You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@hbase.apache.org by mi...@apache.org on 2015/12/23 00:24:49 UTC
[26/51] [partial] hbase-site git commit: Published site at
95a13b51ee052eb73882682e8f009bfa1e914866.
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/32d40534/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/HBaseInterClusterReplicationEndpoint.html
----------------------------------------------------------------------
diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/HBaseInterClusterReplicationEndpoint.html b/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/HBaseInterClusterReplicationEndpoint.html
index e6f7370..66fe895 100644
--- a/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/HBaseInterClusterReplicationEndpoint.html
+++ b/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/HBaseInterClusterReplicationEndpoint.html
@@ -151,9 +151,9 @@
<span class="sourceLineNo">143</span> int sleepMultiplier = 1;<a name="line.143"></a>
<span class="sourceLineNo">144</span><a name="line.144"></a>
<span class="sourceLineNo">145</span> // Connect to peer cluster first, unless we have to stop<a name="line.145"></a>
-<span class="sourceLineNo">146</span> while (this.isRunning() && replicationSinkMgr.getSinks().size() == 0) {<a name="line.146"></a>
+<span class="sourceLineNo">146</span> while (this.isRunning() && replicationSinkMgr.getNumSinks() == 0) {<a name="line.146"></a>
<span class="sourceLineNo">147</span> replicationSinkMgr.chooseSinks();<a name="line.147"></a>
-<span class="sourceLineNo">148</span> if (this.isRunning() && replicationSinkMgr.getSinks().size() == 0) {<a name="line.148"></a>
+<span class="sourceLineNo">148</span> if (this.isRunning() && replicationSinkMgr.getNumSinks() == 0) {<a name="line.148"></a>
<span class="sourceLineNo">149</span> if (sleepForRetries("Waiting for peers", sleepMultiplier)) {<a name="line.149"></a>
<span class="sourceLineNo">150</span> sleepMultiplier++;<a name="line.150"></a>
<span class="sourceLineNo">151</span> }<a name="line.151"></a>
@@ -188,177 +188,192 @@
<span class="sourceLineNo">180</span> List<Entry> entries = replicateContext.getEntries();<a name="line.180"></a>
<span class="sourceLineNo">181</span> String walGroupId = replicateContext.getWalGroupId();<a name="line.181"></a>
<span class="sourceLineNo">182</span> int sleepMultiplier = 1;<a name="line.182"></a>
-<span class="sourceLineNo">183</span><a name="line.183"></a>
-<span class="sourceLineNo">184</span> if (!peersSelected && this.isRunning()) {<a name="line.184"></a>
-<span class="sourceLineNo">185</span> connectToPeers();<a name="line.185"></a>
-<span class="sourceLineNo">186</span> peersSelected = true;<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> if (replicationSinkMgr.getSinks().size() == 0) {<a name="line.189"></a>
-<span class="sourceLineNo">190</span> return false;<a name="line.190"></a>
-<span class="sourceLineNo">191</span> }<a name="line.191"></a>
-<span class="sourceLineNo">192</span> // minimum of: configured threads, number of 100-waledit batches,<a name="line.192"></a>
-<span class="sourceLineNo">193</span> // and number of current sinks<a name="line.193"></a>
-<span class="sourceLineNo">194</span> int n = Math.min(Math.min(this.maxThreads, entries.size()/100+1),<a name="line.194"></a>
-<span class="sourceLineNo">195</span> replicationSinkMgr.getSinks().size());<a name="line.195"></a>
-<span class="sourceLineNo">196</span> List<List<Entry>> entryLists = new ArrayList<List<Entry>>(n);<a name="line.196"></a>
-<span class="sourceLineNo">197</span> if (n == 1) {<a name="line.197"></a>
-<span class="sourceLineNo">198</span> entryLists.add(entries);<a name="line.198"></a>
-<span class="sourceLineNo">199</span> } else {<a name="line.199"></a>
-<span class="sourceLineNo">200</span> for (int i=0; i<n; i++) {<a name="line.200"></a>
-<span class="sourceLineNo">201</span> entryLists.add(new ArrayList<Entry>(entries.size()/n+1));<a name="line.201"></a>
-<span class="sourceLineNo">202</span> }<a name="line.202"></a>
-<span class="sourceLineNo">203</span> // now group by region<a name="line.203"></a>
-<span class="sourceLineNo">204</span> for (Entry e : entries) {<a name="line.204"></a>
-<span class="sourceLineNo">205</span> entryLists.get(Math.abs(Bytes.hashCode(e.getKey().getEncodedRegionName())%n)).add(e);<a name="line.205"></a>
-<span class="sourceLineNo">206</span> }<a name="line.206"></a>
-<span class="sourceLineNo">207</span> }<a name="line.207"></a>
-<span class="sourceLineNo">208</span> while (this.isRunning()) {<a name="line.208"></a>
-<span class="sourceLineNo">209</span> if (!isPeerEnabled()) {<a name="line.209"></a>
