sourceServerName + " to " + destServerName + " failed", e);<a name="line.318"></a>
-<span class="sourceLineNo">319</span>    }<a name="line.319"></a>
-<span class="sourceLineNo">320</span>  }<a name="line.320"></a>
-<span class="sourceLineNo">321</span><a name="line.321"></a>
-<span class="sourceLineNo">322</span>  @Override<a name="line.322"></a>
-<span class="sourceLineNo">323</span>  public void removeReplicatorIfQueueIsEmpty(ServerName serverName) throws ReplicationException {<a name="line.323"></a>
-<span class="sourceLineNo">324</span>    try {<a name="line.324"></a>
-<span class="sourceLineNo">325</span>      ZKUtil.deleteNodeFailSilent(zookeeper, getRsNode(serverName));<a name="line.325"></a>
-<span class="sourceLineNo">326</span>    } catch (NotEmptyException e) {<a name="line.326"></a>
-<span class="sourceLineNo">327</span>      // keep silence to avoid logging too much.<a name="line.327"></a>
-<span class="sourceLineNo">328</span>    } catch (KeeperException e) {<a name="line.328"></a>
-<span class="sourceLineNo">329</span>      throw new ReplicationException("Failed to remove replicator for " + serverName, e);<a name="line.329"></a>
-<span class="sourceLineNo">330</span>    }<a name="line.330"></a>
-<span class="sourceLineNo">331</span>  }<a name="line.331"></a>
-<span class="sourceLineNo">332</span><a name="line.332"></a>
-<span class="sourceLineNo">333</span>  private List&lt;ServerName&gt; getListOfReplicators0() throws KeeperException {<a name="line.333"></a>
-<span class="sourceLineNo">334</span>    return nullToEmpty(ZKUtil.listChildrenNoWatch(zookeeper, queuesZNode)).stream()<a name="line.334"></a>
-<span class="sourceLineNo">335</span>        .map(ServerName::parseServerName).collect(toList());<a name="line.335"></a>
-<span class="sourceLineNo">336</span>  }<a name="line.336"></a>
-<span class="sourceLineNo">337</span><a name="line.337"></a>
-<span class="sourceLineNo">338</span>  @Override<a name="line.338"></a>
-<span class="sourceLineNo">339</span>  public List&lt;ServerName&gt; getListOfReplicators() throws ReplicationException {<a name="line.339"></a>
-<span class="sourceLineNo">340</span>    try {<a name="line.340"></a>
-<span class="sourceLineNo">341</span>      return getListOfReplicators0();<a name="line.341"></a>
-<span class="sourceLineNo">342</span>    } catch (KeeperException e) {<a name="line.342"></a>
-<span class="sourceLineNo">343</span>      throw new ReplicationException("Failed to get list of replicators", e);<a name="line.343"></a>
-<span class="sourceLineNo">344</span>    }<a name="line.344"></a>
-<span class="sourceLineNo">345</span>  }<a name="line.345"></a>
-<span class="sourceLineNo">346</span><a name="line.346"></a>
-<span class="sourceLineNo">347</span>  private List&lt;String&gt; getWALsInQueue0(ServerName serverName, String queueId)<a name="line.347"></a>
-<span class="sourceLineNo">348</span>      throws KeeperException {<a name="line.348"></a>
-<span class="sourceLineNo">349</span>    return nullToEmpty(ZKUtil.listChildrenNoWatch(zookeeper, getQueueNode(serverName, queueId)));<a name="line.349"></a>
-<span class="sourceLineNo">350</span>  }<a name="line.350"></a>
-<span class="sourceLineNo">351</span><a name="line.351"></a>
-<span class="sourceLineNo">352</span>  @Override<a name="line.352"></a>
-<span class="sourceLineNo">353</span>  public List&lt;String&gt; getWALsInQueue(ServerName serverName, String queueId)<a name="line.353"></a>
-<span class="sourceLineNo">354</span>      throws ReplicationException {<a name="line.354"></a>
-<span class="sourceLineNo">355</span>    try {<a name="line.355"></a>
-<span class="sourceLineNo">356</span>      return getWALsInQueue0(serverName, queueId);<a name="line.356"></a>
-<span class="sourceLineNo">357</span>    } catch (KeeperException e) {<a name="line.357"></a>
-<span class="sourceLineNo">358</span>      throw new ReplicationException(<a name="line.358"></a>
-<span class="sourceLineNo">359</span>          "Failed to get wals in queue (serverName=" + serverName + ", queueId=" + queueId + ")",<a name="line.359"></a>
-<span class="sourceLineNo">360</span>          e);<a name="line.360"></a>
-<span class="sourceLineNo">361</span>    }<a name="line.361"></a>
-<span class="sourceLineNo">362</span>  }<a name="line.362"></a>
-<span class="sourceLineNo">363</span><a name="line.363"></a>
-<span class="sourceLineNo">364</span>  private List&lt;String&gt; getAllQueues0(ServerName serverName) throws KeeperException {<a name="line.364"></a>
-<span class="sourceLineNo">365</span>    return nullToEmpty(ZKUtil.listChildrenNoWatch(zookeeper, getRsNode(serverName)));<a name="line.365"></a>
-<span class="sourceLineNo">366</span>  }<a name="line.366"></a>
-<span class="sourceLineNo">367</span><a name="line.367"></a>
-<span class="sourceLineNo">368</span>  @Override<a name="line.368"></a>
-<span class="sourceLineNo">369</span>  public List&lt;String&gt; getAllQueues(ServerName serverName) throws ReplicationException {<a name="line.369"></a>
-<span class="sourceLineNo">370</span>    try {<a name="line.370"></a>
-<span class="sourceLineNo">371</span>      return getAllQueues0(serverName);<a name="line.371"></a>
-<span class="sourceLineNo">372</span>    } catch (KeeperException e) {<a name="line.372"></a>
-<span class="sourceLineNo">373</span>      throw new ReplicationException("Failed to get all queues (serverName=" + serverName + ")", e);<a name="line.373"></a>
-<span class="sourceLineNo">374</span>    }<a name="line.374"></a>
-<span class="sourceLineNo">375</span>  }<a name="line.375"></a>
-<span class="sourceLineNo">376</span><a name="line.376"></a>
