http://git-wip-us.apache.org/repos/asf/hbase-site/blob/6b5143ed/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.html ---------------------------------------------------------------------- diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.html b/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.html index 2ef9b72..be1a1b0 100644 --- a/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.html +++ b/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSource.html @@ -27,552 +27,548 @@ <span class="sourceLineNo">019</span>package org.apache.hadoop.hbase.replication.regionserver;<a name="line.19"></a> <span class="sourceLineNo">020</span><a name="line.20"></a> <span class="sourceLineNo">021</span>import org.apache.hadoop.hbase.shaded.com.google.common.collect.Lists;<a name="line.21"></a> -<span class="sourceLineNo">022</span>import org.apache.hadoop.hbase.shaded.com.google.common.util.concurrent.ListenableFuture;<a name="line.22"></a> -<span class="sourceLineNo">023</span>import org.apache.hadoop.hbase.shaded.com.google.common.util.concurrent.Service;<a name="line.23"></a> -<span class="sourceLineNo">024</span><a name="line.24"></a> -<span class="sourceLineNo">025</span>import java.io.IOException;<a name="line.25"></a> -<span class="sourceLineNo">026</span>import java.util.ArrayList;<a name="line.26"></a> -<span class="sourceLineNo">027</span>import java.util.Collection;<a name="line.27"></a> -<span class="sourceLineNo">028</span>import java.util.Comparator;<a name="line.28"></a> -<span class="sourceLineNo">029</span>import java.util.HashMap;<a name="line.29"></a> -<span class="sourceLineNo">030</span>import java.util.List;<a name="line.30"></a> -<span class="sourceLineNo">031</span>import java.util.Map;<a name="line.31"></a> -<span class="sourceLineNo">032</span>import java.util.UUID;<a name="line.32"></a> -<span class="sourceLineNo">033</span>import java.util.concurrent.ConcurrentHashMap;<a name="line.33"></a> -<span class="sourceLineNo">034</span>import java.util.concurrent.PriorityBlockingQueue;<a name="line.34"></a> -<span class="sourceLineNo">035</span>import java.util.concurrent.TimeUnit;<a name="line.35"></a> -<span class="sourceLineNo">036</span>import java.util.concurrent.TimeoutException;<a name="line.36"></a> -<span class="sourceLineNo">037</span>import java.util.concurrent.atomic.AtomicLong;<a name="line.37"></a> -<span class="sourceLineNo">038</span><a name="line.38"></a> -<span class="sourceLineNo">039</span>import org.apache.commons.lang.StringUtils;<a name="line.39"></a> -<span class="sourceLineNo">040</span>import org.apache.commons.logging.Log;<a name="line.40"></a> -<span class="sourceLineNo">041</span>import org.apache.commons.logging.LogFactory;<a name="line.41"></a> -<span class="sourceLineNo">042</span>import org.apache.hadoop.conf.Configuration;<a name="line.42"></a> -<span class="sourceLineNo">043</span>import org.apache.hadoop.fs.FileSystem;<a name="line.43"></a> -<span class="sourceLineNo">044</span>import org.apache.hadoop.fs.Path;<a name="line.44"></a> -<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.HBaseConfiguration;<a name="line.45"></a> -<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.HConstants;<a name="line.46"></a> -<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.Stoppable;<a name="line.47"></a> -<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.TableName;<a name="line.48"></a> -<span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.classification.InterfaceAudience;<a name="line.49"></a> -<span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.regionserver.RSRpcServices;<a name="line.50"></a> -<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.replication.ChainWALEntryFilter;<a name="line.51"></a> -<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.replication.ClusterMarkingEntryFilter;<a name="line.52"></a> -<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.replication.ReplicationEndpoint;<a name="line.53"></a> -<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.replication.ReplicationException;<a name="line.54"></a> -<span class="sourceLineNo">055</span>import org.apache.hadoop.hbase.replication.ReplicationPeer;<a name="line.55"></a> -<span class="sourceLineNo">056</span>import org.apache.hadoop.hbase.replication.ReplicationPeers;<a name="line.56"></a> -<span class="sourceLineNo">057</span>import org.apache.hadoop.hbase.replication.ReplicationQueueInfo;<a name="line.57"></a> -<span class="sourceLineNo">058</span>import org.apache.hadoop.hbase.replication.ReplicationQueues;<a name="line.58"></a> -<span class="sourceLineNo">059</span>import org.apache.hadoop.hbase.replication.SystemTableWALEntryFilter;<a name="line.59"></a> -<span class="sourceLineNo">060</span>import org.apache.hadoop.hbase.replication.WALEntryFilter;<a name="line.60"></a> -<span class="sourceLineNo">061</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.61"></a> -<span class="sourceLineNo">062</span>import org.apache.hadoop.hbase.util.Pair;<a name="line.62"></a> -<span class="sourceLineNo">063</span>import org.apache.hadoop.hbase.util.Threads;<a name="line.63"></a> -<span class="sourceLineNo">064</span>import org.apache.hadoop.hbase.wal.AbstractFSWALProvider;<a name="line.64"></a> -<span class="sourceLineNo">065</span>import org.apache.hadoop.hbase.wal.WAL.Entry;<a name="line.65"></a> -<span class="sourceLineNo">066</span><a name="line.66"></a> -<span class="sourceLineNo">067</span><a name="line.67"></a> -<span class="sourceLineNo">068</span>/**<a name="line.68"></a> -<span class="sourceLineNo">069</span> * Class that handles the source of a replication stream.<a name="line.69"></a> -<span class="sourceLineNo">070</span> * Currently does not handle more than 1 slave<a name="line.70"></a> -<span class="sourceLineNo">071</span> * For each slave cluster it selects a random number of peers<a name="line.71"></a> -<span class="sourceLineNo">072</span> * using a replication ratio. For example, if replication ration = 0.1<a name="line.72"></a> -<span class="sourceLineNo">073</span> * and slave cluster has 100 region servers, 10 will be selected.<a name="line.73"></a> -<span class="sourceLineNo">074</span> * <p><a name="line.74"></a> -<span class="sourceLineNo">075</span> * A stream is considered down when we cannot contact a region server on the<a name="line.75"></a> -<span class="sourceLineNo">076</span> * peer cluster for more than 55 seconds by default.<a name="line.76"></a> -<span class="sourceLineNo">077</span> * </p><a name="line.77"></a> -<span class="sourceLineNo">078</span> *<a name="line.78"></a> -<span class="sourceLineNo">079</span> */<a name="line.79"></a> -<span class="sourceLineNo">080</span>@InterfaceAudience.Private<a name="line.80"></a> -<span class="sourceLineNo">081</span>public class ReplicationSource extends Thread implements ReplicationSourceInterface {<a name="line.81"></a> -<span class="sourceLineNo">082</span><a name="line.82"></a> -<span class="sourceLineNo">083</span> private static final Log LOG = LogFactory.getLog(ReplicationSource.class);<a name="line.83"></a> -<span class="sourceLineNo">084</span> // Queues of logs to process, entry in format of walGroupId->queue,<a name="line.84"></a> -<span class="sourceLineNo">085</span> // each presents a queue for one wal group<a name="line.85"></a> -<span class="sourceLineNo">086</span> private Map<String, PriorityBlockingQueue<Path>> queues = new HashMap<>();<a name="line.86"></a> -<span class="sourceLineNo">087</span> // per group queue size, keep no more than this number of logs in each wal group<a name="line.87"></a> -<span class="sourceLineNo">088</span> protected int queueSizePerGroup;<a name="line.88"></a> -<span class="sourceLineNo">089</span> protected ReplicationQueues replicationQueues;<a name="line.89"></a> -<span class="sourceLineNo">090</span> private ReplicationPeers replicationPeers;<a name="line.90"></a> -<span class="sourceLineNo">091</span><a name="line.91"></a> -<span class="sourceLineNo">092</span> protected Configuration conf;<a name="line.92"></a> -<span class="sourceLineNo">093</span> protected ReplicationQueueInfo replicationQueueInfo;<a name="line.93"></a> -<span class="sourceLineNo">094</span> // id of the peer cluster this source replicates to<a name="line.94"></a> -<span class="sourceLineNo">095</span> private String peerId;<a name="line.95"></a> -<span class="sourceLineNo">096</span><a name="line.96"></a> -<span class="sourceLineNo">097</span> // The manager of all sources to which we ping back our progress<a name="line.97"></a> -<span class="sourceLineNo">098</span> protected ReplicationSourceManager manager;<a name="line.98"></a> -<span class="sourceLineNo">099</span> // Should we stop everything?<a name="line.99"></a> -<span class="sourceLineNo">100</span> protected Stoppable stopper;<a name="line.100"></a> -<span class="sourceLineNo">101</span> // How long should we sleep for each retry<a name="line.101"></a> -<span class="sourceLineNo">102</span> private long sleepForRetries;<a name="line.102"></a> -<span class="sourceLineNo">103</span> protected FileSystem fs;<a name="line.103"></a> -<span class="sourceLineNo">104</span> // id of this cluster<a name="line.104"></a> -<span class="sourceLineNo">105</span> private UUID clusterId;<a name="line.105"></a> -<span class="sourceLineNo">106</span> // id of the other cluster<a name="line.106"></a> -<span class="sourceLineNo">107</span> private UUID peerClusterId;<a name="line.107"></a> -<span class="sourceLineNo">108</span> // total number of edits we replicated<a name="line.108"></a> -<span class="sourceLineNo">109</span> private AtomicLong totalReplicatedEdits = new AtomicLong(0);<a name="line.109"></a> -<span class="sourceLineNo">110</span> // The znode we currently play with<a name="line.110"></a> -<span class="sourceLineNo">111</span> protected String peerClusterZnode;<a name="line.111"></a> -<span class="sourceLineNo">112</span> // Maximum number of retries before taking bold actions<a name="line.112"></a> -<span class="sourceLineNo">113</span> private int maxRetriesMultiplier;<a name="line.113"></a> -<span class="sourceLineNo">114</span> // Indicates if this particular source is running<a name="line.114"></a> -<span class="sourceLineNo">115</span> private volatile boolean sourceRunning = false;<a name="line.115"></a> -<span class="sourceLineNo">116</span> // Metrics for this source<a name="line.116"></a> -<span class="sourceLineNo">117</span> private MetricsSource metrics;<a name="line.117"></a> -<span class="sourceLineNo">118</span> //WARN threshold for the number of queued logs, defaults to 2<a name="line.118"></a> -<span class="sourceLineNo">119</span> private int logQueueWarnThreshold;<a name="line.119"></a> -<span class="sourceLineNo">120</span> // ReplicationEndpoint which will handle the actual replication<a name="line.120"></a> -<span class="sourceLineNo">121</span> private ReplicationEndpoint replicationEndpoint;<a name="line.121"></a> -<span class="sourceLineNo">122</span> // A filter (or a chain of filters) for the WAL entries.