-<span class="sourceLineNo">210</span> if (sleepForRetries("Replication is disabled", sleepMultiplier)) {<a name="line.210"></a>
-<span class="sourceLineNo">211</span> sleepMultiplier++;<a name="line.211"></a>
-<span class="sourceLineNo">212</span> }<a name="line.212"></a>
-<span class="sourceLineNo">213</span> continue;<a name="line.213"></a>
-<span class="sourceLineNo">214</span> }<a name="line.214"></a>
-<span class="sourceLineNo">215</span> try {<a name="line.215"></a>
-<span class="sourceLineNo">216</span> if (LOG.isTraceEnabled()) {<a name="line.216"></a>
-<span class="sourceLineNo">217</span> LOG.trace("Replicating " + entries.size() +<a name="line.217"></a>
-<span class="sourceLineNo">218</span> " entries of total size " + replicateContext.getSize());<a name="line.218"></a>
-<span class="sourceLineNo">219</span> }<a name="line.219"></a>
-<span class="sourceLineNo">220</span><a name="line.220"></a>
-<span class="sourceLineNo">221</span> int futures = 0;<a name="line.221"></a>
-<span class="sourceLineNo">222</span> for (int i=0; i<entryLists.size(); i++) {<a name="line.222"></a>
-<span class="sourceLineNo">223</span> if (!entryLists.get(i).isEmpty()) {<a name="line.223"></a>
-<span class="sourceLineNo">224</span> if (LOG.isTraceEnabled()) {<a name="line.224"></a>
-<span class="sourceLineNo">225</span> LOG.trace("Submitting " + entryLists.get(i).size() +<a name="line.225"></a>
-<span class="sourceLineNo">226</span> " entries of total size " + replicateContext.getSize());<a name="line.226"></a>
-<span class="sourceLineNo">227</span> }<a name="line.227"></a>
-<span class="sourceLineNo">228</span> // RuntimeExceptions encountered here bubble up and are handled in ReplicationSource<a name="line.228"></a>
-<span class="sourceLineNo">229</span> pool.submit(createReplicator(entryLists.get(i), i));<a name="line.229"></a>
-<span class="sourceLineNo">230</span> futures++;<a name="line.230"></a>
-<span class="sourceLineNo">231</span> }<a name="line.231"></a>
-<span class="sourceLineNo">232</span> }<a name="line.232"></a>
-<span class="sourceLineNo">233</span> IOException iox = null;<a name="line.233"></a>
-<span class="sourceLineNo">234</span><a name="line.234"></a>
-<span class="sourceLineNo">235</span> for (int i=0; i<futures; i++) {<a name="line.235"></a>
-<span class="sourceLineNo">236</span> try {<a name="line.236"></a>
-<span class="sourceLineNo">237</span> // wait for all futures, remove successful parts<a name="line.237"></a>
-<span class="sourceLineNo">238</span> // (only the remaining parts will be retried)<a name="line.238"></a>
-<span class="sourceLineNo">239</span> Future<Integer> f = pool.take();<a name="line.239"></a>
-<span class="sourceLineNo">240</span> entryLists.set(f.get().intValue(), Collections.<Entry>emptyList());<a name="line.240"></a>
-<span class="sourceLineNo">241</span> } catch (InterruptedException ie) {<a name="line.241"></a>
-<span class="sourceLineNo">242</span> iox = new IOException(ie);<a name="line.242"></a>
-<span class="sourceLineNo">243</span> } catch (ExecutionException ee) {<a name="line.243"></a>
-<span class="sourceLineNo">244</span> // cause must be an IOException<a name="line.244"></a>
-<span class="sourceLineNo">245</span> iox = (IOException)ee.getCause();<a name="line.245"></a>
-<span class="sourceLineNo">246</span> }<a name="line.246"></a>
-<span class="sourceLineNo">247</span> }<a name="line.247"></a>
-<span class="sourceLineNo">248</span> if (iox != null) {<a name="line.248"></a>
-<span class="sourceLineNo">249</span> // if we had any exceptions, try again<a name="line.249"></a>
-<span class="sourceLineNo">250</span> throw iox;<a name="line.250"></a>
-<span class="sourceLineNo">251</span> }<a name="line.251"></a>
-<span class="sourceLineNo">252</span> // update metrics<a name="line.252"></a>
-<span class="sourceLineNo">253</span> this.metrics.setAgeOfLastShippedOp(entries.get(entries.size() - 1).getKey().getWriteTime(),<a name="line.253"></a>
-<span class="sourceLineNo">254</span> walGroupId);<a name="line.254"></a>
-<span class="sourceLineNo">255</span> return true;<a name="line.255"></a>
-<span class="sourceLineNo">256</span><a name="line.256"></a>
-<span class="sourceLineNo">257</span> } catch (IOException ioe) {<a name="line.257"></a>
-<span class="sourceLineNo">258</span> // Didn't ship anything, but must still age the last time we did<a name="line.258"></a>
-<span class="sourceLineNo">259</span> this.metrics.refreshAgeOfLastShippedOp(walGroupId);<a name="line.259"></a>
-<span class="sourceLineNo">260</span> if (ioe instanceof RemoteException) {<a name="line.260"></a>
-<span class="sourceLineNo">261</span> ioe = ((RemoteException) ioe).unwrapRemoteException();<a name="line.261"></a>