-<span class="sourceLineNo">377</span>  // will be overridden in UTs<a name="line.377"></a>
-<span class="sourceLineNo">378</span>  @VisibleForTesting<a name="line.378"></a>
-<span class="sourceLineNo">379</span>  protected int getQueuesZNodeCversion() throws KeeperException {<a name="line.379"></a>
-<span class="sourceLineNo">380</span>    Stat stat = new Stat();<a name="line.380"></a>
-<span class="sourceLineNo">381</span>    ZKUtil.getDataNoWatch(this.zookeeper, this.queuesZNode, stat);<a name="line.381"></a>
-<span class="sourceLineNo">382</span>    return stat.getCversion();<a name="line.382"></a>
-<span class="sourceLineNo">383</span>  }<a name="line.383"></a>
-<span class="sourceLineNo">384</span><a name="line.384"></a>
-<span class="sourceLineNo">385</span>  @Override<a name="line.385"></a>
-<span class="sourceLineNo">386</span>  public Set&lt;String&gt; getAllWALs() throws ReplicationException {<a name="line.386"></a>
-<span class="sourceLineNo">387</span>    try {<a name="line.387"></a>
-<span class="sourceLineNo">388</span>      for (int retry = 0;; retry++) {<a name="line.388"></a>
-<span class="sourceLineNo">389</span>        int v0 = getQueuesZNodeCversion();<a name="line.389"></a>
-<span class="sourceLineNo">390</span>        List&lt;ServerName&gt; rss = getListOfReplicators0();<a name="line.390"></a>
-<span class="sourceLineNo">391</span>        if (rss.isEmpty()) {<a name="line.391"></a>
-<span class="sourceLineNo">392</span>          LOG.debug("Didn't find a RegionServer that replicates, won't prevent deletions.");<a name="line.392"></a>
-<span class="sourceLineNo">393</span>          return Collections.emptySet();<a name="line.393"></a>
-<span class="sourceLineNo">394</span>        }<a name="line.394"></a>
-<span class="sourceLineNo">395</span>        Set&lt;String&gt; wals = new HashSet&lt;&gt;();<a name="line.395"></a>
-<span class="sourceLineNo">396</span>        for (ServerName rs : rss) {<a name="line.396"></a>
-<span class="sourceLineNo">397</span>          for (String queueId : getAllQueues0(rs)) {<a name="line.397"></a>
-<span class="sourceLineNo">398</span>            wals.addAll(getWALsInQueue0(rs, queueId));<a name="line.398"></a>
-<span class="sourceLineNo">399</span>          }<a name="line.399"></a>
-<span class="sourceLineNo">400</span>        }<a name="line.400"></a>
-<span class="sourceLineNo">401</span>        int v1 = getQueuesZNodeCversion();<a name="line.401"></a>
-<span class="sourceLineNo">402</span>        if (v0 == v1) {<a name="line.402"></a>
-<span class="sourceLineNo">403</span>          return wals;<a name="line.403"></a>
+<span class="sourceLineNo">168</span>  public void removeQueue(ServerName serverName, String queueId) throws ReplicationException {<a name="line.168"></a>
+<span class="sourceLineNo">169</span>    try {<a name="line.169"></a>
+<span class="sourceLineNo">170</span>      ZKUtil.deleteNodeRecursively(zookeeper, getQueueNode(serverName, queueId));<a name="line.170"></a>
+<span class="sourceLineNo">171</span>    } catch (KeeperException e) {<a name="line.171"></a>
+<span class="sourceLineNo">172</span>      throw new ReplicationException(<a name="line.172"></a>
+<span class="sourceLineNo">173</span>          "Failed to delete queue (serverName=" + serverName + ", queueId=" + queueId + ")", e);<a name="line.173"></a>
+<span class="sourceLineNo">174</span>    }<a name="line.174"></a>
+<span class="sourceLineNo">175</span>  }<a name="line.175"></a>
+<span class="sourceLineNo">176</span><a name="line.176"></a>
+<span class="sourceLineNo">177</span>  @Override<a name="line.177"></a>
+<span class="sourceLineNo">178</span>  public void addWAL(ServerName serverName, String queueId, String fileName)<a name="line.178"></a>
+<span class="sourceLineNo">179</span>      throws ReplicationException {<a name="line.179"></a>
+<span class="sourceLineNo">180</span>    try {<a name="line.180"></a>
+<span class="sourceLineNo">181</span>      ZKUtil.createWithParents(zookeeper, getFileNode(serverName, queueId, fileName));<a name="line.181"></a>
+<span class="sourceLineNo">182</span>    } catch (KeeperException e) {<a name="line.182"></a>
+<span class="sourceLineNo">183</span>      throw new ReplicationException("Failed to add wal to queue (serverName=" + serverName<a name="line.183"></a>
+<span class="sourceLineNo">184</span>          + ", queueId=" + queueId + ", fileName=" + fileName + ")", e);<a name="line.184"></a>
+<span class="sourceLineNo">185</span>    }<a name="line.185"></a>
+<span class="sourceLineNo">186</span>  }<a name="line.186"></a>
+<span class="sourceLineNo">187</span><a name="line.187"></a>
+<span class="sourceLineNo">188</span>  @Override<a name="line.188"></a>
+<span class="sourceLineNo">189</span>  public void removeWAL(ServerName serverName, String queueId, String fileName)<a name="line.189"></a>
+<span class="sourceLineNo">190</span>      throws ReplicationException {<a name="line.190"></a>
+<span class="sourceLineNo">191</span>    String fileNode = getFileNode(serverName, queueId, fileName);<a name="line.191"></a>
+<span class="sourceLineNo">192</span>    try {<a name="line.192"></a>
+<span class="sourceLineNo">193</span>      ZKUtil.deleteNode(zookeeper, fileNode);<a name="line.193"></a>
+<span class="sourceLineNo">194</span>    } catch (NoNodeException e) {<a name="line.194"></a>
+<span class="sourceLineNo">195</span>      LOG.warn("{} already deleted when removing log", fileNode);<a name="line.195"></a>
+<span class="sourceLineNo">196</span>    } catch (KeeperException e) {<a name="line.196"></a>