<a name="line.122"></a> -<span class="sourceLineNo">123</span> protected WALEntryFilter walEntryFilter;<a name="line.123"></a> -<span class="sourceLineNo">124</span> // throttler<a name="line.124"></a> -<span class="sourceLineNo">125</span> private ReplicationThrottler throttler;<a name="line.125"></a> -<span class="sourceLineNo">126</span> private long defaultBandwidth;<a name="line.126"></a> -<span class="sourceLineNo">127</span> private long currentBandwidth;<a name="line.127"></a> -<span class="sourceLineNo">128</span> protected final ConcurrentHashMap<String, ReplicationSourceShipper> workerThreads =<a name="line.128"></a> -<span class="sourceLineNo">129</span> new ConcurrentHashMap<>();<a name="line.129"></a> +<span class="sourceLineNo">022</span><a name="line.22"></a> +<span class="sourceLineNo">023</span>import java.io.IOException;<a name="line.23"></a> +<span class="sourceLineNo">024</span>import java.util.ArrayList;<a name="line.24"></a> +<span class="sourceLineNo">025</span>import java.util.Collection;<a name="line.25"></a> +<span class="sourceLineNo">026</span>import java.util.Comparator;<a name="line.26"></a> +<span class="sourceLineNo">027</span>import java.util.HashMap;<a name="line.27"></a> +<span class="sourceLineNo">028</span>import java.util.List;<a name="line.28"></a> +<span class="sourceLineNo">029</span>import java.util.Map;<a name="line.29"></a> +<span class="sourceLineNo">030</span>import java.util.UUID;<a name="line.30"></a> +<span class="sourceLineNo">031</span>import java.util.concurrent.ConcurrentHashMap;<a name="line.31"></a> +<span class="sourceLineNo">032</span>import java.util.concurrent.PriorityBlockingQueue;<a name="line.32"></a> +<span class="sourceLineNo">033</span>import java.util.concurrent.TimeUnit;<a name="line.33"></a> +<span class="sourceLineNo">034</span>import java.util.concurrent.TimeoutException;<a name="line.34"></a> +<span class="sourceLineNo">035</span>import java.util.concurrent.atomic.AtomicLong;<a name="line.35"></a> +<span class="sourceLineNo">036</span><a name="line.36"></a> +<span class="sourceLineNo">037</span>import org.apache.commons.lang.StringUtils;<a name="line.37"></a> +<span class="sourceLineNo">038</span>import org.apache.commons.logging.Log;<a name="line.38"></a> +<span class="sourceLineNo">039</span>import org.apache.commons.logging.LogFactory;<a name="line.39"></a> +<span class="sourceLineNo">040</span>import org.apache.hadoop.conf.Configuration;<a name="line.40"></a> +<span class="sourceLineNo">041</span>import org.apache.hadoop.fs.FileSystem;<a name="line.41"></a> +<span class="sourceLineNo">042</span>import org.apache.hadoop.fs.Path;<a name="line.42"></a> +<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.HBaseConfiguration;<a name="line.43"></a> +<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.HConstants;<a name="line.44"></a> +<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.Stoppable;<a name="line.45"></a> +<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.TableName;<a name="line.46"></a> +<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.classification.InterfaceAudience;<a name="line.47"></a> +<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.regionserver.RSRpcServices;<a name="line.48"></a> +<span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.replication.ChainWALEntryFilter;<a name="line.49"></a> +<span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.replication.ClusterMarkingEntryFilter;<a name="line.50"></a> +<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.replication.ReplicationEndpoint;<a name="line.51"></a> +<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.replication.ReplicationException;<a name="line.52"></a> +<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.replication.ReplicationPeer;<a name="line.53"></a> +<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.replication.ReplicationPeers;<a name="line.54"></a> +<span class="sourceLineNo">055</span>import org.apache.hadoop.hbase.replication.ReplicationQueueInfo;<a name="line.55"></a> +<span class="sourceLineNo">056</span>import org.apache.hadoop.hbase.replication.ReplicationQueues;<a name="line.56"></a> +<span class="sourceLineNo">057</span>import org.apache.hadoop.hbase.replication.SystemTableWALEntryFilter;<a name="line.57"></a> +<span class="sourceLineNo">058</span>import org.apache.hadoop.hbase.replication.WALEntryFilter;<a name="line.58"></a> +<span class="sourceLineNo">059</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.59"></a> +<span class="sourceLineNo">060</span>import org.apache.hadoop.hbase.util.Pair;<a name="line.60"></a> +<span class="sourceLineNo">061</span>import org.apache.hadoop.hbase.util.Threads;<a name="line.61"></a> +<span class="sourceLineNo">062</span>import org.apache.hadoop.hbase.wal.AbstractFSWALProvider;<a name="line.62"></a> +<span class="sourceLineNo">063</span>import org.apache.hadoop.hbase.wal.WAL.Entry;<a name="line.63"></a> +<span class="sourceLineNo">064</span><a name="line.64"></a> +<span class="sourceLineNo">065</span><a name="line.65"></a> +<span class="sourceLineNo">066</span>/**<a name="line.66"></a> +<span class="sourceLineNo">067</span> * Class that handles the source of a replication stream.<a name="line.67"></a> +<span class="sourceLineNo">068</span> * Currently does not handle more than 1 slave<a name="line.68"></a> +<span class="sourceLineNo">069</span> * For each slave cluster it selects a random number of peers<a name="line.69"></a> +<span class="sourceLineNo">070</span> * using a replication ratio. For example, if replication ration = 0.1<a name="line.70"></a> +<span class="sourceLineNo">071</span> * and slave cluster has 100 region servers, 10 will be selected.<a name="line.71"></a> +<span class="sourceLineNo">072</span> * <p><a name="line.72"></a> +<span class="sourceLineNo">073</span> * A stream is considered down when we cannot contact a region server on the<a name="line.73"></a> +<span class="sourceLineNo">074</span> * peer cluster for more than 55 seconds by default.<a name="line.74"></a> +<span class="sourceLineNo">075</span> * </p><a name="line.75"></a> +<span class="sourceLineNo">076</span> *<a name="line.76"></a> +<span class="sourceLineNo">077</span> */<a name="line.77"></a> +<span class="sourceLineNo">078</span>@InterfaceAudience.Private<a name="line.78"></a> +<span class="sourceLineNo">079</span>public class ReplicationSource extends Thread implements ReplicationSourceInterface {<a name="line.79"></a> +<span class="sourceLineNo">080</span><a name="line.80"></a> +<span class="sourceLineNo">081</span> private static final Log LOG = LogFactory.getLog(ReplicationSource.class);<a name="line.81"></a> +<span class="sourceLineNo">082</span> // Queues of logs to process, entry in format of walGroupId->queue,<a name="line.82"></a> +<span class="sourceLineNo">083</span> // each presents a queue for one wal group<a name="line.83"></a> +<span class="sourceLineNo">084</span> private Map<String, PriorityBlockingQueue<Path>> queues = new HashMap<>();<a name="line.84"></a> +<span class="sourceLineNo">085</span> // per group queue size, keep no more than this number of logs in each wal group<a name="line.85"></a> +<span class="sourceLineNo">086</span> protected int queueSizePerGroup;<a name="line.86"></a> +<span class="sourceLineNo">087</span> protected ReplicationQueues replicationQueues;<a name="line.87"></a> +<span class="sourceLineNo">088</span> private ReplicationPeers replicationPeers;<a name="line.88"></a> +<span class="sourceLineNo">089</span><a name="line.89"></a> +<span class="sourceLineNo">090</span> protected Configuration conf;<a name="line.90"></a> +<span class="sourceLineNo">091</span> protected ReplicationQueueInfo replicationQueueInfo;<a name="line.91"></a> +<span class="sourceLineNo">092</span> // id of the peer cluster this source replicates to<a name="line.92"></a> +<span class="sourceLineNo">093</span> private String peerId;<a name="line.93"></a> +<span class="sourceLineNo">094</span><a name="line.94"></a> +<span class="sourceLineNo">095</span> // The manager of all sources to which we ping back our progress<a name="line.95"></a> +<span class="sourceLineNo">096</span> protected ReplicationSourceManager manager;<a name="line.96"></a> +<span class="sourceLineNo">097</span> // Should we stop everything?<a name="line.97"></a> +<span class="sourceLineNo">098</span> protected Stoppable stopper;<a name="line.98"></a> +<span class="sourceLineNo">099</span> // How long should we sleep for each retry<a name="line.99"></a> +<span class="sourceLineNo">100</span> private long sleepForRetries;<a name="line.100"></a> +<span class="sourceLineNo">101</span> protected FileSystem fs;<a name="line.101"></a> +<span class="sourceLineNo">102</span> // id of this cluster<a name="line.102"></a> +<span class="sourceLineNo">103</span> private UUID clusterId;<a name="line.103"></a> +<span class="sourceLineNo">104</span> // id of the other cluster<a name="line.104"></a> +<span class="sourceLineNo">105</span> private UUID peerClusterId;<a name="line.105"></a> +<span class="sourceLineNo">106</span> // total number of edits we replicated<a name="line.106"></a> +<span class="sourceLineNo">107</span> private AtomicLong totalReplicatedEdits = new AtomicLong(0);<a name="line.107"></a> +<span class="sourceLineNo">108</span> // The znode we currently play with<a name="line.108"></a> +<span class="sourceLineNo">109</span> protected String peerClusterZnode;<a name="line.109"></a> +<span class="sourceLineNo">110</span> // Maximum number of retries before taking bold actions<a name="line.110"></a> +<span class="sourceLineNo">111</span> private int maxRetriesMultiplier;<a name="line.111"></a> +<span class="sourceLineNo">112</span> // Indicates if this particular source is running<a name="line.112"></a> +<span class="sourceLineNo">113</span> private volatile boolean sourceRunning = false;<a name="line.113"></a> +<span class="sourceLineNo">114</span> // Metrics for this source<a name="line.114"></a> +<span class="sourceLineNo">115</span> private MetricsSource metrics;<a name="line.115"></a> +<span class="sourceLineNo">116</span> //WARN threshold for the number of queued logs, defaults to 2<a name="line.116"></a> +<span class="sourceLineNo">117</span> private int logQueueWarnThreshold;<a name="line.117"></a> +<span class="sourceLineNo">118</span> // ReplicationEndpoint which will handle the actual replication<a name="line.118"></a> +<span class="sourceLineNo">119</span> private ReplicationEndpoint replicationEndpoint;<a name="line.119"></a> +<span class="sourceLineNo">120</span> // A filter (or a chain of filters) for the WAL entries.