-<span class="sourceLineNo">262</span> LOG.warn("Can't replicate because of an error on the remote cluster: ", ioe);<a name="line.262"></a>
-<span class="sourceLineNo">263</span> if (ioe instanceof TableNotFoundException) {<a name="line.263"></a>
-<span class="sourceLineNo">264</span> if (sleepForRetries("A table is missing in the peer cluster. "<a name="line.264"></a>
-<span class="sourceLineNo">265</span> + "Replication cannot proceed without losing data.", sleepMultiplier)) {<a name="line.265"></a>
-<span class="sourceLineNo">266</span> sleepMultiplier++;<a name="line.266"></a>
-<span class="sourceLineNo">267</span> }<a name="line.267"></a>
-<span class="sourceLineNo">268</span> }<a name="line.268"></a>
-<span class="sourceLineNo">269</span> } else {<a name="line.269"></a>
-<span class="sourceLineNo">270</span> if (ioe instanceof SocketTimeoutException) {<a name="line.270"></a>
-<span class="sourceLineNo">271</span> // This exception means we waited for more than 60s and nothing<a name="line.271"></a>
-<span class="sourceLineNo">272</span> // happened, the cluster is alive and calling it right away<a name="line.272"></a>
-<span class="sourceLineNo">273</span> // even for a test just makes things worse.<a name="line.273"></a>
-<span class="sourceLineNo">274</span> sleepForRetries("Encountered a SocketTimeoutException. Since the " +<a name="line.274"></a>
-<span class="sourceLineNo">275</span> "call to the remote cluster timed out, which is usually " +<a name="line.275"></a>
-<span class="sourceLineNo">276</span> "caused by a machine failure or a massive slowdown",<a name="line.276"></a>
-<span class="sourceLineNo">277</span> this.socketTimeoutMultiplier);<a name="line.277"></a>
-<span class="sourceLineNo">278</span> } else if (ioe instanceof ConnectException) {<a name="line.278"></a>
-<span class="sourceLineNo">279</span> LOG.warn("Peer is unavailable, rechecking all sinks: ", ioe);<a name="line.279"></a>
-<span class="sourceLineNo">280</span> replicationSinkMgr.chooseSinks();<a name="line.280"></a>
-<span class="sourceLineNo">281</span> } else {<a name="line.281"></a>
-<span class="sourceLineNo">282</span> LOG.warn("Can't replicate because of a local or network error: ", ioe);<a name="line.282"></a>
+<span class="sourceLineNo">183</span> int numReplicated = 0;<a name="line.183"></a>
+<span class="sourceLineNo">184</span><a name="line.184"></a>
+<span class="sourceLineNo">185</span> if (!peersSelected && this.isRunning()) {<a name="line.185"></a>
+<span class="sourceLineNo">186</span> connectToPeers();<a name="line.186"></a>
+<span class="sourceLineNo">187</span> peersSelected = true;<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> int numSinks = replicationSinkMgr.getNumSinks();<a name="line.190"></a>
+<span class="sourceLineNo">191</span> if (numSinks == 0) {<a name="line.191"></a>
+<span class="sourceLineNo">192</span> LOG.warn("No replication sinks found, returning without replicating. The source should retry"<a name="line.192"></a>
+<span class="sourceLineNo">193</span> + " with the same set of edits.");<a name="line.193"></a>
+<span class="sourceLineNo">194</span> return false;<a name="line.194"></a>
+<span class="sourceLineNo">195</span> }<a name="line.195"></a>
+<span class="sourceLineNo">196</span><a name="line.196"></a>
+<span class="sourceLineNo">197</span> // minimum of: configured threads, number of 100-waledit batches,<a name="line.197"></a>
+<span class="sourceLineNo">198</span> // and number of current sinks<a name="line.198"></a>
+<span class="sourceLineNo">199</span> int n = Math.min(Math.min(this.maxThreads, entries.size()/100+1), numSinks);<a name="line.199"></a>
+<span class="sourceLineNo">200</span><a name="line.200"></a>
+<span class="sourceLineNo">201</span> List<List<Entry>> entryLists = new ArrayList<List<Entry>>(n);<a name="line.201"></a>
+<span class="sourceLineNo">202</span> if (n == 1) {<a name="line.202"></a>
+<span class="sourceLineNo">203</span> entryLists.add(entries);<a name="line.203"></a>
+<span class="sourceLineNo">204</span> } else {<a name="line.204"></a>
+<span class="sourceLineNo">205</span> for (int i=0; i<n; i++) {<a name="line.205"></a>
+<span class="sourceLineNo">206</span> entryLists.add(new ArrayList<Entry>(entries.size()/n+1));<a name="line.206"></a>
+<span class="sourceLineNo">207</span> }<a name="line.207"></a>
+<span class="sourceLineNo">208</span> // now group by region<a name="line.208"></a>
+<span class="sourceLineNo">209</span> for (Entry e : entries) {<a name="line.209"></a>
+<span class="sourceLineNo">210</span> entryLists.get(Math.abs(Bytes.hashCode(e.getKey().getEncodedRegionName())%n)).add(e);<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> while (this.isRunning()) {<a name="line.213"></a>