+<span class="sourceLineNo">197</span>      throw new ReplicationException("Failed to remove wal from queue (serverName=" + serverName +<a name="line.197"></a>
+<span class="sourceLineNo">198</span>        ", queueId=" + queueId + ", fileName=" + fileName + ")", e);<a name="line.198"></a>
+<span class="sourceLineNo">199</span>    }<a name="line.199"></a>
+<span class="sourceLineNo">200</span>  }<a name="line.200"></a>
+<span class="sourceLineNo">201</span><a name="line.201"></a>
+<span class="sourceLineNo">202</span>  @Override<a name="line.202"></a>
+<span class="sourceLineNo">203</span>  public void setWALPosition(ServerName serverName, String queueId, String fileName, long position,<a name="line.203"></a>
+<span class="sourceLineNo">204</span>      Map&lt;String, Long&gt; lastSeqIds) throws ReplicationException {<a name="line.204"></a>
+<span class="sourceLineNo">205</span>    try {<a name="line.205"></a>
+<span class="sourceLineNo">206</span>      List&lt;ZKUtilOp&gt; listOfOps = new ArrayList&lt;&gt;();<a name="line.206"></a>
+<span class="sourceLineNo">207</span>      if (position &gt; 0) {<a name="line.207"></a>
+<span class="sourceLineNo">208</span>        listOfOps.add(ZKUtilOp.setData(getFileNode(serverName, queueId, fileName),<a name="line.208"></a>
+<span class="sourceLineNo">209</span>          ZKUtil.positionToByteArray(position)));<a name="line.209"></a>
+<span class="sourceLineNo">210</span>      }<a name="line.210"></a>
+<span class="sourceLineNo">211</span>      // Persist the max sequence id(s) of regions for serial replication atomically.<a name="line.211"></a>
+<span class="sourceLineNo">212</span>      for (Entry&lt;String, Long&gt; lastSeqEntry : lastSeqIds.entrySet()) {<a name="line.212"></a>
+<span class="sourceLineNo">213</span>        String peerId = new ReplicationQueueInfo(queueId).getPeerId();<a name="line.213"></a>
+<span class="sourceLineNo">214</span>        String path = getSerialReplicationRegionPeerNode(lastSeqEntry.getKey(), peerId);<a name="line.214"></a>
+<span class="sourceLineNo">215</span>        /*<a name="line.215"></a>
+<span class="sourceLineNo">216</span>         * Make sure the existence of path<a name="line.216"></a>
+<span class="sourceLineNo">217</span>         * /hbase/replication/regions/&lt;hash&gt;/&lt;encoded-region-name&gt;-&lt;peer-id&gt;. As the javadoc in<a name="line.217"></a>
+<span class="sourceLineNo">218</span>         * multiOrSequential() method said, if received a NodeExistsException, all operations will<a name="line.218"></a>
+<span class="sourceLineNo">219</span>         * fail. So create the path here, and in fact, no need to add this operation to listOfOps,<a name="line.219"></a>
+<span class="sourceLineNo">220</span>         * because only need to make sure that update file position and sequence id atomically.<a name="line.220"></a>
+<span class="sourceLineNo">221</span>         */<a name="line.221"></a>
+<span class="sourceLineNo">222</span>        ZKUtil.createWithParents(zookeeper, path);<a name="line.222"></a>
+<span class="sourceLineNo">223</span>        // Persist the max sequence id of region to zookeeper.<a name="line.223"></a>
+<span class="sourceLineNo">224</span>        listOfOps.add(ZKUtilOp.setData(path, ZKUtil.positionToByteArray(lastSeqEntry.getValue())));<a name="line.224"></a>
+<span class="sourceLineNo">225</span>      }<a name="line.225"></a>
+<span class="sourceLineNo">226</span>      if (!listOfOps.isEmpty()) {<a name="line.226"></a>
+<span class="sourceLineNo">227</span>        ZKUtil.multiOrSequential(zookeeper, listOfOps, false);<a name="line.227"></a>
+<span class="sourceLineNo">228</span>      }<a name="line.228"></a>
+<span class="sourceLineNo">229</span>    } catch (KeeperException e) {<a name="line.229"></a>
+<span class="sourceLineNo">230</span>      throw new ReplicationException("Failed to set log position (serverName=" + serverName<a name="line.230"></a>
+<span class="sourceLineNo">231</span>          + ", queueId=" + queueId + ", fileName=" + fileName + ", position=" + position + ")", e);<a name="line.231"></a>
+<span class="sourceLineNo">232</span>    }<a name="line.232"></a>
+<span class="sourceLineNo">233</span>  }<a name="line.233"></a>
+<span class="sourceLineNo">234</span><a name="line.234"></a>
+<span class="sourceLineNo">235</span>  @Override<a name="line.235"></a>
+<span class="sourceLineNo">236</span>  public long getLastSequenceId(String encodedRegionName, String peerId)<a name="line.236"></a>
+<span class="sourceLineNo">237</span>      throws ReplicationException {<a name="line.237"></a>
+<span class="sourceLineNo">238</span>    byte[] data;<a name="line.238"></a>
+<span class="sourceLineNo">239</span>    try {<a name="line.239"></a>
+<span class="sourceLineNo">240</span>      data =<a name="line.240"></a>
+<span class="sourceLineNo">241</span>          ZKUtil.getData(zookeeper, getSerialReplicationRegionPeerNode(encodedRegionName, peerId));<a name="line.241"></a>
+<span class="sourceLineNo">242</span>    } catch (KeeperException | InterruptedException e) {<a name="line.242"></a>
+<span class="sourceLineNo">243</span>      throw new ReplicationException("Failed to get the last sequence id(region="<a name="line.243"></a>
+<span class="sourceLineNo">244</span>          + encodedRegionName + ", peerId=" + peerId + ")");<a name="line.244"></a>
+<span class="sourceLineNo">245</span>    }<a name="line.245"></a>
+<span class="sourceLineNo">246</span>    try {<a name="line.246"></a>
+<span class="sourceLineNo">247</span>      return ZKUtil.parseWALPositionFrom(data);<a name="line.247"></a>