<a name="line.120"></a> +<span class="sourceLineNo">121</span> protected WALEntryFilter walEntryFilter;<a name="line.121"></a> +<span class="sourceLineNo">122</span> // throttler<a name="line.122"></a> +<span class="sourceLineNo">123</span> private ReplicationThrottler throttler;<a name="line.123"></a> +<span class="sourceLineNo">124</span> private long defaultBandwidth;<a name="line.124"></a> +<span class="sourceLineNo">125</span> private long currentBandwidth;<a name="line.125"></a> +<span class="sourceLineNo">126</span> protected final ConcurrentHashMap<String, ReplicationSourceShipper> workerThreads =<a name="line.126"></a> +<span class="sourceLineNo">127</span> new ConcurrentHashMap<>();<a name="line.127"></a> +<span class="sourceLineNo">128</span><a name="line.128"></a> +<span class="sourceLineNo">129</span> private AtomicLong totalBufferUsed;<a name="line.129"></a> <span class="sourceLineNo">130</span><a name="line.130"></a> -<span class="sourceLineNo">131</span> private AtomicLong totalBufferUsed;<a name="line.131"></a> -<span class="sourceLineNo">132</span><a name="line.132"></a> -<span class="sourceLineNo">133</span> /**<a name="line.133"></a> -<span class="sourceLineNo">134</span> * Instantiation method used by region servers<a name="line.134"></a> -<span class="sourceLineNo">135</span> *<a name="line.135"></a> -<span class="sourceLineNo">136</span> * @param conf configuration to use<a name="line.136"></a> -<span class="sourceLineNo">137</span> * @param fs file system to use<a name="line.137"></a> -<span class="sourceLineNo">138</span> * @param manager replication manager to ping to<a name="line.138"></a> -<span class="sourceLineNo">139</span> * @param stopper the atomic boolean to use to stop the regionserver<a name="line.139"></a> -<span class="sourceLineNo">140</span> * @param peerClusterZnode the name of our znode<a name="line.140"></a> -<span class="sourceLineNo">141</span> * @param clusterId unique UUID for the cluster<a name="line.141"></a> -<span class="sourceLineNo">142</span> * @param replicationEndpoint the replication endpoint implementation<a name="line.142"></a> -<span class="sourceLineNo">143</span> * @param metrics metrics for replication source<a name="line.143"></a> -<span class="sourceLineNo">144</span> * @throws IOException<a name="line.144"></a> -<span class="sourceLineNo">145</span> */<a name="line.145"></a> -<span class="sourceLineNo">146</span> @Override<a name="line.146"></a> -<span class="sourceLineNo">147</span> public void init(final Configuration conf, final FileSystem fs,<a name="line.147"></a> -<span class="sourceLineNo">148</span> final ReplicationSourceManager manager, final ReplicationQueues replicationQueues,<a name="line.148"></a> -<span class="sourceLineNo">149</span> final ReplicationPeers replicationPeers, final Stoppable stopper,<a name="line.149"></a> -<span class="sourceLineNo">150</span> final String peerClusterZnode, final UUID clusterId, ReplicationEndpoint replicationEndpoint,<a name="line.150"></a> -<span class="sourceLineNo">151</span> final MetricsSource metrics)<a name="line.151"></a> -<span class="sourceLineNo">152</span> throws IOException {<a name="line.152"></a> -<span class="sourceLineNo">153</span> this.stopper = stopper;<a name="line.153"></a> -<span class="sourceLineNo">154</span> this.conf = HBaseConfiguration.create(conf);<a name="line.154"></a> -<span class="sourceLineNo">155</span> decorateConf();<a name="line.155"></a> -<span class="sourceLineNo">156</span> this.sleepForRetries =<a name="line.156"></a> -<span class="sourceLineNo">157</span> this.conf.getLong("replication.source.sleepforretries", 1000); // 1 second<a name="line.157"></a> -<span class="sourceLineNo">158</span> this.maxRetriesMultiplier =<a name="line.158"></a> -<span class="sourceLineNo">159</span> this.conf.getInt("replication.source.maxretriesmultiplier", 300); // 5 minutes @ 1 sec per<a name="line.159"></a> -<span class="sourceLineNo">160</span> this.queueSizePerGroup = this.conf.getInt("hbase.regionserver.maxlogs", 32);<a name="line.160"></a> -<span class="sourceLineNo">161</span> this.replicationQueues = replicationQueues;<a name="line.161"></a> -<span class="sourceLineNo">162</span> this.replicationPeers = replicationPeers;<a name="line.162"></a> -<span class="sourceLineNo">163</span> this.manager = manager;<a name="line.163"></a> -<span class="sourceLineNo">164</span> this.fs = fs;<a name="line.164"></a> -<span class="sourceLineNo">165</span> this.metrics = metrics;<a name="line.165"></a> -<span class="sourceLineNo">166</span> this.clusterId = clusterId;<a name="line.166"></a> -<span class="sourceLineNo">167</span><a name="line.167"></a> -<span class="sourceLineNo">168</span> this.peerClusterZnode = peerClusterZnode;<a name="line.168"></a> -<span class="sourceLineNo">169</span> this.replicationQueueInfo = new ReplicationQueueInfo(peerClusterZnode);<a name="line.169"></a> -<span class="sourceLineNo">170</span> // ReplicationQueueInfo parses the peerId out of the znode for us<a name="line.170"></a> -<span class="sourceLineNo">171</span> this.peerId = this.replicationQueueInfo.getPeerId();<a name="line.171"></a> -<span class="sourceLineNo">172</span> this.logQueueWarnThreshold = this.conf.getInt("replication.source.log.queue.warn", 2);<a name="line.172"></a> -<span class="sourceLineNo">173</span> this.replicationEndpoint = replicationEndpoint;<a name="line.173"></a> -<span class="sourceLineNo">174</span><a name="line.174"></a> -<span class="sourceLineNo">175</span> defaultBandwidth = this.conf.getLong("replication.source.per.peer.node.bandwidth", 0);<a name="line.175"></a> -<span class="sourceLineNo">176</span> currentBandwidth = getCurrentBandwidth();<a name="line.176"></a> -<span class="sourceLineNo">177</span> this.throttler = new ReplicationThrottler((double) currentBandwidth / 10.0);<a name="line.177"></a> -<span class="sourceLineNo">178</span> this.totalBufferUsed = manager.getTotalBufferUsed();<a name="line.178"></a> -<span class="sourceLineNo">179</span> LOG.info("peerClusterZnode=" + peerClusterZnode + ", ReplicationSource : " + peerId<a name="line.179"></a> -<span class="sourceLineNo">180</span> + ", currentBandwidth=" + this.currentBandwidth);<a name="line.180"></a> -<span class="sourceLineNo">181</span> }<a name="line.181"></a> -<span class="sourceLineNo">182</span><a name="line.182"></a> -<span class="sourceLineNo">183</span> private void decorateConf() {<a name="line.183"></a> -<span class="sourceLineNo">184</span> String replicationCodec = this.conf.get(HConstants.REPLICATION_CODEC_CONF_KEY);<a name="line.184"></a> -<span class="sourceLineNo">185</span> if (StringUtils.isNotEmpty(replicationCodec)) {<a name="line.185"></a> -<span class="sourceLineNo">186</span> this.conf.set(HConstants.RPC_CODEC_CONF_KEY, replicationCodec);<a name="line.186"></a> -<span class="sourceLineNo">187</span> }<a name="line.187"></a> -<span class="sourceLineNo">188</span> }<a name="line.188"></a> -<span class="sourceLineNo">189</span><a name="line.189"></a> -<span class="sourceLineNo">190</span> @Override<a name="line.190"></a> -<span class="sourceLineNo">191</span> public void enqueueLog(Path log) {<a name="line.191"></a> -<span class="sourceLineNo">192</span> String logPrefix = AbstractFSWALProvider.getWALPrefixFromWALName(log.getName());<a name="line.192"></a> -<span class="sourceLineNo">193</span> PriorityBlockingQueue<Path> queue = queues.get(logPrefix);<a name="line.193"></a> -<span class="sourceLineNo">194</span> if (queue == null) {<a name="line.194"></a> -<span class="sourceLineNo">195</span> queue = new PriorityBlockingQueue<>(queueSizePerGroup, new LogsComparator());<a name="line.195"></a> -<span class="sourceLineNo">196</span> queues.put(logPrefix, queue);<a name="line.196"></a> -<span class="sourceLineNo">197</span> if (this.sourceRunning) {<a name="line.197"></a> -<span class="sourceLineNo">198</span> // new wal group observed after source startup, start a new worker thread to track it<a name="line.198"></a> -<span class="sourceLineNo">199</span> // notice: it's possible that log enqueued when this.running is set but worker thread<a name="line.199"></a> -<span class="sourceLineNo">200</span> // still not launched, so it's necessary to check workerThreads before start the worker<a name="line.200"></a> -<span class="sourceLineNo">201</span> tryStartNewShipper(logPrefix, queue);<a name="line.201"></a> -<span class="sourceLineNo">202</span> }<a name="line.202"></a> -<span class="sourceLineNo">203</span> }<a name="line.203"></a> -<span class="sourceLineNo">204</span> queue.put(log);<a name="line.204"></a> -<span class="sourceLineNo">205</span> this.metrics.incrSizeOfLogQueue();<a name="line.205"></a> -<span class="sourceLineNo">206</span> // This will log a warning for each new log that gets created above the warn threshold<a name="line.206"></a> -<span class="sourceLineNo">207</span> int queueSize = queue.size();<a name="line.207"></a> -<span class="sourceLineNo">208</span> if (queueSize > this.logQueueWarnThreshold) {<a name="line.208"></a> -<span class="sourceLineNo">209</span> LOG.warn("WAL group " + logPrefix + " queue size: " + queueSize<a name="line.209"></a> -<span class="sourceLineNo">210</span> + " exceeds value of replication.source.log.queue.warn: " + logQueueWarnThreshold);<a name="line.210"></a> -<span class="sourceLineNo">211</span> }<a name="line.211"></a> -<span class="sourceLineNo">212</span> }<a name="line.212"></a> -<span class="sourceLineNo">213</span><a name="line.213"></a> -<span class="sourceLineNo">214</span> @Override<a name="line.214"></a> -<span class="sourceLineNo">215</span> public void addHFileRefs(TableName tableName, byte[] family, List<Pair<Path, Path>> pairs)<a name="line.215"></a> -<span class="sourceLineNo">216</span> throws ReplicationException {<a name="line.216"></a> -<span class="sourceLineNo">217</span> String peerId = peerClusterZnode;<a name="line.217"></a> -<span class="sourceLineNo">218</span> if (peerId.contains("-")) {<a name="line.218"></a> -<span class="sourceLineNo">219</span> // peerClusterZnode will be in the form peerId + "-" + rsZNode.