+<span class="sourceLineNo">214</span> if (!isPeerEnabled()) {<a name="line.214"></a>
+<span class="sourceLineNo">215</span> if (sleepForRetries("Replication is disabled", sleepMultiplier)) {<a name="line.215"></a>
+<span class="sourceLineNo">216</span> sleepMultiplier++;<a name="line.216"></a>
+<span class="sourceLineNo">217</span> }<a name="line.217"></a>
+<span class="sourceLineNo">218</span> continue;<a name="line.218"></a>
+<span class="sourceLineNo">219</span> }<a name="line.219"></a>
+<span class="sourceLineNo">220</span> try {<a name="line.220"></a>
+<span class="sourceLineNo">221</span> if (LOG.isTraceEnabled()) {<a name="line.221"></a>
+<span class="sourceLineNo">222</span> LOG.trace("Replicating " + entries.size() +<a name="line.222"></a>
+<span class="sourceLineNo">223</span> " entries of total size " + replicateContext.getSize());<a name="line.223"></a>
+<span class="sourceLineNo">224</span> }<a name="line.224"></a>
+<span class="sourceLineNo">225</span><a name="line.225"></a>
+<span class="sourceLineNo">226</span> int futures = 0;<a name="line.226"></a>
+<span class="sourceLineNo">227</span> for (int i=0; i<entryLists.size(); i++) {<a name="line.227"></a>
+<span class="sourceLineNo">228</span> if (!entryLists.get(i).isEmpty()) {<a name="line.228"></a>
+<span class="sourceLineNo">229</span> if (LOG.isTraceEnabled()) {<a name="line.229"></a>
+<span class="sourceLineNo">230</span> LOG.trace("Submitting " + entryLists.get(i).size() +<a name="line.230"></a>
+<span class="sourceLineNo">231</span> " entries of total size " + replicateContext.getSize());<a name="line.231"></a>
+<span class="sourceLineNo">232</span> }<a name="line.232"></a>
+<span class="sourceLineNo">233</span> // RuntimeExceptions encountered here bubble up and are handled in ReplicationSource<a name="line.233"></a>
+<span class="sourceLineNo">234</span> pool.submit(createReplicator(entryLists.get(i), i));<a name="line.234"></a>
+<span class="sourceLineNo">235</span> futures++;<a name="line.235"></a>
+<span class="sourceLineNo">236</span> }<a name="line.236"></a>
+<span class="sourceLineNo">237</span> }<a name="line.237"></a>
+<span class="sourceLineNo">238</span> IOException iox = null;<a name="line.238"></a>
+<span class="sourceLineNo">239</span><a name="line.239"></a>
+<span class="sourceLineNo">240</span> for (int i=0; i<futures; i++) {<a name="line.240"></a>
+<span class="sourceLineNo">241</span> try {<a name="line.241"></a>
+<span class="sourceLineNo">242</span> // wait for all futures, remove successful parts<a name="line.242"></a>
+<span class="sourceLineNo">243</span> // (only the remaining parts will be retried)<a name="line.243"></a>
+<span class="sourceLineNo">244</span> Future<Integer> f = pool.take();<a name="line.244"></a>
+<span class="sourceLineNo">245</span> int index = f.get().intValue();<a name="line.245"></a>
+<span class="sourceLineNo">246</span> int batchSize = entryLists.get(index).size();<a name="line.246"></a>
+<span class="sourceLineNo">247</span> entryLists.set(index, Collections.<Entry>emptyList());<a name="line.247"></a>
+<span class="sourceLineNo">248</span> // Now, we have marked the batch as done replicating, record its size<a name="line.248"></a>
+<span class="sourceLineNo">249</span> numReplicated += batchSize;<a name="line.249"></a>
+<span class="sourceLineNo">250</span> } catch (InterruptedException ie) {<a name="line.250"></a>
+<span class="sourceLineNo">251</span> iox = new IOException(ie);<a name="line.251"></a>
+<span class="sourceLineNo">252</span> } catch (ExecutionException ee) {<a name="line.252"></a>
+<span class="sourceLineNo">253</span> // cause must be an IOException<a name="line.253"></a>
+<span class="sourceLineNo">254</span> iox = (IOException)ee.getCause();<a name="line.254"></a>
+<span class="sourceLineNo">255</span> }<a name="line.255"></a>
+<span class="sourceLineNo">256</span> }<a name="line.256"></a>
+<span class="sourceLineNo">257</span> if (iox != null) {<a name="line.257"></a>
+<span class="sourceLineNo">258</span> // if we had any exceptions, try again<a name="line.258"></a>
+<span class="sourceLineNo">259</span> throw iox;<a name="line.259"></a>
+<span class="sourceLineNo">260</span> }<a name="line.260"></a>
+<span class="sourceLineNo">261</span> if (numReplicated != entries.size()) {<a name="line.261"></a>
+<span class="sourceLineNo">262</span> // Something went wrong here and we don't know what, let's just fail and retry.<a name="line.262"></a>
+<span class="sourceLineNo">263</span> LOG.warn("The number of edits replicated is different from the number received,"<a name="line.263"></a>