+<span class="sourceLineNo">248</span>    } catch (DeserializationException de) {<a name="line.248"></a>
+<span class="sourceLineNo">249</span>      LOG.warn("Failed to parse log position (region=" + encodedRegionName + ", peerId=" + peerId<a name="line.249"></a>
+<span class="sourceLineNo">250</span>          + "), data=" + Bytes.toStringBinary(data));<a name="line.250"></a>
+<span class="sourceLineNo">251</span>    }<a name="line.251"></a>
+<span class="sourceLineNo">252</span>    return HConstants.NO_SEQNUM;<a name="line.252"></a>
+<span class="sourceLineNo">253</span>  }<a name="line.253"></a>
+<span class="sourceLineNo">254</span><a name="line.254"></a>
+<span class="sourceLineNo">255</span>  @Override<a name="line.255"></a>
+<span class="sourceLineNo">256</span>  public long getWALPosition(ServerName serverName, String queueId, String fileName)<a name="line.256"></a>
+<span class="sourceLineNo">257</span>      throws ReplicationException {<a name="line.257"></a>
+<span class="sourceLineNo">258</span>    byte[] bytes;<a name="line.258"></a>
+<span class="sourceLineNo">259</span>    try {<a name="line.259"></a>
+<span class="sourceLineNo">260</span>      bytes = ZKUtil.getData(zookeeper, getFileNode(serverName, queueId, fileName));<a name="line.260"></a>
+<span class="sourceLineNo">261</span>    } catch (KeeperException | InterruptedException e) {<a name="line.261"></a>
+<span class="sourceLineNo">262</span>      throw new ReplicationException("Failed to get log position (serverName=" + serverName +<a name="line.262"></a>
+<span class="sourceLineNo">263</span>        ", queueId=" + queueId + ", fileName=" + fileName + ")", e);<a name="line.263"></a>
+<span class="sourceLineNo">264</span>    }<a name="line.264"></a>
+<span class="sourceLineNo">265</span>    try {<a name="line.265"></a>
+<span class="sourceLineNo">266</span>      return ZKUtil.parseWALPositionFrom(bytes);<a name="line.266"></a>
+<span class="sourceLineNo">267</span>    } catch (DeserializationException de) {<a name="line.267"></a>
+<span class="sourceLineNo">268</span>      LOG.warn("Failed parse log position (serverName={}, queueId={}, fileName={})",<a name="line.268"></a>
+<span class="sourceLineNo">269</span>          serverName, queueId, fileName);<a name="line.269"></a>
+<span class="sourceLineNo">270</span>    }<a name="line.270"></a>
+<span class="sourceLineNo">271</span>    // if we can not parse the position, start at the beginning of the wal file again<a name="line.271"></a>
+<span class="sourceLineNo">272</span>    return 0;<a name="line.272"></a>
+<span class="sourceLineNo">273</span>  }<a name="line.273"></a>
+<span class="sourceLineNo">274</span><a name="line.274"></a>
+<span class="sourceLineNo">275</span>  @Override<a name="line.275"></a>
+<span class="sourceLineNo">276</span>  public Pair&lt;String, SortedSet&lt;String&gt;&gt; claimQueue(ServerName sourceServerName, String queueId,<a name="line.276"></a>
+<span class="sourceLineNo">277</span>      ServerName destServerName) throws ReplicationException {<a name="line.277"></a>
+<span class="sourceLineNo">278</span>    LOG.info("Atomically moving {}/{}'s WALs to {}", sourceServerName, queueId, destServerName);<a name="line.278"></a>
+<span class="sourceLineNo">279</span>    try {<a name="line.279"></a>
+<span class="sourceLineNo">280</span>      ZKUtil.createWithParents(zookeeper, getRsNode(destServerName));<a name="line.280"></a>
+<span class="sourceLineNo">281</span>    } catch (KeeperException e) {<a name="line.281"></a>
+<span class="sourceLineNo">282</span>      throw new ReplicationException(<a name="line.282"></a>
+<span class="sourceLineNo">283</span>          "Claim queue queueId=" + queueId + " from " + sourceServerName + " to " + destServerName +<a name="line.283"></a>
+<span class="sourceLineNo">284</span>            " failed when creating the node for " + destServerName,<a name="line.284"></a>
+<span class="sourceLineNo">285</span>          e);<a name="line.285"></a>
+<span class="sourceLineNo">286</span>    }<a name="line.286"></a>
+<span class="sourceLineNo">287</span>    try {<a name="line.287"></a>
+<span class="sourceLineNo">288</span>      String oldQueueNode = getQueueNode(sourceServerName, queueId);<a name="line.288"></a>
+<span class="sourceLineNo">289</span>      List&lt;String&gt; wals = ZKUtil.listChildrenNoWatch(zookeeper, oldQueueNode);<a name="line.289"></a>
+<span class="sourceLineNo">290</span>      String newQueueId = queueId + "-" + sourceServerName;<a name="line.290"></a>
+<span class="sourceLineNo">291</span>      if (CollectionUtils.isEmpty(wals)) {<a name="line.291"></a>
+<span class="sourceLineNo">292</span>        ZKUtil.deleteNodeFailSilent(zookeeper, oldQueueNode);<a name="line.292"></a>
+<span class="sourceLineNo">293</span>        LOG.info("Removed empty {}/{}", sourceServerName, queueId);<a name="line.293"></a>
+<span class="sourceLineNo">294</span>        return new Pair&lt;&gt;(newQueueId, Collections.emptySortedSet());<a name="line.294"></a>
+<span class="sourceLineNo">295</span>      }<a name="line.295"></a>
+<span class="sourceLineNo">296</span>      String newQueueNode = getQueueNode(destServerName, newQueueId);<a name="line.296"></a>
+<span class="sourceLineNo">297</span>      List&lt;ZKUtilOp&gt; listOfOps = new ArrayList&lt;&gt;();<a name="line.297"></a>
+<span class="sourceLineNo">298</span>      SortedSet&lt;String&gt; logQueue = new TreeSet&lt;&gt;();<a name="line.298"></a>
+<span class="sourceLineNo">299</span>      // create the new cluster znode<a name="line.299"></a>