<a name="line.219"></a> -<span class="sourceLineNo">220</span> // A peerId will not have "-" in its name, see HBASE-11394<a name="line.220"></a> -<span class="sourceLineNo">221</span> peerId = peerClusterZnode.split("-")[0];<a name="line.221"></a> -<span class="sourceLineNo">222</span> }<a name="line.222"></a> -<span class="sourceLineNo">223</span> Map<TableName, List<String>> tableCFMap = replicationPeers.getConnectedPeer(peerId).getTableCFs();<a name="line.223"></a> -<span class="sourceLineNo">224</span> if (tableCFMap != null) {<a name="line.224"></a> -<span class="sourceLineNo">225</span> List<String> tableCfs = tableCFMap.get(tableName);<a name="line.225"></a> -<span class="sourceLineNo">226</span> if (tableCFMap.containsKey(tableName)<a name="line.226"></a> -<span class="sourceLineNo">227</span> && (tableCfs == null || tableCfs.contains(Bytes.toString(family)))) {<a name="line.227"></a> -<span class="sourceLineNo">228</span> this.replicationQueues.addHFileRefs(peerId, pairs);<a name="line.228"></a> -<span class="sourceLineNo">229</span> metrics.incrSizeOfHFileRefsQueue(pairs.size());<a name="line.229"></a> -<span class="sourceLineNo">230</span> } else {<a name="line.230"></a> -<span class="sourceLineNo">231</span> LOG.debug("HFiles will not be replicated belonging to the table " + tableName + " family "<a name="line.231"></a> -<span class="sourceLineNo">232</span> + Bytes.toString(family) + " to peer id " + peerId);<a name="line.232"></a> -<span class="sourceLineNo">233</span> }<a name="line.233"></a> -<span class="sourceLineNo">234</span> } else {<a name="line.234"></a> -<span class="sourceLineNo">235</span> // user has explicitly not defined any table cfs for replication, means replicate all the<a name="line.235"></a> -<span class="sourceLineNo">236</span> // data<a name="line.236"></a> -<span class="sourceLineNo">237</span> this.replicationQueues.addHFileRefs(peerId, pairs);<a name="line.237"></a> -<span class="sourceLineNo">238</span> metrics.incrSizeOfHFileRefsQueue(pairs.size());<a name="line.238"></a> -<span class="sourceLineNo">239</span> }<a name="line.239"></a> -<span class="sourceLineNo">240</span> }<a name="line.240"></a> -<span class="sourceLineNo">241</span><a name="line.241"></a> -<span class="sourceLineNo">242</span> @Override<a name="line.242"></a> -<span class="sourceLineNo">243</span> public void run() {<a name="line.243"></a> -<span class="sourceLineNo">244</span> // mark we are running now<a name="line.244"></a> -<span class="sourceLineNo">245</span> this.sourceRunning = true;<a name="line.245"></a> -<span class="sourceLineNo">246</span> try {<a name="line.246"></a> -<span class="sourceLineNo">247</span> // start the endpoint, connect to the cluster<a name="line.247"></a> -<span class="sourceLineNo">248</span> Service service = replicationEndpoint.startAsync();<a name="line.248"></a> -<span class="sourceLineNo">249</span> final int waitTime = 10;<a name="line.249"></a> -<span class="sourceLineNo">250</span> service.awaitRunning(waitTime, TimeUnit.SECONDS);<a name="line.250"></a> -<span class="sourceLineNo">251</span> if (!service.isRunning()) {<a name="line.251"></a> -<span class="sourceLineNo">252</span> LOG.warn("ReplicationEndpoint was not started after waiting " + waitTime +<a name="line.252"></a> -<span class="sourceLineNo">253</span> " + seconds. Exiting");<a name="line.253"></a> -<span class="sourceLineNo">254</span> uninitialize();<a name="line.254"></a> -<span class="sourceLineNo">255</span> return;<a name="line.255"></a> -<span class="sourceLineNo">256</span> }<a name="line.256"></a> -<span class="sourceLineNo">257</span> } catch (Exception ex) {<a name="line.257"></a> -<span class="sourceLineNo">258</span> LOG.warn("Error starting ReplicationEndpoint, exiting", ex);<a name="line.258"></a> -<span class="sourceLineNo">259</span> throw new RuntimeException(ex);<a name="line.259"></a> -<span class="sourceLineNo">260</span> }<a name="line.260"></a> -<span class="sourceLineNo">261</span><a name="line.261"></a> -<span class="sourceLineNo">262</span> int sleepMultiplier = 1;<a name="line.262"></a> -<span class="sourceLineNo">263</span> // delay this until we are in an asynchronous thread<a name="line.263"></a> -<span class="sourceLineNo">264</span> while (this.isSourceActive() && this.peerClusterId == null) {<a name="line.264"></a> -<span class="sourceLineNo">265</span> this.peerClusterId = replicationEndpoint.getPeerUUID();<a name="line.265"></a> -<span class="sourceLineNo">266</span> if (this.isSourceActive() && this.peerClusterId == null) {<a name="line.266"></a> -<span class="sourceLineNo">267</span> if (sleepForRetries("Cannot contact the peer's zk ensemble", sleepMultiplier)) {<a name="line.267"></a> -<span class="sourceLineNo">268</span> sleepMultiplier++;<a name="line.268"></a> -<span class="sourceLineNo">269</span> }<a name="line.269"></a> -<span class="sourceLineNo">270</span> }<a name="line.270"></a> -<span class="sourceLineNo">271</span> }<a name="line.271"></a> -<span class="sourceLineNo">272</span><a name="line.272"></a> -<span class="sourceLineNo">273</span> // In rare case, zookeeper setting may be messed up. That leads to the incorrect<a name="line.273"></a> -<span class="sourceLineNo">274</span> // peerClusterId value, which is the same as the source clusterId<a name="line.274"></a> -<span class="sourceLineNo">275</span> if (clusterId.equals(peerClusterId) && !replicationEndpoint.canReplicateToSameCluster()) {<a name="line.275"></a> -<span class="sourceLineNo">276</span> this.terminate("ClusterId " + clusterId + " is replicating to itself: peerClusterId "<a name="line.276"></a> -<span class="sourceLineNo">277</span> + peerClusterId + " which is not allowed by ReplicationEndpoint:"<a name="line.277"></a> -<span class="sourceLineNo">278</span> + replicationEndpoint.getClass().getName(), null, false);<a name="line.278"></a> -<span class="sourceLineNo">279</span> this.manager.closeQueue(this);<a name="line.279"></a> -<span class="sourceLineNo">280</span> return;<a name="line.280"></a> -<span class="sourceLineNo">281</span> }<a name="line.281"></a> -<span class="sourceLineNo">282</span> LOG.info("Replicating " + clusterId + " -> " + peerClusterId);<a name="line.282"></a> -<span class="sourceLineNo">283</span><a name="line.283"></a> -<span class="sourceLineNo">284</span> initializeWALEntryFilter();<a name="line.284"></a> -<span class="sourceLineNo">285</span> // start workers<a name="line.285"></a> -<span class="sourceLineNo">286</span> for (Map.Entry<String, PriorityBlockingQueue<Path>> entry : queues.entrySet()) {<a name="line.286"></a> -<span class="sourceLineNo">287</span> String walGroupId = entry.getKey();<a name="line.287"></a> -<span class="sourceLineNo">288</span> PriorityBlockingQueue<Path> queue = entry.getValue();<a name="line.288"></a> -<span class="sourceLineNo">289</span> tryStartNewShipper(walGroupId, queue);<a name="line.289"></a> -<span class="sourceLineNo">290</span> }<a name="line.290"></a> -<span class="sourceLineNo">291</span> }<a name="line.291"></a> -<span class="sourceLineNo">292</span><a name="line.292"></a> -<span class="sourceLineNo">293</span> private void initializeWALEntryFilter() {<a name="line.293"></a> -<span class="sourceLineNo">294</span> // get the WALEntryFilter from ReplicationEndpoint and add it to default filters<a name="line.294"></a> -<span class="sourceLineNo">295</span> ArrayList<WALEntryFilter> filters = Lists.newArrayList(<a name="line.295"></a> -<span class="sourceLineNo">296</span> (WALEntryFilter)new SystemTableWALEntryFilter());<a name="line.296"></a> -<span class="sourceLineNo">297</span> WALEntryFilter filterFromEndpoint = this.replicationEndpoint.getWALEntryfilter();<a name="line.297"></a> -<span class="sourceLineNo">298</span> if (filterFromEndpoint != null) {<a name="line.298"></a> -<span class="sourceLineNo">299</span> filters.add(filterFromEndpoint);<a name="line.299"></a> -<span class="sourceLineNo">300</span> }<a name="line.300"></a> -<span class="sourceLineNo">301</span> filters.add(new ClusterMarkingEntryFilter(clusterId, peerClusterId, replicationEndpoint));<a name="line.301"></a> -<span class="sourceLineNo">302</span> this.walEntryFilter = new ChainWALEntryFilter(filters);<a name="line.302"></a> -<span class="sourceLineNo">303</span> }<a name="line.303"></a> -<span class="sourceLineNo">304</span><a name="line.304"></a> -<span class="sourceLineNo">305</span> protected void tryStartNewShipper(String walGroupId, PriorityBlockingQueue<Path> queue) {<a name="line.305"></a> -<span class="sourceLineNo">306</span> final ReplicationSourceShipper worker = new ReplicationSourceShipper(conf,<a name="line.306"></a> -<span class="sourceLineNo">307</span> walGroupId, queue, this);<a name="line.307"></a> -<span class="sourceLineNo">308</span> ReplicationSourceShipper extant = workerThreads.putIfAbsent(walGroupId, worker);<a name="line.308"></a> -<span class="sourceLineNo">309</span> if (extant != null) {<a name="line.309"></a> -<span class="sourceLineNo">310</span> LOG.debug("Someone has beat us to start a worker thread for wal group " + walGroupId);<a name="line.310"></a> -<span class="sourceLineNo">311</span> } else {<a name="line.311"></a> -<span class="sourceLineNo">312</span> LOG.debug("Starting up worker for wal group " + walGroupId);<a name="line.312"></a> -<span class="sourceLineNo">313</span> worker.startup(getUncaughtExceptionHandler());<a name="line.313"></a> -<span class="sourceLineNo">314</span> worker.setWALReader(startNewWALReader(worker.getName(), walGroupId, queue,<a name="line.314"></a> -<span class="sourceLineNo">315</span> worker.getStartPosition()));<a name="line.315"></a> -<span class="sourceLineNo">316</span> workerThreads.put(walGroupId, worker);<a name="line.316"></a> -<span class="sourceLineNo">317</span> }<a name="line.317"></a> -<span class="sourceLineNo">318</span> }<a name="line.318"></a> -<span class="sourceLineNo">319</span><a name="line.319"></a> -<span class="sourceLineNo">320</span> protected ReplicationSourceWALReader startNewWALReader(String threadName, String walGroupId,<a name="line.320"></a> -<span class="sourceLineNo">321</span> PriorityBlockingQueue<Path> queue, long startPosition) {<a name="line.321"></a> -<span class="sourceLineNo">322</span> ReplicationSourceWALReader walReader =<a name="line.322"></a> -<span class="sourceLineNo">323</span> new ReplicationSourceWALReader(fs, conf, queue, startPosition, walEntryFilter, this);<a name="line.323"></a> -<span