+<span class="sourceLineNo">264</span> + " failing for now.");<a name="line.264"></a>
+<span class="sourceLineNo">265</span> return false;<a name="line.265"></a>
+<span class="sourceLineNo">266</span> }<a name="line.266"></a>
+<span class="sourceLineNo">267</span> // update metrics<a name="line.267"></a>
+<span class="sourceLineNo">268</span> this.metrics.setAgeOfLastShippedOp(entries.get(entries.size() - 1).getKey().getWriteTime(),<a name="line.268"></a>
+<span class="sourceLineNo">269</span> walGroupId);<a name="line.269"></a>
+<span class="sourceLineNo">270</span> return true;<a name="line.270"></a>
+<span class="sourceLineNo">271</span><a name="line.271"></a>
+<span class="sourceLineNo">272</span> } catch (IOException ioe) {<a name="line.272"></a>
+<span class="sourceLineNo">273</span> // Didn't ship anything, but must still age the last time we did<a name="line.273"></a>
+<span class="sourceLineNo">274</span> this.metrics.refreshAgeOfLastShippedOp(walGroupId);<a name="line.274"></a>
+<span class="sourceLineNo">275</span> if (ioe instanceof RemoteException) {<a name="line.275"></a>
+<span class="sourceLineNo">276</span> ioe = ((RemoteException) ioe).unwrapRemoteException();<a name="line.276"></a>
+<span class="sourceLineNo">277</span> LOG.warn("Can't replicate because of an error on the remote cluster: ", ioe);<a name="line.277"></a>
+<span class="sourceLineNo">278</span> if (ioe instanceof TableNotFoundException) {<a name="line.278"></a>
+<span class="sourceLineNo">279</span> if (sleepForRetries("A table is missing in the peer cluster. "<a name="line.279"></a>
+<span class="sourceLineNo">280</span> + "Replication cannot proceed without losing data.", sleepMultiplier)) {<a name="line.280"></a>
+<span class="sourceLineNo">281</span> sleepMultiplier++;<a name="line.281"></a>
+<span class="sourceLineNo">282</span> }<a name="line.282"></a>
<span class="sourceLineNo">283</span> }<a name="line.283"></a>
-<span class="sourceLineNo">284</span> }<a name="line.284"></a>
-<span class="sourceLineNo">285</span> if (sleepForRetries("Since we are unable to replicate", sleepMultiplier)) {<a name="line.285"></a>
-<span class="sourceLineNo">286</span> sleepMultiplier++;<a name="line.286"></a>
-<span class="sourceLineNo">287</span> }<a name="line.287"></a>
-<span class="sourceLineNo">288</span> }<a name="line.288"></a>
-<span class="sourceLineNo">289</span> }<a name="line.289"></a>
-<span class="sourceLineNo">290</span> return false; // in case we exited before replicating<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> protected boolean isPeerEnabled() {<a name="line.293"></a>
-<span class="sourceLineNo">294</span> return ctx.getReplicationPeer().getPeerState() == PeerState.ENABLED;<a name="line.294"></a>
-<span class="sourceLineNo">295</span> }<a name="line.295"></a>
-<span class="sourceLineNo">296</span><a name="line.296"></a>
-<span class="sourceLineNo">297</span> @Override<a name="line.297"></a>
-<span class="sourceLineNo">298</span> protected void doStop() {<a name="line.298"></a>
-<span class="sourceLineNo">299</span> disconnect(); //don't call super.doStop()<a name="line.299"></a>
-<span class="sourceLineNo">300</span> if (this.conn != null) {<a name="line.300"></a>
-<span class="sourceLineNo">301</span> try {<a name="line.301"></a>
-<span class="sourceLineNo">302</span> this.conn.close();<a name="line.302"></a>
-<span class="sourceLineNo">303</span> this.conn = null;<a name="line.303"></a>
-<span class="sourceLineNo">304</span> } catch (IOException e) {<a name="line.304"></a>
-<span class="sourceLineNo">305</span> LOG.warn("Failed to close the connection");<a name="line.305"></a>
-<span class="sourceLineNo">306</span> }<a name="line.306"></a>
-<span class="sourceLineNo">307</span> }<a name="line.307"></a>
-<span class="sourceLineNo">308</span> exec.shutdownNow();<a name="line.308"></a>
-<span class="sourceLineNo">309</span> notifyStopped();<a name="line.309"></a>
+<span class="sourceLineNo">284</span> } else {<a name="line.284"></a>
+<span class="sourceLineNo">285</span> if (ioe instanceof SocketTimeoutException) {<a name="line.285"></a>
+<span class="sourceLineNo">286</span> // This exception means we waited for more than 60s and nothing<a name="line.286"></a>
+<span class="sourceLineNo">287</span> // happened, the cluster is alive and calling it right away<a name="line.287"></a>
+<span class="sourceLineNo">288</span> // even for a test just makes things worse.<a name="line.288"></a>
+<span class="sourceLineNo">289</span> sleepForRetries("Encountered a SocketTimeoutException. Since the " +<a name="line.289"></a>
+<span class="sourceLineNo">290</span> "call to the remote cluster timed out, which is usually " +<a name="line.290"></a>