+<span class="sourceLineNo">300</span>      listOfOps.add(ZKUtilOp.createAndFailSilent(newQueueNode, HConstants.EMPTY_BYTE_ARRAY));<a name="line.300"></a>
+<span class="sourceLineNo">301</span>      // get the offset of the logs and set it to new znodes<a name="line.301"></a>
+<span class="sourceLineNo">302</span>      for (String wal : wals) {<a name="line.302"></a>
+<span class="sourceLineNo">303</span>        String oldWalNode = getFileNode(oldQueueNode, wal);<a name="line.303"></a>
+<span class="sourceLineNo">304</span>        byte[] logOffset = ZKUtil.getData(this.zookeeper, oldWalNode);<a name="line.304"></a>
+<span class="sourceLineNo">305</span>        LOG.debug("Creating {} with data {}", wal, Bytes.toStringBinary(logOffset));<a name="line.305"></a>
+<span class="sourceLineNo">306</span>        String newWalNode = getFileNode(newQueueNode, wal);<a name="line.306"></a>
+<span class="sourceLineNo">307</span>        listOfOps.add(ZKUtilOp.createAndFailSilent(newWalNode, logOffset));<a name="line.307"></a>
+<span class="sourceLineNo">308</span>        listOfOps.add(ZKUtilOp.deleteNodeFailSilent(oldWalNode));<a name="line.308"></a>
+<span class="sourceLineNo">309</span>        logQueue.add(wal);<a name="line.309"></a>
+<span class="sourceLineNo">310</span>      }<a name="line.310"></a>
+<span class="sourceLineNo">311</span>      // add delete op for peer<a name="line.311"></a>
+<span class="sourceLineNo">312</span>      listOfOps.add(ZKUtilOp.deleteNodeFailSilent(oldQueueNode));<a name="line.312"></a>
+<span class="sourceLineNo">313</span><a name="line.313"></a>
+<span class="sourceLineNo">314</span>      LOG.trace("The multi list size is {}", listOfOps.size());<a name="line.314"></a>
+<span class="sourceLineNo">315</span>      ZKUtil.multiOrSequential(zookeeper, listOfOps, false);<a name="line.315"></a>
+<span class="sourceLineNo">316</span><a name="line.316"></a>
+<span class="sourceLineNo">317</span>      LOG.info("Atomically moved {}/{}'s WALs to {}", sourceServerName, queueId, destServerName);<a name="line.317"></a>
+<span class="sourceLineNo">318</span>      return new Pair&lt;&gt;(newQueueId, logQueue);<a name="line.318"></a>
+<span class="sourceLineNo">319</span>    } catch (NoNodeException | NodeExistsException | NotEmptyException | BadVersionException e) {<a name="line.319"></a>
+<span class="sourceLineNo">320</span>      // Multi call failed; it looks like some other regionserver took away the logs.<a name="line.320"></a>
+<span class="sourceLineNo">321</span>      // These exceptions mean that zk tells us the request can not be execute so it is safe to just<a name="line.321"></a>
+<span class="sourceLineNo">322</span>      // return a null. For other types of exception should be thrown out to notify the upper layer.<a name="line.322"></a>
+<span class="sourceLineNo">323</span>      LOG.info("Claim queue queueId={} from {} to {} failed with {}, someone else took the log?",<a name="line.323"></a>
+<span class="sourceLineNo">324</span>          queueId,sourceServerName, destServerName, e.toString());<a name="line.324"></a>
+<span class="sourceLineNo">325</span>      return null;<a name="line.325"></a>
+<span class="sourceLineNo">326</span>    } catch (KeeperException | InterruptedException e) {<a name="line.326"></a>
+<span class="sourceLineNo">327</span>      throw new ReplicationException("Claim queue queueId=" + queueId + " from " +<a name="line.327"></a>
+<span class="sourceLineNo">328</span>        sourceServerName + " to " + destServerName + " failed", e);<a name="line.328"></a>
+<span class="sourceLineNo">329</span>    }<a name="line.329"></a>
+<span class="sourceLineNo">330</span>  }<a name="line.330"></a>
+<span class="sourceLineNo">331</span><a name="line.331"></a>
+<span class="sourceLineNo">332</span>  @Override<a name="line.332"></a>
+<span class="sourceLineNo">333</span>  public void removeReplicatorIfQueueIsEmpty(ServerName serverName) throws ReplicationException {<a name="line.333"></a>
+<span class="sourceLineNo">334</span>    try {<a name="line.334"></a>
+<span class="sourceLineNo">335</span>      ZKUtil.deleteNodeFailSilent(zookeeper, getRsNode(serverName));<a name="line.335"></a>
+<span class="sourceLineNo">336</span>    } catch (NotEmptyException e) {<a name="line.336"></a>
+<span class="sourceLineNo">337</span>      // keep silence to avoid logging too much.<a name="line.337"></a>
+<span class="sourceLineNo">338</span>    } catch (KeeperException e) {<a name="line.338"></a>
+<span class="sourceLineNo">339</span>      throw new ReplicationException("Failed to remove replicator for " + serverName, e);<a name="line.339"></a>
+<span class="sourceLineNo">340</span>    }<a name="line.340"></a>
+<span class="sourceLineNo">341</span>  }<a name="line.341"></a>
+<span class="sourceLineNo">342</span><a name="line.342"></a>
+<span class="sourceLineNo">343</span>  private List&lt;ServerName&gt; getListOfReplicators0() throws KeeperException {<a name="line.343"></a>
+<span class="sourceLineNo">344</span>    return nullToEmpty(ZKUtil.listChildrenNoWatch(zookeeper, queuesZNode)).stream()<a name="line.344"></a>
+<span class="sourceLineNo">345</span>        .map(ServerName::parseServerName).collect(toList());<a name="line.345"></a>
+<span class="sourceLineNo">346</span>  }<a name="line.346"></a>
+<span class="sourceLineNo">347</span><a name="line.347"></a>
+<span class="sourceLineNo">348</span>  @Override<a name="line.348"></a>
+<span class="sourceLineNo">349</span>  public List&lt;ServerName&gt; getListOfReplicators() throws ReplicationException {<a name="line.349"></a>
+<span class="sourceLineNo">350</span>    try {<a name="line.350"></a>