class="sourceLineNo">324</span> return (ReplicationSourceWALReader) Threads.setDaemonThreadRunning(walReader,<a name="line.324"></a> -<span class="sourceLineNo">325</span> threadName + ".replicationSource.wal-reader." + walGroupId + "," + peerClusterZnode,<a name="line.325"></a> -<span class="sourceLineNo">326</span> getUncaughtExceptionHandler());<a name="line.326"></a> -<span class="sourceLineNo">327</span> }<a name="line.327"></a> -<span class="sourceLineNo">328</span><a name="line.328"></a> -<span class="sourceLineNo">329</span> public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {<a name="line.329"></a> -<span class="sourceLineNo">330</span> return new Thread.UncaughtExceptionHandler() {<a name="line.330"></a> -<span class="sourceLineNo">331</span> @Override<a name="line.331"></a> -<span class="sourceLineNo">332</span> public void uncaughtException(final Thread t, final Throwable e) {<a name="line.332"></a> -<span class="sourceLineNo">333</span> RSRpcServices.exitIfOOME(e);<a name="line.333"></a> -<span class="sourceLineNo">334</span> LOG.error("Unexpected exception in " + t.getName() + " currentPath=" + getCurrentPath(), e);<a name="line.334"></a> -<span class="sourceLineNo">335</span> stopper.stop("Unexpected exception in " + t.getName());<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> }<a name="line.338"></a> -<span class="sourceLineNo">339</span><a name="line.339"></a> -<span class="sourceLineNo">340</span> @Override<a name="line.340"></a> -<span class="sourceLineNo">341</span> public ReplicationEndpoint getReplicationEndpoint() {<a name="line.341"></a> -<span class="sourceLineNo">342</span> return this.replicationEndpoint;<a name="line.342"></a> -<span class="sourceLineNo">343</span> }<a name="line.343"></a> -<span class="sourceLineNo">344</span><a name="line.344"></a> -<span class="sourceLineNo">345</span> @Override<a name="line.345"></a> -<span class="sourceLineNo">346</span> public ReplicationSourceManager getSourceManager() {<a name="line.346"></a> -<span class="sourceLineNo">347</span> return this.manager;<a name="line.347"></a> -<span class="sourceLineNo">348</span> }<a name="line.348"></a> -<span class="sourceLineNo">349</span><a name="line.349"></a> -<span class="sourceLineNo">350</span> @Override<a name="line.350"></a> -<span class="sourceLineNo">351</span> public void tryThrottle(int batchSize) throws InterruptedException {<a name="line.351"></a> -<span class="sourceLineNo">352</span> checkBandwidthChangeAndResetThrottler();<a name="line.352"></a> -<span class="sourceLineNo">353</span> if (throttler.isEnabled()) {<a name="line.353"></a> -<span class="sourceLineNo">354</span> long sleepTicks = throttler.getNextSleepInterval(batchSize);<a name="line.354"></a> -<span class="sourceLineNo">355</span> if (sleepTicks > 0) {<a name="line.355"></a> -<span class="sourceLineNo">356</span> if (LOG.isTraceEnabled()) {<a name="line.356"></a> -<span class="sourceLineNo">357</span> LOG.trace("To sleep " + sleepTicks + "ms for throttling control");<a name="line.357"></a> -<span class="sourceLineNo">358</span> }<a name="line.358"></a> -<span class="sourceLineNo">359</span> Thread.sleep(sleepTicks);<a name="line.359"></a> -<span class="sourceLineNo">360</span> // reset throttler's cycle start tick when sleep for throttling occurs<a name="line.360"></a> -<span class="sourceLineNo">361</span> throttler.resetStartTick();<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> }<a name="line.364"></a> -<span class="sourceLineNo">365</span><a name="line.365"></a> -<span class="sourceLineNo">366</span> private void checkBandwidthChangeAndResetThrottler() {<a name="line.366"></a> -<span class="sourceLineNo">367</span> long peerBandwidth = getCurrentBandwidth();<a name="line.367"></a> -<span class="sourceLineNo">368</span> if (peerBandwidth != currentBandwidth) {<a name="line.368"></a> -<span class="sourceLineNo">369</span> currentBandwidth = peerBandwidth;<a name="line.369"></a> -<span class="sourceLineNo">370</span> throttler.setBandwidth((double) currentBandwidth / 10.0);<a name="line.370"></a> -<span class="sourceLineNo">371</span> LOG.info("ReplicationSource : " + peerId<a name="line.371"></a> -<span class="sourceLineNo">372</span> + " bandwidth throttling changed, currentBandWidth=" + currentBandwidth);<a name="line.372"></a> -<span class="sourceLineNo">373</span> }<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> private long getCurrentBandwidth() {<a name="line.376"></a> -<span class="sourceLineNo">377</span> ReplicationPeer replicationPeer = this.replicationPeers.getConnectedPeer(peerId);<a name="line.377"></a> -<span class="sourceLineNo">378</span> long peerBandwidth = replicationPeer != null ? replicationPeer.getPeerBandwidth() : 0;<a name="line.378"></a> -<span class="sourceLineNo">379</span> // user can set peer bandwidth to 0 to use default bandwidth<a name="line.379"></a> -<span class="sourceLineNo">380</span> return peerBandwidth != 0 ? peerBandwidth : defaultBandwidth;<a name="line.380"></a> -<span class="sourceLineNo">381</span> }<a name="line.381"></a> -<span class="sourceLineNo">382</span><a name="line.382"></a> -<span class="sourceLineNo">383</span> private void uninitialize() {<a name="line.383"></a> -<span class="sourceLineNo">384</span> LOG.debug("Source exiting " + this.peerId);<a name="line.384"></a> -<span class="sourceLineNo">385</span> metrics.clear();<a name="line.385"></a> -<span class="sourceLineNo">386</span> if (replicationEndpoint.state() == Service.State.STARTING<a name="line.386"></a> -<span class="sourceLineNo">387</span> || replicationEndpoint.state() == Service.State.RUNNING) {<a name="line.387"></a> -<span class="sourceLineNo">388</span> replicationEndpoint.stopAsync();<a name="line.388"></a> -<span class="sourceLineNo">389</span> final int waitTime = 10;<a name="line.389"></a> -<span class="sourceLineNo">390</span> try {<a name="line.390"></a> -<span class="sourceLineNo">391</span> replicationEndpoint.awaitTerminated(waitTime, TimeUnit.SECONDS);<a name="line.391"></a> -<span class="sourceLineNo">392</span> } catch (TimeoutException e) {<a name="line.392"></a> -<span class="sourceLineNo">393</span> LOG.warn("Failed termination after " + waitTime + " seconds.");<a name="line.393"></a> -<span class="sourceLineNo">394</span> }<a name="line.394"></a> -<span class="sourceLineNo">395</span> }<a name="line.395"></a> -<span class="sourceLineNo">396</span> }<a name="line.396"></a> -<span class="sourceLineNo">397</span><a name="line.397"></a> -<span class="sourceLineNo">398</span> /**<a name="line.398"></a> -<span class="sourceLineNo">399</span> * Do the sleeping logic<a name="line.399"></a> -<span class="sourceLineNo">400</span> * @param msg Why we sleep<a name="line.400"></a> -<span class="sourceLineNo">401</span> * @param sleepMultiplier by how many times the default sleeping time is augmented<a name="line.401"></a> -<span class="sourceLineNo">402</span> * @return True if <code>sleepMultiplier</code> is &lt; <code>maxRetriesMultiplier</code><a name="line.402"></a> -<span class="sourceLineNo">403</span> */<a name="line.403"></a> -<span class="sourceLineNo">404</span> protected boolean sleepForRetries(String msg, int sleepMultiplier) {<a name="line.404"></a> -<span class="sourceLineNo">405</span> try {<a name="line.405"></a> -<span class="sourceLineNo">406</span> if (LOG.isTraceEnabled()) {<a name="line.406"></a> -<span class="sourceLineNo">407</span> LOG.trace(msg + ", sleeping " + sleepForRetries + " times " + sleepMultiplier);<a name="line.407"></a> -<span class="sourceLineNo">408</span> }<a name="line.408"></a> -<span class="sourceLineNo">409</span> Thread.sleep(this.sleepForRetries * sleepMultiplier);<a name="line.409"></a> -<span class="sourceLineNo">410</span> } catch (InterruptedException e) {<a name="line.410"></a> -<span class="sourceLineNo">411</span> LOG.debug("Interrupted while sleeping between retries");<a name="line.411"></a> -<span class="sourceLineNo">412</span> Thread.currentThread().interrupt();<a name="line.412"></a> -<span class="sourceLineNo">413</span> }<a name="line.413"></a> -<span class="sourceLineNo">414</span> return sleepMultiplier < maxRetriesMultiplier;<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> /**<a name="line.417"></a> -<span class="sourceLineNo">418</span> * check whether the peer is enabled or not<a name="line.418"></a> -<span class="sourceLineNo">419</span> *<a name="line.419"></a> -<span class="sourceLineNo">420</span> * @return true if the peer is enabled, otherwise false<a name="line.420"></a> -<span class="sourceLineNo">421</span> */<a name="line.421"></a> -<span class="sourceLineNo">422</span> @Override<a name="line.422"></a> -<span class="sourceLineNo">423</span> public boolean isPeerEnabled() {<a name="line.423"></a> -<span class="sourceLineNo">424</span> return this.replicationPeers.getStatusOfPeer(this.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> @Override<a name="line.427"></a> -<span class="sourceLineNo">428</span> public void startup() {<a name="line.428"></a> -<span class="sourceLineNo">429</span> String n = Thread.currentThread().getName();<a name="line.429"></a> -<span class="sourceLineNo">430</span> Thread.UncaughtExceptionHandler handler = new Thread.UncaughtExceptionHandler() {<a name="line.430"></a> -<span class="sourceLineNo">431</span> @Override<a name="line.431"></a> -<span class="sourceLineNo">432</span> public void uncaughtException(final Thread t, final Throwable e) {<a name="line.432"></a> -<span class="sourceLineNo">433</span> LOG.error("Unexpected exception in ReplicationSource", e);<a name="line.433"></a> -<span class="sourceLineNo">434</span> }<a name="line.434"></a> -<span class="sourceLineNo">435</span> };<a name="line.435"></a> -<span class="sourceLineNo">436</span> Threads<a name="line.436"></a> -<span class="sourceLineNo">437</span> .setDaemonThreadRunning(this, n + ".replicationSource," + this.peerClusterZnode, handler);<a name="line.437"></a> -<span class="sourceLineNo">438</span> }<a name="line.438"></a> -<span class="sourceLineNo">439</span><a name="line.439"></a> -<span class="sourceLineNo">440</span> @Override<a name="line.440"></a> -<span class="sourceLineNo">441</span> public void terminate(String reason) {<a name="line.441"></a> -<span class="sourceLineNo">442</span> terminate(reason, null);<a name="line.442"></a> -<span class="sourceLineNo">443</span> }<a name="line.443"></a> -<span class="sourceLineNo">444</span><a name="line.444"></a> -<span class="sourceLineNo">445</span> @Override<a name="line.445"></a> -<span