+<span class="sourceLineNo">291</span> "caused by a machine failure or a massive slowdown",<a name="line.291"></a>
+<span class="sourceLineNo">292</span> this.socketTimeoutMultiplier);<a name="line.292"></a>
+<span class="sourceLineNo">293</span> } else if (ioe instanceof ConnectException) {<a name="line.293"></a>
+<span class="sourceLineNo">294</span> LOG.warn("Peer is unavailable, rechecking all sinks: ", ioe);<a name="line.294"></a>
+<span class="sourceLineNo">295</span> replicationSinkMgr.chooseSinks();<a name="line.295"></a>
+<span class="sourceLineNo">296</span> } else {<a name="line.296"></a>
+<span class="sourceLineNo">297</span> LOG.warn("Can't replicate because of a local or network error: ", ioe);<a name="line.297"></a>
+<span class="sourceLineNo">298</span> }<a name="line.298"></a>
+<span class="sourceLineNo">299</span> }<a name="line.299"></a>
+<span class="sourceLineNo">300</span> if (sleepForRetries("Since we are unable to replicate", sleepMultiplier)) {<a name="line.300"></a>
+<span class="sourceLineNo">301</span> sleepMultiplier++;<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> }<a name="line.304"></a>
+<span class="sourceLineNo">305</span> return false; // in case we exited before replicating<a name="line.305"></a>
+<span class="sourceLineNo">306</span> }<a name="line.306"></a>
+<span class="sourceLineNo">307</span><a name="line.307"></a>
+<span class="sourceLineNo">308</span> protected boolean isPeerEnabled() {<a name="line.308"></a>
+<span class="sourceLineNo">309</span> return ctx.getReplicationPeer().getPeerState() == PeerState.ENABLED;<a name="line.309"></a>
<span class="sourceLineNo">310</span> }<a name="line.310"></a>
<span class="sourceLineNo">311</span><a name="line.311"></a>
-<span class="sourceLineNo">312</span> // is this needed? Nobody else will call doStop() otherwise<a name="line.312"></a>
-<span class="sourceLineNo">313</span> @Override<a name="line.313"></a>
-<span class="sourceLineNo">314</span> public State stopAndWait() {<a name="line.314"></a>
-<span class="sourceLineNo">315</span> doStop();<a name="line.315"></a>
-<span class="sourceLineNo">316</span> return super.stopAndWait();<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> @VisibleForTesting<a name="line.319"></a>
-<span class="sourceLineNo">320</span> protected Replicator createReplicator(List<Entry> entries, int ordinal) {<a name="line.320"></a>
-<span class="sourceLineNo">321</span> return new Replicator(entries, ordinal);<a name="line.321"></a>
-<span class="sourceLineNo">322</span> }<a name="line.322"></a>
-<span class="sourceLineNo">323</span><a name="line.323"></a>
-<span class="sourceLineNo">324</span> @VisibleForTesting<a name="line.324"></a>
-<span class="sourceLineNo">325</span> protected class Replicator implements Callable<Integer> {<a name="line.325"></a>
-<span class="sourceLineNo">326</span> private List<Entry> entries;<a name="line.326"></a>
-<span class="sourceLineNo">327</span> private int ordinal;<a name="line.327"></a>
-<span class="sourceLineNo">328</span> public Replicator(List<Entry> entries, int ordinal) {<a name="line.328"></a>
-<span class="sourceLineNo">329</span> this.entries = entries;<a name="line.329"></a>
-<span class="sourceLineNo">330</span> this.ordinal = ordinal;<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> @Override<a name="line.333"></a>
-<span class="sourceLineNo">334</span> public Integer call() throws IOException {<a name="line.334"></a>
-<span class="sourceLineNo">335</span> SinkPeer sinkPeer = null;<a name="line.335"></a>
-<span class="sourceLineNo">336</span> try {<a name="line.336"></a>
-<span class="sourceLineNo">337</span> sinkPeer = replicationSinkMgr.getReplicationSink();<a name="line.337"></a>
-<span class="sourceLineNo">338</span> BlockingInterface rrs = sinkPeer.getRegionServer();<a name="line.338"></a>
-<span class="sourceLineNo">339</span> ReplicationProtbufUtil.replicateWALEntry(rrs, entries.toArray(new Entry[entries.size()]),<a name="line.339"></a>
-<span class="sourceLineNo">340</span> replicationClusterId, baseNamespaceDir, hfileArchiveDir);<a name="line.340"></a>
-<span class="sourceLineNo">341</span> replicationSinkMgr.reportSinkSuccess(sinkPeer);<a name="line.341"></a>
-<span class="sourceLineNo">342</span> return ordinal;<a name="line.342"></a>
-<span class="sourceLineNo">343</span><a name="line.343"></a>
-<span class="sourceLineNo">344</span> } catch (IOException ioe) {<a name="line.344"></a>
-<span class="sourceLineNo">345</span> if (sinkPeer != null) {<a name="line.345"></a>
-<span class="sourceLineNo">346</span> replicationSinkMgr.reportBadSink(sinkPeer);<a name="line.346"></a>