+<span class="sourceLineNo">351</span>      return getListOfReplicators0();<a name="line.351"></a>
+<span class="sourceLineNo">352</span>    } catch (KeeperException e) {<a name="line.352"></a>
+<span class="sourceLineNo">353</span>      throw new ReplicationException("Failed to get list of replicators", e);<a name="line.353"></a>
+<span class="sourceLineNo">354</span>    }<a name="line.354"></a>
+<span class="sourceLineNo">355</span>  }<a name="line.355"></a>
+<span class="sourceLineNo">356</span><a name="line.356"></a>
+<span class="sourceLineNo">357</span>  private List&lt;String&gt; getWALsInQueue0(ServerName serverName, String queueId)<a name="line.357"></a>
+<span class="sourceLineNo">358</span>      throws KeeperException {<a name="line.358"></a>
+<span class="sourceLineNo">359</span>    return nullToEmpty(ZKUtil.listChildrenNoWatch(zookeeper, getQueueNode(serverName, queueId)));<a name="line.359"></a>
+<span class="sourceLineNo">360</span>  }<a name="line.360"></a>
+<span class="sourceLineNo">361</span><a name="line.361"></a>
+<span class="sourceLineNo">362</span>  @Override<a name="line.362"></a>
+<span class="sourceLineNo">363</span>  public List&lt;String&gt; getWALsInQueue(ServerName serverName, String queueId)<a name="line.363"></a>
+<span class="sourceLineNo">364</span>      throws ReplicationException {<a name="line.364"></a>
+<span class="sourceLineNo">365</span>    try {<a name="line.365"></a>
+<span class="sourceLineNo">366</span>      return getWALsInQueue0(serverName, queueId);<a name="line.366"></a>
+<span class="sourceLineNo">367</span>    } catch (KeeperException e) {<a name="line.367"></a>
+<span class="sourceLineNo">368</span>      throw new ReplicationException(<a name="line.368"></a>
+<span class="sourceLineNo">369</span>          "Failed to get wals in queue (serverName=" + serverName + ", queueId=" + queueId + ")",<a name="line.369"></a>
+<span class="sourceLineNo">370</span>          e);<a name="line.370"></a>
+<span class="sourceLineNo">371</span>    }<a name="line.371"></a>
+<span class="sourceLineNo">372</span>  }<a name="line.372"></a>
+<span class="sourceLineNo">373</span><a name="line.373"></a>
+<span class="sourceLineNo">374</span>  private List&lt;String&gt; getAllQueues0(ServerName serverName) throws KeeperException {<a name="line.374"></a>
+<span class="sourceLineNo">375</span>    return nullToEmpty(ZKUtil.listChildrenNoWatch(zookeeper, getRsNode(serverName)));<a name="line.375"></a>
+<span class="sourceLineNo">376</span>  }<a name="line.376"></a>
+<span class="sourceLineNo">377</span><a name="line.377"></a>
+<span class="sourceLineNo">378</span>  @Override<a name="line.378"></a>
+<span class="sourceLineNo">379</span>  public List&lt;String&gt; getAllQueues(ServerName serverName) throws ReplicationException {<a name="line.379"></a>
+<span class="sourceLineNo">380</span>    try {<a name="line.380"></a>
+<span class="sourceLineNo">381</span>      return getAllQueues0(serverName);<a name="line.381"></a>
+<span class="sourceLineNo">382</span>    } catch (KeeperException e) {<a name="line.382"></a>
+<span class="sourceLineNo">383</span>      throw new ReplicationException("Failed to get all queues (serverName=" + serverName + ")", e);<a name="line.383"></a>
+<span class="sourceLineNo">384</span>    }<a name="line.384"></a>
+<span class="sourceLineNo">385</span>  }<a name="line.385"></a>
+<span class="sourceLineNo">386</span><a name="line.386"></a>
+<span class="sourceLineNo">387</span>  // will be overridden in UTs<a name="line.387"></a>
+<span class="sourceLineNo">388</span>  @VisibleForTesting<a name="line.388"></a>
+<span class="sourceLineNo">389</span>  protected int getQueuesZNodeCversion() throws KeeperException {<a name="line.389"></a>
+<span class="sourceLineNo">390</span>    Stat stat = new Stat();<a name="line.390"></a>
+<span class="sourceLineNo">391</span>    ZKUtil.getDataNoWatch(this.zookeeper, this.queuesZNode, stat);<a name="line.391"></a>
+<span class="sourceLineNo">392</span>    return stat.getCversion();<a name="line.392"></a>
+<span class="sourceLineNo">393</span>  }<a name="line.393"></a>
+<span class="sourceLineNo">394</span><a name="line.394"></a>
+<span class="sourceLineNo">395</span>  @Override<a name="line.395"></a>
+<span class="sourceLineNo">396</span>  public Set&lt;String&gt; getAllWALs() throws ReplicationException {<a name="line.396"></a>
+<span class="sourceLineNo">397</span>    try {<a name="line.397"></a>
+<span class="sourceLineNo">398</span>      for (int retry = 0;; retry++) {<a name="line.398"></a>
+<span class="sourceLineNo">399</span>        int v0 = getQueuesZNodeCversion();<a name="line.399"></a>
+<span class="sourceLineNo">400</span>        List&lt;ServerName&gt; rss = getListOfReplicators0();<a name="line.400"></a>
+<span class="sourceLineNo">401</span>        if (rss.isEmpty()) {<a name="line.401"></a>
+<span class="sourceLineNo">402</span>          LOG.debug("Didn't find a RegionServer that replicates, won't prevent deletions.");<a name="line.402"></a>
+<span class="sourceLineNo">403</span>          return Collections.emptySet();<a name="line.403"></a>
 <span class="sourceLineNo">404</span>        }<a name="line.404"></a>
-<span class="sourceLineNo">405</span>        LOG.info("Replication queue node cversion changed from %d to %d, retry = %d",<a name="line.405"></a>
-<span class="sourceLineNo">406</span>            v0, v1, retry);<a name="line.406"></a>
-<span class="sourceLineNo">407</span>      }<a name="line.407"></a>
-<span class="sourceLineNo">408</span>    } catch (KeeperException e) {<a name="line.408"></a>