class="sourceLineNo">446</span> public void terminate(String reason, Exception cause) {<a name="line.446"></a> -<span class="sourceLineNo">447</span> terminate(reason, cause, true);<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> public void terminate(String reason, Exception cause, boolean join) {<a name="line.450"></a> -<span class="sourceLineNo">451</span> if (cause == null) {<a name="line.451"></a> -<span class="sourceLineNo">452</span> LOG.info("Closing source "<a name="line.452"></a> -<span class="sourceLineNo">453</span> + this.peerClusterZnode + " because: " + reason);<a name="line.453"></a> -<span class="sourceLineNo">454</span><a name="line.454"></a> -<span class="sourceLineNo">455</span> } else {<a name="line.455"></a> -<span class="sourceLineNo">456</span> LOG.error("Closing source " + this.peerClusterZnode<a name="line.456"></a> -<span class="sourceLineNo">457</span> + " because an error occurred: " + reason, cause);<a name="line.457"></a> -<span class="sourceLineNo">458</span> }<a name="line.458"></a> -<span class="sourceLineNo">459</span> this.sourceRunning = false;<a name="line.459"></a> -<span class="sourceLineNo">460</span> Collection<ReplicationSourceShipper> workers = workerThreads.values();<a name="line.460"></a> -<span class="sourceLineNo">461</span> for (ReplicationSourceShipper worker : workers) {<a name="line.461"></a> -<span class="sourceLineNo">462</span> worker.stopWorker();<a name="line.462"></a> -<span class="sourceLineNo">463</span> worker.entryReader.interrupt();<a name="line.463"></a> -<span class="sourceLineNo">464</span> worker.interrupt();<a name="line.464"></a> +<span class="sourceLineNo">131</span> public static final String WAIT_ON_ENDPOINT_SECONDS =<a name="line.131"></a> +<span class="sourceLineNo">132</span> "hbase.replication.wait.on.endpoint.seconds";<a name="line.132"></a> +<span class="sourceLineNo">133</span> public static final int DEFAULT_WAIT_ON_ENDPOINT_SECONDS = 30;<a name="line.133"></a> +<span class="sourceLineNo">134</span> private int waitOnEndpointSeconds = -1;<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> * Instantiation method used by region servers<a name="line.137"></a> +<span class="sourceLineNo">138</span> *<a name="line.138"></a> +<span class="sourceLineNo">139</span> * @param conf configuration to use<a name="line.139"></a> +<span class="sourceLineNo">140</span> * @param fs file system to use<a name="line.140"></a> +<span class="sourceLineNo">141</span> * @param manager replication manager to ping to<a name="line.141"></a> +<span class="sourceLineNo">142</span> * @param stopper the atomic boolean to use to stop the regionserver<a name="line.142"></a> +<span class="sourceLineNo">143</span> * @param peerClusterZnode the name of our znode<a name="line.143"></a> +<span class="sourceLineNo">144</span> * @param clusterId unique UUID for the cluster<a name="line.144"></a> +<span class="sourceLineNo">145</span> * @param replicationEndpoint the replication endpoint implementation<a name="line.145"></a> +<span class="sourceLineNo">146</span> * @param metrics metrics for replication source<a name="line.146"></a> +<span class="sourceLineNo">147</span> * @throws IOException<a name="line.147"></a> +<span class="sourceLineNo">148</span> */<a name="line.148"></a> +<span class="sourceLineNo">149</span> @Override<a name="line.149"></a> +<span class="sourceLineNo">150</span> public void init(final Configuration conf, final FileSystem fs,<a name="line.150"></a> +<span class="sourceLineNo">151</span> final ReplicationSourceManager manager, final ReplicationQueues replicationQueues,<a name="line.151"></a> +<span class="sourceLineNo">152</span> final ReplicationPeers replicationPeers, final Stoppable stopper,<a name="line.152"></a> +<span class="sourceLineNo">153</span> final String peerClusterZnode, final UUID clusterId, ReplicationEndpoint replicationEndpoint,<a name="line.153"></a> +<span class="sourceLineNo">154</span> final MetricsSource metrics)<a name="line.154"></a> +<span class="sourceLineNo">155</span> throws IOException {<a name="line.155"></a> +<span class="sourceLineNo">156</span> this.stopper = stopper;<a name="line.156"></a> +<span class="sourceLineNo">157</span> this.conf = HBaseConfiguration.create(conf);<a name="line.157"></a> +<span class="sourceLineNo">158</span> this.waitOnEndpointSeconds =<a name="line.158"></a> +<span class="sourceLineNo">159</span> this.conf.getInt(WAIT_ON_ENDPOINT_SECONDS, DEFAULT_WAIT_ON_ENDPOINT_SECONDS);<a name="line.159"></a> +<span class="sourceLineNo">160</span> decorateConf();<a name="line.160"></a> +<span class="sourceLineNo">161</span> this.sleepForRetries =<a name="line.161"></a> +<span class="sourceLineNo">162</span> this.conf.getLong("replication.source.sleepforretries", 1000); // 1 second<a name="line.162"></a> +<span class="sourceLineNo">163</span> this.maxRetriesMultiplier =<a name="line.163"></a> +<span class="sourceLineNo">164</span> this.conf.getInt("replication.source.maxretriesmultiplier", 300); // 5 minutes @ 1 sec per<a name="line.164"></a> +<span class="sourceLineNo">165</span> this.queueSizePerGroup = this.conf.getInt("hbase.regionserver.maxlogs", 32);<a name="line.165"></a> +<span class="sourceLineNo">166</span> this.replicationQueues = replicationQueues;<a name="line.166"></a> +<span class="sourceLineNo">167</span> this.replicationPeers = replicationPeers;<a name="line.167"></a> +<span class="sourceLineNo">168</span> this.manager = manager;<a name="line.168"></a> +<span class="sourceLineNo">169</span> this.fs = fs;<a name="line.169"></a> +<span class="sourceLineNo">170</span> this.metrics = metrics;<a name="line.170"></a> +<span class="sourceLineNo">171</span> this.clusterId = clusterId;<a name="line.171"></a> +<span class="sourceLineNo">172</span><a name="line.172"></a> +<span class="sourceLineNo">173</span> this.peerClusterZnode = peerClusterZnode;<a name="line.173"></a> +<span class="sourceLineNo">174</span> this.replicationQueueInfo = new ReplicationQueueInfo(peerClusterZnode);<a name="line.174"></a> +<span class="sourceLineNo">175</span> // ReplicationQueueInfo parses the peerId out of the znode for us<a name="line.175"></a> +<span class="sourceLineNo">176</span> this.peerId = this.replicationQueueInfo.getPeerId();<a name="line.176"></a> +<span class="sourceLineNo">177</span> this.logQueueWarnThreshold = this.conf.getInt("replication.source.log.queue.warn", 2);<a name="line.177"></a> +<span class="sourceLineNo">178</span> this.replicationEndpoint = replicationEndpoint;<a name="line.178"></a> +<span class="sourceLineNo">179</span><a name="line.179"></a> +<span class="sourceLineNo">180</span> defaultBandwidth = this.conf.getLong("replication.source.per.peer.node.bandwidth", 0);<a name="line.180"></a> +<span class="sourceLineNo">181</span> currentBandwidth = getCurrentBandwidth();<a name="line.181"></a> +<span class="sourceLineNo">182</span> this.throttler = new ReplicationThrottler((double) currentBandwidth / 10.0);<a name="line.182"></a> +<span class="sourceLineNo">183</span> this.totalBufferUsed = manager.getTotalBufferUsed();<a name="line.183"></a> +<span class="sourceLineNo">184</span> LOG.info("peerClusterZnode=" + peerClusterZnode + ", ReplicationSource : " + peerId<a name="line.184"></a> +<span class="sourceLineNo">185</span> + ", currentBandwidth=" + this.currentBandwidth);<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> private void decorateConf() {<a name="line.188"></a> +<span class="sourceLineNo">189</span> String replicationCodec = this.conf.get(HConstants.REPLICATION_CODEC_CONF_KEY);<a name="line.189"></a> +<span class="sourceLineNo">190</span> if (StringUtils.isNotEmpty(replicationCodec)) {<a name="line.190"></a> +<span class="sourceLineNo">191</span> this.conf.set(HConstants.RPC_CODEC_CONF_KEY, replicationCodec);<a name="line.191"></a> +<span class="sourceLineNo">192</span> }<a name="line.192"></a> +<span class="sourceLineNo">193</span> }<a name="line.193"></a> +<span class="sourceLineNo">194</span><a name="line.194"></a> +<span class="sourceLineNo">195</span> @Override<a name="line.195"></a> +<span class="sourceLineNo">196</span> public void enqueueLog(Path log) {<a name="line.196"></a> +<span class="sourceLineNo">197</span> String logPrefix = AbstractFSWALProvider.getWALPrefixFromWALName(log.getName());<a name="line.197"></a> +<span class="sourceLineNo">198</span> PriorityBlockingQueue<Path> queue = queues.get(logPrefix);<a name="line.198"></a> +<span class="sourceLineNo">199</span> if (queue == null) {<a name="line.199"></a> +<span class="sourceLineNo">200</span> queue = new PriorityBlockingQueue<>(queueSizePerGroup, new LogsComparator());<a name="line.200"></a> +<span class="sourceLineNo">201</span> queues.put(logPrefix, queue);<a name="line.201"></a> +<span class="sourceLineNo">202</span> if (this.sourceRunning) {<a name="line.202"></a> +<span class="sourceLineNo">203</span> // new wal group observed after source startup, start a new worker thread to track it<a name="line.203"></a> +<span class="sourceLineNo">204</span> // notice: it's possible that log enqueued when this.running is set but worker thread<a name="line.204"></a> +<span class="sourceLineNo">205</span> // still not launched, so it's necessary to check workerThreads before start the worker<a name="line.205"></a> +<span class="sourceLineNo">206</span> tryStartNewShipper(logPrefix, queue);<a name="line.206"></a> +<span class="sourceLineNo">207</span> }<a name="line.207"></a> +<span class="sourceLineNo">208</span> }<a name="line.208"></a> +<span class="sourceLineNo">209</span> queue.put(log);<a name="line.209"></a> +<span class="sourceLineNo">210</span> this.metrics.incrSizeOfLogQueue();<a name="line.210"></a> +<span class="sourceLineNo">211</span> // This will log a warning for each new log that gets created above the warn threshold<a name="line.211"></a> +<span class="sourceLineNo">212</span> int queueSize = queue.size();<a name="line.212"></a> +<span class="sourceLineNo">213</span> if (queueSize > this.logQueueWarnThreshold) {<a name="line.213"></a> +<span class="sourceLineNo">214</span> LOG.warn("WAL group " + logPrefix + " queue size: " + queueSize<a name="line.214"></a> +<span class="sourceLineNo">215</span> + " exceeds value of replication.source.log.queue.warn: " + logQueueWarnThreshold);<a name="line.215"></a> +<span class="sourceLineNo">216</span> }<a name="line.216"></a> +<span class="sourceLineNo">217</span> }<a name="line.217"></a> +<span class="sourceLineNo">218</span><a name="line.218"></a> +<span class="sourceLineNo">219</span> @Override<a name="line.219"></a> +<span class="sourceLineNo">220</span> public void addHFileRefs(TableName tableName, byte[] family, List<Pair<Path, Path>> pairs)<a name="line.220"></a> +<span class="sourceLineNo">221</span> throws ReplicationException {<a name="line.221"></a> +<span class="sourceLineNo">222</span> String peerId = peerClusterZnode;<a name="line.222"></a> +<span class="sourceLineNo">223</span> if (peerId.contains("-")) {<a name="line.223"></a> +<span class="sourceLineNo">224</span> // peerClusterZnode will be in the form peerId + "-" + rsZNode.