-<span class="sourceLineNo">347</span> }<a name="line.347"></a>
-<span class="sourceLineNo">348</span> throw ioe;<a name="line.348"></a>
-<span class="sourceLineNo">349</span> }<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> }<a name="line.352"></a>
-<span class="sourceLineNo">353</span>}<a name="line.353"></a>
+<span class="sourceLineNo">312</span> @Override<a name="line.312"></a>
+<span class="sourceLineNo">313</span> protected void doStop() {<a name="line.313"></a>
+<span class="sourceLineNo">314</span> disconnect(); //don't call super.doStop()<a name="line.314"></a>
+<span class="sourceLineNo">315</span> if (this.conn != null) {<a name="line.315"></a>
+<span class="sourceLineNo">316</span> try {<a name="line.316"></a>
+<span class="sourceLineNo">317</span> this.conn.close();<a name="line.317"></a>
+<span class="sourceLineNo">318</span> this.conn = null;<a name="line.318"></a>
+<span class="sourceLineNo">319</span> } catch (IOException e) {<a name="line.319"></a>
+<span class="sourceLineNo">320</span> LOG.warn("Failed to close the connection");<a name="line.320"></a>
+<span class="sourceLineNo">321</span> }<a name="line.321"></a>
+<span class="sourceLineNo">322</span> }<a name="line.322"></a>
+<span class="sourceLineNo">323</span> exec.shutdownNow();<a name="line.323"></a>
+<span class="sourceLineNo">324</span> notifyStopped();<a name="line.324"></a>
+<span class="sourceLineNo">325</span> }<a name="line.325"></a>
+<span class="sourceLineNo">326</span><a name="line.326"></a>
+<span class="sourceLineNo">327</span> // is this needed? Nobody else will call doStop() otherwise<a name="line.327"></a>
+<span class="sourceLineNo">328</span> @Override<a name="line.328"></a>
+<span class="sourceLineNo">329</span> public State stopAndWait() {<a name="line.329"></a>
+<span class="sourceLineNo">330</span> doStop();<a name="line.330"></a>
+<span class="sourceLineNo">331</span> return super.stopAndWait();<a name="line.331"></a>
+<span class="sourceLineNo">332</span> }<a name="line.332"></a>
+<span class="sourceLineNo">333</span><a name="line.333"></a>
+<span class="sourceLineNo">334</span> @VisibleForTesting<a name="line.334"></a>
+<span class="sourceLineNo">335</span> protected Replicator createReplicator(List<Entry> entries, int ordinal) {<a name="line.335"></a>
+<span class="sourceLineNo">336</span> return new Replicator(entries, ordinal);<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> @VisibleForTesting<a name="line.339"></a>
+<span class="sourceLineNo">340</span> protected class Replicator implements Callable<Integer> {<a name="line.340"></a>
+<span class="sourceLineNo">341</span> private List<Entry> entries;<a name="line.341"></a>
+<span class="sourceLineNo">342</span> private int ordinal;<a name="line.342"></a>
+<span class="sourceLineNo">343</span> public Replicator(List<Entry> entries, int ordinal) {<a name="line.343"></a>
+<span class="sourceLineNo">344</span> this.entries = entries;<a name="line.344"></a>
+<span class="sourceLineNo">345</span> this.ordinal = ordinal;<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 Integer call() throws IOException {<a name="line.349"></a>
+<span class="sourceLineNo">350</span> SinkPeer sinkPeer = null;<a name="line.350"></a>
+<span class="sourceLineNo">351</span> try {<a name="line.351"></a>
+<span class="sourceLineNo">352</span> sinkPeer = replicationSinkMgr.getReplicationSink();<a name="line.352"></a>
+<span class="sourceLineNo">353</span> BlockingInterface rrs = sinkPeer.getRegionServer();<a name="line.353"></a>
+<span class="sourceLineNo">354</span> ReplicationProtbufUtil.replicateWALEntry(rrs, entries.toArray(new Entry[entries.size()]),<a name="line.354"></a>
+<span class="sourceLineNo">355</span> replicationClusterId, baseNamespaceDir, hfileArchiveDir);<a name="line.355"></a>
+<span class="sourceLineNo">356</span> replicationSinkMgr.reportSinkSuccess(sinkPeer);<a name="line.356"></a>
+<span class="sourceLineNo">357</span> return ordinal;<a name="line.357"></a>
+<span class="sourceLineNo">358</span><a name="line.358"></a>
+<span class="sourceLineNo">359</span> } catch (IOException ioe) {<a name="line.359"></a>
+<span class="sourceLineNo">360</span> if (sinkPeer != null) {<a name="line.360"></a>
+<span class="sourceLineNo">361</span> replicationSinkMgr.reportBadSink(sinkPeer);<a name="line.361"></a>
+<span class="sourceLineNo">362</span> }<a name="line.362"></a>
+<span class="sourceLineNo">363</span> throw ioe;<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><a name="line.366"></a>
+<span class="sourceLineNo">367</span> }<a name="line.367"></a>
+<span class="sourceLineNo">368</span>}<a name="line.368"></a>