-<span class="sourceLineNo">409</span>      throw new ReplicationException("Failed to get all wals", e);<a name="line.409"></a>
-<span class="sourceLineNo">410</span>    }<a name="line.410"></a>
-<span class="sourceLineNo">411</span>  }<a name="line.411"></a>
-<span class="sourceLineNo">412</span><a name="line.412"></a>
-<span class="sourceLineNo">413</span>  private String getHFileRefsPeerNode(String peerId) {<a name="line.413"></a>
-<span class="sourceLineNo">414</span>    return ZNodePaths.joinZNode(hfileRefsZNode, peerId);<a name="line.414"></a>
-<span class="sourceLineNo">415</span>  }<a name="line.415"></a>
-<span class="sourceLineNo">416</span><a name="line.416"></a>
-<span class="sourceLineNo">417</span>  private String getHFileNode(String peerNode, String fileName) {<a name="line.417"></a>
-<span class="sourceLineNo">418</span>    return ZNodePaths.joinZNode(peerNode, fileName);<a name="line.418"></a>
-<span class="sourceLineNo">419</span>  }<a name="line.419"></a>
-<span class="sourceLineNo">420</span><a name="line.420"></a>
-<span class="sourceLineNo">421</span>  @Override<a name="line.421"></a>
-<span class="sourceLineNo">422</span>  public void addPeerToHFileRefs(String peerId) throws ReplicationException {<a name="line.422"></a>
-<span class="sourceLineNo">423</span>    String peerNode = getHFileRefsPeerNode(peerId);<a name="line.423"></a>
-<span class="sourceLineNo">424</span>    try {<a name="line.424"></a>
-<span class="sourceLineNo">425</span>      if (ZKUtil.checkExists(zookeeper, peerNode) == -1) {<a name="line.425"></a>
-<span class="sourceLineNo">426</span>        LOG.info("Adding peer {} to hfile reference queue.", peerId);<a name="line.426"></a>
-<span class="sourceLineNo">427</span>        ZKUtil.createWithParents(zookeeper, peerNode);<a name="line.427"></a>
-<span class="sourceLineNo">428</span>      }<a name="line.428"></a>
-<span class="sourceLineNo">429</span>    } catch (KeeperException e) {<a name="line.429"></a>
-<span class="sourceLineNo">430</span>      throw new ReplicationException("Failed to add peer " + peerId + " to hfile reference queue.",<a name="line.430"></a>
-<span class="sourceLineNo">431</span>          e);<a name="line.431"></a>
-<span class="sourceLineNo">432</span>    }<a name="line.432"></a>
-<span class="sourceLineNo">433</span>  }<a name="line.433"></a>
-<span class="sourceLineNo">434</span><a name="line.434"></a>
-<span class="sourceLineNo">435</span>  @Override<a name="line.435"></a>
-<span class="sourceLineNo">436</span>  public void removePeerFromHFileRefs(String peerId) throws ReplicationException {<a name="line.436"></a>
-<span class="sourceLineNo">437</span>    String peerNode = getHFileRefsPeerNode(peerId);<a name="line.437"></a>
-<span class="sourceLineNo">438</span>    try {<a name="line.438"></a>
-<span class="sourceLineNo">439</span>      if (ZKUtil.checkExists(zookeeper, peerNode) == -1) {<a name="line.439"></a>
-<span class="sourceLineNo">440</span>          LOG.debug("Peer {} not found in hfile reference queue.", peerNode);<a name="line.440"></a>
-<span class="sourceLineNo">441</span>      } else {<a name="line.441"></a>
-<span class="sourceLineNo">442</span>        LOG.info("Removing peer {} from hfile reference queue.", peerNode);<a name="line.442"></a>
-<span class="sourceLineNo">443</span>        ZKUtil.deleteNodeRecursively(zookeeper, peerNode);<a name="line.443"></a>
-<span class="sourceLineNo">444</span>      }<a name="line.444"></a>
-<span class="sourceLineNo">445</span>    } catch (KeeperException e) {<a name="line.445"></a>
-<span class="sourceLineNo">446</span>      throw new ReplicationException(<a name="line.446"></a>
-<span class="sourceLineNo">447</span>          "Failed to remove peer " + peerId + " from hfile reference queue.", e);<a name="line.447"></a>
-<span class="sourceLineNo">448</span>    }<a name="line.448"></a>
-<span class="sourceLineNo">449</span>  }<a name="line.449"></a>
-<span class="sourceLineNo">450</span><a name="line.450"></a>
-<span class="sourceLineNo">451</span>  @Override<a name="line.451"></a>
-<span class="sourceLineNo">452</span>  public void addHFileRefs(String peerId, List&lt;Pair&lt;Path, Path&gt;&gt; pairs)<a name="line.452"></a>
-<span class="sourceLineNo">453</span>      throws ReplicationException {<a name="line.453"></a>
-<span class="sourceLineNo">454</span>    String peerNode = getHFileRefsPeerNode(peerId);<a name="line.454"></a>
-<span class="sourceLineNo">455</span>    LOG.debug("Adding hfile references {} in queue {}", pairs, peerNode);<a name="line.455"></a>
-<span class="sourceLineNo">456</span>    List&lt;ZKUtilOp&gt; listOfOps = pairs.stream().map(p -&gt; p.getSecond().getName())<a name="line.456"></a>
-<span class="sourceLineNo">457</span>        .map(n -&gt; getHFileNode(peerNode, n))<a name="line.457"></a>
-<span class="sourceLineNo">458</span>        .map(f -&gt; ZKUtilOp.createAndFailSilent(f, HConstants.EMPTY_BYTE_ARRAY)).collect(toList());<a name="line.458"></a>
-<span class="sourceLineNo">459</span>      LOG.debug("The multi list size for adding hfile references in zk for node {} is {}",<a name="line.459"></a>
-<span class="sourceLineNo">460</span>          peerNode, listOfOps.size());<a name="line.460"></a>
-<span class="sourceLineNo">461</span>    try {<a name="line.461"></a>
-<span class="sourceLineNo">462</span>      ZKUtil.multiOrSequential(this.zookeeper, listOfOps, true);<a name="line.462"></a>
-<span class="sourceLineNo">463</span>    } catch (KeeperException e) {<a name="line.463"></a>