<a name="line.224"></a> +<span class="sourceLineNo">225</span> // A peerId will not have "-" in its name, see HBASE-11394<a name="line.225"></a> +<span class="sourceLineNo">226</span> peerId = peerClusterZnode.split("-")[0];<a name="line.226"></a> +<span class="sourceLineNo">227</span> }<a name="line.227"></a> +<span class="sourceLineNo">228</span> Map<TableName, List<String>> tableCFMap = replicationPeers.getConnectedPeer(peerId).getTableCFs();<a name="line.228"></a> +<span class="sourceLineNo">229</span> if (tableCFMap != null) {<a name="line.229"></a> +<span class="sourceLineNo">230</span> List<String> tableCfs = tableCFMap.get(tableName);<a name="line.230"></a> +<span class="sourceLineNo">231</span> if (tableCFMap.containsKey(tableName)<a name="line.231"></a> +<span class="sourceLineNo">232</span> && (tableCfs == null || tableCfs.contains(Bytes.toString(family)))) {<a name="line.232"></a> +<span class="sourceLineNo">233</span> this.replicationQueues.addHFileRefs(peerId, pairs);<a name="line.233"></a> +<span class="sourceLineNo">234</span> metrics.incrSizeOfHFileRefsQueue(pairs.size());<a name="line.234"></a> +<span class="sourceLineNo">235</span> } else {<a name="line.235"></a> +<span class="sourceLineNo">236</span> LOG.debug("HFiles will not be replicated belonging to the table " + tableName + " family "<a name="line.236"></a> +<span class="sourceLineNo">237</span> + Bytes.toString(family) + " to peer id " + peerId);<a name="line.237"></a> +<span class="sourceLineNo">238</span> }<a name="line.238"></a> +<span class="sourceLineNo">239</span> } else {<a name="line.239"></a> +<span class="sourceLineNo">240</span> // user has explicitly not defined any table cfs for replication, means replicate all the<a name="line.240"></a> +<span class="sourceLineNo">241</span> // data<a name="line.241"></a> +<span class="sourceLineNo">242</span> this.replicationQueues.addHFileRefs(peerId, pairs);<a name="line.242"></a> +<span class="sourceLineNo">243</span> metrics.incrSizeOfHFileRefsQueue(pairs.size());<a name="line.243"></a> +<span class="sourceLineNo">244</span> }<a name="line.244"></a> +<span class="sourceLineNo">245</span> }<a name="line.245"></a> +<span class="sourceLineNo">246</span><a name="line.246"></a> +<span class="sourceLineNo">247</span> @Override<a name="line.247"></a> +<span class="sourceLineNo">248</span> public void run() {<a name="line.248"></a> +<span class="sourceLineNo">249</span> // mark we are running now<a name="line.249"></a> +<span class="sourceLineNo">250</span> this.sourceRunning = true;<a name="line.250"></a> +<span class="sourceLineNo">251</span> try {<a name="line.251"></a> +<span class="sourceLineNo">252</span> // start the endpoint, connect to the cluster<a name="line.252"></a> +<span class="sourceLineNo">253</span> this.replicationEndpoint.start();<a name="line.253"></a> +<span class="sourceLineNo">254</span> this.replicationEndpoint.awaitRunning(this.waitOnEndpointSeconds, TimeUnit.SECONDS);<a name="line.254"></a> +<span class="sourceLineNo">255</span> } catch (Exception ex) {<a name="line.255"></a> +<span class="sourceLineNo">256</span> LOG.warn("Error starting ReplicationEndpoint, exiting", ex);<a name="line.256"></a> +<span class="sourceLineNo">257</span> uninitialize();<a name="line.257"></a> +<span class="sourceLineNo">258</span> throw new RuntimeException(ex);<a name="line.258"></a> +<span class="sourceLineNo">259</span> }<a name="line.259"></a> +<span class="sourceLineNo">260</span><a name="line.260"></a> +<span class="sourceLineNo">261</span> int sleepMultiplier = 1;<a name="line.261"></a> +<span class="sourceLineNo">262</span> // delay this until we are in an asynchronous thread<a name="line.262"></a> +<span class="sourceLineNo">263</span> while (this.isSourceActive() && this.peerClusterId == null) {<a name="line.263"></a> +<span class="sourceLineNo">264</span> this.peerClusterId = replicationEndpoint.getPeerUUID();<a name="line.264"></a> +<span class="sourceLineNo">265</span> if (this.isSourceActive() && this.peerClusterId == null) {<a name="line.265"></a> +<span class="sourceLineNo">266</span> if (sleepForRetries("Cannot contact the peer's zk ensemble", sleepMultiplier)) {<a name="line.266"></a> +<span class="sourceLineNo">267</span> sleepMultiplier++;<a name="line.267"></a> +<span class="sourceLineNo">268</span> }<a name="line.268"></a> +<span class="sourceLineNo">269</span> }<a name="line.269"></a> +<span class="sourceLineNo">270</span> }<a name="line.270"></a> +<span class="sourceLineNo">271</span><a name="line.271"></a> +<span class="sourceLineNo">272</span> // In rare case, zookeeper setting may be messed up. That leads to the incorrect<a name="line.272"></a> +<span class="sourceLineNo">273</span> // peerClusterId value, which is the same as the source clusterId<a name="line.273"></a> +<span class="sourceLineNo">274</span> if (clusterId.equals(peerClusterId) && !replicationEndpoint.canReplicateToSameCluster()) {<a name="line.274"></a> +<span class="sourceLineNo">275</span> this.terminate("ClusterId " + clusterId + " is replicating to itself: peerClusterId "<a name="line.275"></a> +<span class="sourceLineNo">276</span> + peerClusterId + " which is not allowed by ReplicationEndpoint:"<a name="line.276"></a> +<span class="sourceLineNo">277</span> + replicationEndpoint.getClass().getName(), null, false);<a name="line.277"></a> +<span class="sourceLineNo">278</span> this.manager.closeQueue(this);<a name="line.278"></a> +<span class="sourceLineNo">279</span> return;<a name="line.279"></a> +<span class="sourceLineNo">280</span> }<a name="line.280"></a> +<span class="sourceLineNo">281</span> LOG.info("Replicating " + clusterId + " -> " + peerClusterId);<a name="line.281"></a> +<span class="sourceLineNo">282</span><a name="line.282"></a> +<span class="sourceLineNo">283</span> initializeWALEntryFilter();<a name="line.283"></a> +<span class="sourceLineNo">284</span> // start workers<a name="line.284"></a> +<span class="sourceLineNo">285</span> for (Map.Entry<String, PriorityBlockingQueue<Path>> entry : queues.entrySet()) {<a name="line.285"></a> +<span class="sourceLineNo">286</span> String walGroupId = entry.getKey();<a name="line.286"></a> +<span class="sourceLineNo">287</span> PriorityBlockingQueue<Path> queue = entry.getValue();<a name="line.287"></a> +<span class="sourceLineNo">288</span> tryStartNewShipper(walGroupId, queue);<a name="line.288"></a> +<span class="sourceLineNo">289</span> }<a name="line.289"></a> +<span class="sourceLineNo">290</span> }<a name="line.290"></a> +<span class="sourceLineNo">291</span><a name="line.291"></a> +<span class="sourceLineNo">292</span> private void initializeWALEntryFilter() {<a name="line.292"></a> +<span class="sourceLineNo">293</span> // get the WALEntryFilter from ReplicationEndpoint and add it to default filters<a name="line.293"></a> +<span class="sourceLineNo">294</span> ArrayList<WALEntryFilter> filters = Lists.newArrayList(<a name="line.294"></a> +<span class="sourceLineNo">295</span> (WALEntryFilter)new SystemTableWALEntryFilter());<a name="line.295"></a> +<span class="sourceLineNo">296</span> WALEntryFilter filterFromEndpoint = this.replicationEndpoint.getWALEntryfilter();<a name="line.296"></a> +<span class="sourceLineNo">297</span> if (filterFromEndpoint != null) {<a name="line.297"></a> +<span class="sourceLineNo">298</span> filters.add(filterFromEndpoint);<a name="line.298"></a> +<span class="sourceLineNo">299</span> }<a name="line.299"></a> +<span class="sourceLineNo">300</span> filters.add(new ClusterMarkingEntryFilter(clusterId, peerClusterId, replicationEndpoint));<a name="line.300"></a> +<span class="sourceLineNo">301</span> this.walEntryFilter = new ChainWALEntryFilter(filters);<a name="line.301"></a> +<span class="sourceLineNo">302</span> }<a name="line.302"></a> +<span class="sourceLineNo">303</span><a name="line.303"></a> +<span class="sourceLineNo">304</span> protected void tryStartNewShipper(String walGroupId, PriorityBlockingQueue<Path> queue) {<a name="line.304"></a> +<span class="sourceLineNo">305</span> final ReplicationSourceShipper worker = new ReplicationSourceShipper(conf,<a name="line.305"></a> +<span class="sourceLineNo">306</span> walGroupId, queue, this);<a name="line.306"></a> +<span class="sourceLineNo">307</span> ReplicationSourceShipper extant = workerThreads.putIfAbsent(walGroupId, worker);<a name="line.307"></a> +<span class="sourceLineNo">308</span> if (extant != null) {<a name="line.308"></a> +<span class="sourceLineNo">309</span> LOG.debug("Someone has beat us to start a worker thread for wal group " + walGroupId);<a name="line.309"></a> +<span class="sourceLineNo">310</span> } else {<a name="line.310"></a> +<span class="sourceLineNo">311</span> LOG.debug("Starting up worker for wal group " + walGroupId);<a name="line.311"></a> +<span class="sourceLineNo">312</span> worker.startup(getUncaughtExceptionHandler());<a name="line.312"></a> +<span class="sourceLineNo">313</span> worker.setWALReader(startNewWALReader(worker.getName(), walGroupId, queue,<a name="line.313"></a> +<span class="sourceLineNo">314</span> worker.getStartPosition()));<a name="line.314"></a> +<span class="sourceLineNo">315</span> workerThreads.put(walGroupId, worker);<a name="line.315"></a> +<span class="sourceLineNo">316</span> }<a name="line.316"></a> +<span class="sourceLineNo">317</span> }<a name="line.317"></a> +<span class="sourceLineNo">318</span><a name="line.318"></a> +<span class="sourceLineNo">319</span> protected ReplicationSourceWALReader startNewWALReader(String threadName, String walGroupId,<a name="line.319"></a> +<span class="sourceLineNo">320</span> PriorityBlockingQueue<Path> queue, long startPosition) {<a name="line.320"></a> +<span class="sourceLineNo">321</span> ReplicationSourceWALReader walReader =<a name="line.321"></a> +<span class="sourceLineNo">322</span> new ReplicationSourceWALReader(fs, conf, queue, startPosition, walEntryFilter, this);<a name="line.322"></a> +<span class="sourceLineNo">323</span> return (ReplicationSourceWALReader) Threads.setDaemonThreadRunning(walReader,<a name="line.323"></a> +<span class="sourceLineNo">324</span> threadName + ".replicationSource.wal-reader." + walGroupId + "," + peerClusterZnode,<a name="line.324"></a> +<span class="sourceLineNo">325</span> getUncaughtExceptionHandler());<a name="line.325"></a> +<span class="sourceLineNo">326</span> }<a name="line.326"></a> +<span class="sourceLineNo">327</span><a name="line.327"></a> +<span class="sourceLineNo">328</span> public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {<a name="line.328"></a> +<span class="sourceLineNo">329</span> return new Thread.UncaughtExceptionHandler() {<a name="line.329"></a> +<span class="sourceLineNo">330</span> @Override<a name="line.330"></a> +<span class="sourceLineNo">331</span> public void uncaughtException(final Thread t, final Throwable e) {<a name="line.331"></a> +<span class="sourceLineNo">332</span> RSRpcServices.exitIfOOME(e);<a name="line.332"></a> +<span class="sourceLineNo">333</span> LOG.error("Unexpected exception in " + t.getName() + " currentPath=" + getCurrentPath(), e);<a name="line.333"></a> +<span class="sourceLineNo">334</span> stopper.stop("Unexpected exception in " + t.getName());<a name="line.334"></a> +<span class="sourceLineNo">335</span> }<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><a name="line.338"></a> +<span class="sourceLineNo">339</span> @Override<a name="line.339"></a> +<span class="sourceLineNo">340</span> public ReplicationEndpoint getReplicationEndpoint() {<a name="line.340"></a> +<span class="sourceLineNo">341</span> return this.replicationEndpoint;<a name="line.341"></a> +<span class="sourceLineNo">342</span> }<a name="line.342"></a> +<span class="sourceLineNo">343</span><a name="line.343"></a> +<span class="sourceLineNo">344</span> @Override<a name="line.344"></a> +<span class="sourceLineNo">345</span> public ReplicationSourceManager getSourceManager() {<a name="line.345"></a> +<span class="sourceLineNo">346</span> return this.manager;<a name="line.346"></a> +<span class="sourceLineNo">347</span> }<a name="line.347"></a> +<span class="sourceLineNo">348</span><a name="line.348"></a> +<span class="sourceLineNo">349</span> @Override<a name="line.349"></a> +<span class="sourceLineNo">350</span> public void tryThrottle(int batchSize) throws InterruptedException {<a name="line.350"></a> +<span class="sourceLineNo">351</span> checkBandwidthChangeAndResetThrottler();<a name="line.351"></a> +<span class="sourceLineNo">352</span> if (throttler.isEnabled()) {<a name="line.352"></a> +<span class="sourceLineNo">353</span> long sleepTicks = throttler.getNextSleepInterval(batchSize);<a name="line.353"></a> +<span class="sourceLineNo">354</span> if (sleepTicks > 0) {<a name="line.354"></a> +<span class="sourceLineNo">355</span> if (LOG.isTraceEnabled()) {<a name="line.355"></a> +<span class="sourceLineNo">356</span> LOG.trace("To sleep " + sleepTicks + "ms for throttling control");<a name="line.356"></a> +<span class="sourceLineNo">357</span> }<a name="line.357"></a> +<span class="sourceLineNo">358</span> Thread.sleep(sleepTicks);<a name="line.358"></a> +<span class="sourceLineNo">359</span> // reset throttler's cycle start tick when sleep for throttling occurs<a name="line.359"></a> +<span class="sourceLineNo">360</span> throttler.resetStartTick();<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><a name="line.364"></a> +<span class="sourceLineNo">365</span> private void checkBandwidthChangeAndResetThrottler() {<a name="line.365"></a> +<span class="sourceLineNo">366</span> long peerBandwidth = getCurrentBandwidth();<a name="line.366"></a> +<span class="sourceLineNo">367</span> if (peerBandwidth != currentBandwidth) {<a name="line.367"></a> +<span class="sourceLineNo">368</span> currentBandwidth = peerBandwidth;<a name="line.368"></a> +<span class="sourceLineNo">369</span> throttler.setBandwidth((double) currentBandwidth / 10.0);<a name="line.369"></a> +<span class="sourceLineNo">370</span> LOG.info("ReplicationSource : " + peerId<a name="line.370"></a> +<span class="sourceLineNo">371</span> + " bandwidth throttling changed, currentBandWidth=" + currentBandwidth);<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><a name="line.374"></a> +<span class="sourceLineNo">375</span> private long getCurrentBandwidth() {<a name="line.375"></a> +<span class="sourceLineNo">376</span> ReplicationPeer replicationPeer = this.replicationPeers.getConnectedPeer(peerId);<a name="line.376"></a> +<span class="sourceLineNo">377</span> long peerBandwidth = replicationPeer != null ? replicationPeer.getPeerBandwidth() : 0;<a name="line.377"></a> +<span class="sourceLineNo">378</span> // user can set peer bandwidth to 0 to use default bandwidth<a name="line.378"></a> +<span class="sourceLineNo">379</span> return peerBandwidth != 0 ? peerBandwidth : defaultBandwidth;<a name="line.379"></a> +<span class="sourceLineNo">380</span> }<a name="line.380"></a> +<span class="sourceLineNo">381</span><a name="line.381"></a> +<span class="sourceLineNo">382</span> private void uninitialize() {<a name="line.382"></a> +<span class="sourceLineNo">383</span> LOG.debug("Source exiting " + this.peerId);<a name="line.383"></a> +<span class="sourceLineNo">384</span> metrics.clear();<a name="line.384"></a> +<span class="sourceLineNo">385</span> if (this.replicationEndpoint.isRunning() || this.replicationEndpoint.isStarting()) {<a name="line.385"></a> +<span class="sourceLineNo">386</span> this.replicationEndpoint.stop();<a name="line.386"></a> +<span class="sourceLineNo">387</span> try {<a name="line.387"></a> +<span class="sourceLineNo">388</span> this.replicationEndpoint.awaitTerminated(this.waitOnEndpointSeconds, TimeUnit.SECONDS);<a name="line.388"></a> +<span class="sourceLineNo">389</span> } catch (TimeoutException e) {<a name="line.389"></a> +<span class="sourceLineNo">390</span> LOG.warn("Failed termination after " + this.waitOnEndpointSeconds + " seconds.");<a name="line.390"></a> +<span class="sourceLineNo">391</span> }<a name="line.391"></a> +<span class="sourceLineNo">392</span> }<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> /**<a name="line.395"></a> +<span class="sourceLineNo">396</span> * Do the sleeping logic<a name="line.396"></a> +<span class="sourceLineNo">397</span> * @param msg Why we sleep<a name="line.397"></a> +<span class="sourceLineNo">398</span> * @param sleepMultiplier by how many times the default sleeping time is augmented<a name="line.398"></a> +<span class="sourceLineNo">399</span> * @return True if <code>sleepMultiplier</code> is &lt; <code>maxRetriesMultiplier</code><a name="line.399"></a> +<span class="sourceLineNo">400</span> */<a name="line.400"></a> +<span class="sourceLineNo">401</span> protected boolean sleepForRetries(String msg, int sleepMultiplier) {<a name="line.401"></a> +<span class="sourceLineNo">402</span> try {<a name="line.402"></a> +<span class="sourceLineNo">403</span> if (LOG.isTraceEnabled()) {<a name="line.403"></a> +<span class="sourceLineNo">404</span> LOG.trace(msg + ", sleeping " + sleepForRetries + " times " + sleepMultiplier);<a name="line.404"></a> +<span class="sourceLineNo">405</span> }<a name="line.405"></a> +<span class="sourceLineNo">406</span> Thread.sleep(this.sleepForRetries * sleepMultiplier);<a name="line.406"></a> +<span class="sourceLineNo">407</span> } catch (InterruptedException e) {<a name="line.407"></a> +<span class="sourceLineNo">408</span> LOG.debug("Interrupted while sleeping between retries");<a name="line.408"></a> +<span class="sourceLineNo">409</span> Thread.currentThread().interrupt();<a name="line.409"></a> +<span class="sourceLineNo">410</span> }<a name="line.410"></a> +<span class="sourceLineNo">411</span> return sleepMultiplier < maxRetriesMultiplier;<a name="line.411"></a> +<span class="sourceLineNo">412</span> }<a name="line.412"></a> +<span class="sourceLineNo">413</span><a name="line.413"></a> +<span class="sourceLineNo">414</span> /**<a name="line.414"></a> +<span class="sourceLineNo">415</span> * check whether the peer is enabled or not<a name="line.415"></a> +<span class="sourceLineNo">416</span> *<a name="line.416"></a> +<span class="sourceLineNo">417</span> * @return true if the peer is enabled, otherwise false<a name="line.417"></a> +<span class="sourceLineNo">418</span> */<a name="line.418"></a> +<span class="sourceLineNo">419</span> @Override<a name="line.419"></a> +<span class="sourceLineNo">420</span> public boolean isPeerEnabled() {<a name="line.420"></a> +<span class="sourceLineNo">421</span> return this.replicationPeers.getStatusOfPeer(this.peerId);<a name="line.421"></a> +<span class="sourceLineNo">422</span> }<a name="line.422"></a> +<span class="sourceLineNo">423</span><a name="line.423"></a> +<span class="sourceLineNo">424</span> @Override<a name="line.424"></a> +<span class="sourceLineNo">425</span> public void startup() {<a name="line.425"></a> +<span class="sourceLineNo">426</span> String n = Thread.currentThread().getName();<a name="line.426"></a> +<span class="sourceLineNo">427</span> Thread.UncaughtExceptionHandler handler = new Thread.UncaughtExceptionHandler() {<a name="line.427"></a> +<span class="sourceLineNo">428</span> @Override<a name="line.428"></a> +<span class="sourceLineNo">429</span> public void uncaughtException(final Thread t, final Throwable e) {<a name="line.429"></a> +<span class="sourceLineNo">430</span> LOG.error("Unexpected exception in ReplicationSource", e);<a name="line.430"></a> +<span class="sourceLineNo">431</span> }<a name="line.431"></a> +<span class="sourceLineNo">432</span> };<a name="line.432"></a> +<span class="sourceLineNo">433</span> Threads<a name="line.433"></a> +<span class="sourceLineNo">434</span> .setDaemonThreadRunning(this, n + ".replicationSource," + this.peerClusterZnode, handler);<a name="line.434"></a> +<span class="sourceLineNo">435</span> }<a name="line.435"></a> +<span class="sourceLineNo">436</span><a name="line.436"></a> +<span class="sourceLineNo">437</span> @Override<a name="line.437"></a> +<span class="sourceLineNo">438</span> public void terminate(String reason) {<a name="line.438"></a> +<span class="sourceLineNo">439</span> terminate(reason, null);<a name="line.439"></a> +<span class="sourceLineNo">440</span> }<a name="line.440"></a> +<span class="sourceLineNo">441</span><a name="line.441"></a> +<span class="sourceLineNo">442</span> @Override<a name="line.442"></a> +<span class="sourceLineNo">443</span> public void terminate(String reason, Exception cause) {<a name="line.443"></a> +<span class="sourceLineNo">444</s
<TRUNCATED>