-<span class="sourceLineNo">464</span>      throw new ReplicationException("Failed to add hfile reference to peer " + peerId, e);<a name="line.464"></a>
-<span class="sourceLineNo">465</span>    }<a name="line.465"></a>
-<span class="sourceLineNo">466</span>  }<a name="line.466"></a>
-<span class="sourceLineNo">467</span><a name="line.467"></a>
-<span class="sourceLineNo">468</span>  @Override<a name="line.468"></a>
-<span class="sourceLineNo">469</span>  public void removeHFileRefs(String peerId, List&lt;String&gt; files) throws ReplicationException {<a name="line.469"></a>
-<span class="sourceLineNo">470</span>    String peerNode = getHFileRefsPeerNode(peerId);<a name="line.470"></a>
-<span class="sourceLineNo">471</span>    LOG.debug("Removing hfile references {} from queue {}", files, peerNode);<a name="line.471"></a>
-<span class="sourceLineNo">472</span><a name="line.472"></a>
-<span class="sourceLineNo">473</span>    List&lt;ZKUtilOp&gt; listOfOps = files.stream().map(n -&gt; getHFileNode(peerNode, n))<a name="line.473"></a>
-<span class="sourceLineNo">474</span>        .map(ZKUtilOp::deleteNodeFailSilent).collect(toList());<a name="line.474"></a>
-<span class="sourceLineNo">475</span>    LOG.debug("The multi list size for removing hfile references in zk for node {} is {}",<a name="line.475"></a>
-<span class="sourceLineNo">476</span>        peerNode, listOfOps.size());<a name="line.476"></a>
-<span class="sourceLineNo">477</span>    try {<a name="line.477"></a>
-<span class="sourceLineNo">478</span>      ZKUtil.multiOrSequential(this.zookeeper, listOfOps, true);<a name="line.478"></a>
-<span class="sourceLineNo">479</span>    } catch (KeeperException e) {<a name="line.479"></a>
-<span class="sourceLineNo">480</span>      throw new ReplicationException("Failed to remove hfile reference from peer " + peerId, e);<a name="line.480"></a>
-<span class="sourceLineNo">481</span>    }<a name="line.481"></a>
-<span class="sourceLineNo">482</span>  }<a name="line.482"></a>
-<span class="sourceLineNo">483</span><a name="line.483"></a>
-<span class="sourceLineNo">484</span>  private List&lt;String&gt; getAllPeersFromHFileRefsQueue0() throws KeeperException {<a name="line.484"></a>
-<span class="sourceLineNo">485</span>    return nullToEmpty(ZKUtil.listChildrenNoWatch(zookeeper, hfileRefsZNode));<a name="line.485"></a>
-<span class="sourceLineNo">486</span>  }<a name="line.486"></a>
-<span class="sourceLineNo">487</span><a name="line.487"></a>
-<span class="sourceLineNo">488</span>  @Override<a name="line.488"></a>
-<span class="sourceLineNo">489</span>  public List&lt;String&gt; getAllPeersFromHFileRefsQueue() throws ReplicationException {<a name="line.489"></a>
-<span class="sourceLineNo">490</span>    try {<a name="line.490"></a>
-<span class="sourceLineNo">491</span>      return getAllPeersFromHFileRefsQueue0();<a name="line.491"></a>
-<span class="sourceLineNo">492</span>    } catch (KeeperException e) {<a name="line.492"></a>
-<span class="sourceLineNo">493</span>      throw new ReplicationException("Failed to get list of all peers in hfile references node.",<a name="line.493"></a>
-<span class="sourceLineNo">494</span>          e);<a name="line.494"></a>
-<span class="sourceLineNo">495</span>    }<a name="line.495"></a>
+<span class="sourceLineNo">405</span>        Set&lt;String&gt; wals = new HashSet&lt;&gt;();<a name="line.405"></a>
+<span class="sourceLineNo">406</span>        for (ServerName rs : rss) {<a name="line.406"></a>
+<span class="sourceLineNo">407</span>          for (String queueId : getAllQueues0(rs)) {<a name="line.407"></a>
+<span class="sourceLineNo">408</span>            wals.addAll(getWALsInQueue0(rs, queueId));<a name="line.408"></a>
+<span class="sourceLineNo">409</span>          }<a name="line.409"></a>
+<span class="sourceLineNo">410</span>        }<a name="line.410"></a>
+<span class="sourceLineNo">411</span>        int v1 = getQueuesZNodeCversion();<a name="line.411"></a>
+<span class="sourceLineNo">412</span>        if (v0 == v1) {<a name="line.412"></a>
+<span class="sourceLineNo">413</span>          return wals;<a name="line.413"></a>
+<span class="sourceLineNo">414</span>        }<a name="line.414"></a>
+<span class="sourceLineNo">415</span>        LOG.info("Replication queue node cversion changed from %d to %d, retry = %d",<a name="line.415"></a>
+<span class="sourceLineNo">416</span>            v0, v1, retry);<a name="line.416"></a>
+<span class="sourceLineNo">417</span>      }<a name="line.417"></a>
+<span class="sourceLineNo">418</span>    } catch (KeeperException e) {<a name="line.418"></a>
+<span class="sourceLineNo">419</span>      throw new ReplicationException("Failed to get all wals", e);<a name="line.419"></a>
+<span class="sourceLineNo">420</span>    }<a name="line.420"></a>
+<span class="sourceLineNo">421</span>  }<a name="line.421"></a>
+<span class="sourceLineNo">422</span><a name="line.422"></a>
+<span class="sourceLineNo">423</span>  private String getHFileRefsPeerNode(String peerId) {<a name="line.423"></a>
+<span class="sourceLineNo">424</span>    return ZNodePaths.joinZNode(hfileRefsZNode, peerId);<a name="line.424"></a>
+<span class="sourceLineNo">425</span>  }<a name="line.425"></a>
+<span class="sourceLineNo">426</span><a name="line.426"></a>
+<span class="sourceLineNo">427</span>  private String getHFileNode(String peerNode, String fileName) {<a name="line.427"></a>
+<span class="sourceLineNo">428</span>    return ZNodePaths.joinZNode(peerNode, fileName);<a name="line.428"></a>
+<span class="sourceLineNo">429</span>  }<a name="line.429"></a>
+<span class="sourceLineNo">430</s

<TRUNCATED>