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/32d40534/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSinkManager.SinkPeer.html
----------------------------------------------------------------------
diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSinkManager.SinkPeer.html b/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSinkManager.SinkPeer.html
index 3e8f4ec..00bdffe 100644
--- a/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSinkManager.SinkPeer.html
+++ b/devapidocs/src-html/org/apache/hadoop/hbase/replication/regionserver/ReplicationSinkManager.SinkPeer.html
@@ -31,170 +31,179 @@
<span class="sourceLineNo">023</span>import java.util.Map;<a name="line.23"></a>
<span class="sourceLineNo">024</span>import java.util.Random;<a name="line.24"></a>
<span class="sourceLineNo">025</span><a name="line.25"></a>
-<span class="sourceLineNo">026</span>import org.apache.commons.logging.Log;<a name="line.26"></a>
-<span class="sourceLineNo">027</span>import org.apache.commons.logging.LogFactory;<a name="line.27"></a>
-<span class="sourceLineNo">028</span>import org.apache.hadoop.conf.Configuration;<a name="line.28"></a>
-<span class="sourceLineNo">029</span>import org.apache.hadoop.hbase.ServerName;<a name="line.29"></a>
-<span class="sourceLineNo">030</span>import org.apache.hadoop.hbase.client.HConnection;<a name="line.30"></a>
-<span class="sourceLineNo">031</span>import org.apache.hadoop.hbase.protobuf.generated.AdminProtos.AdminService;<a name="line.31"></a>
-<span class="sourceLineNo">032</span>import org.apache.hadoop.hbase.replication.HBaseReplicationEndpoint;<a name="line.32"></a>
-<span class="sourceLineNo">033</span>import com.google.common.collect.Lists;<a name="line.33"></a>
-<span class="sourceLineNo">034</span>import com.google.common.collect.Maps;<a name="line.34"></a>
-<span class="sourceLineNo">035</span><a name="line.35"></a>
-<span class="sourceLineNo">036</span>/**<a name="line.36"></a>
-<span class="sourceLineNo">037</span> * Maintains a collection of peers to replicate to, and randomly selects a<a name="line.37"></a>
-<span class="sourceLineNo">038</span> * single peer to replicate to per set of data to replicate. Also handles<a name="line.38"></a>
-<span class="sourceLineNo">039</span> * keeping track of peer availability.<a name="line.39"></a>
-<span class="sourceLineNo">040</span> */<a name="line.40"></a>
-<span class="sourceLineNo">041</span>public class ReplicationSinkManager {<a name="line.41"></a>
-<span class="sourceLineNo">042</span><a name="line.42"></a>
-<span class="sourceLineNo">043</span> private static final Log LOG = LogFactory.getLog(ReplicationSinkManager.class);<a name="line.43"></a>
-<span class="sourceLineNo">044</span><a name="line.44"></a>
-<span class="sourceLineNo">045</span> /**<a name="line.45"></a>
-<span class="sourceLineNo">046</span> * Default maximum number of times a replication sink can be reported as bad before<a name="line.46"></a>
-<span class="sourceLineNo">047</span> * it will no longer be provided as a sink for replication without the pool of<a name="line.47"></a>
-<span class="sourceLineNo">048</span> * replication sinks being refreshed.<a name="line.48"></a>
-<span class="sourceLineNo">049</span> */<a name="line.49"></a>
-<span class="sourceLineNo">050</span> static final int DEFAULT_BAD_SINK_THRESHOLD = 3;<a name="line.50"></a>
-<span class="sourceLineNo">051</span><a name="line.51"></a>
-<span class="sourceLineNo">052</span> /**<a name="line.52"></a>
-<span class="sourceLineNo">053</span> * Default ratio of the total number of peer cluster region servers to consider<a name="line.53"></a>
-<span class="sourceLineNo">054</span> * replicating to.<a name="line.54"></a>
-<span class="sourceLineNo">055</span> */<a name="line.55"></a>
-<span class="sourceLineNo">056</span> static final float DEFAULT_REPLICATION_SOURCE_RATIO = 0.1f;<a name="line.56"></a>
-<span class="sourceLineNo">057</span><a name="line.57"></a>
+<span class="sourceLineNo">026</span>import com.google.common.annotations.VisibleForTesting;<a name="line.26"></a>
+<span class="sourceLineNo">027</span>import org.apache.commons.logging.Log;<a name="line.27"></a>
+<span class="sourceLineNo">028</span>import org.apache.commons.logging.LogFactory;<a name="line.28"></a>
+<span class="sourceLineNo">029</span>import org.apache.hadoop.conf.Configuration;<a name="line.29"></a>
+<span class="sourceLineNo">030</span>import org.apache.hadoop.hbase.ServerName;<a name="line.30"></a>
+<span class="sourceLineNo">031</span>import org.apache.hadoop.hbase.client.HConnection;<a name="line.31"></a>
+<span class="sourceLineNo">032</span>import org.apache.hadoop.hbase.protobuf.generated.AdminProtos.AdminService;<a name="line.32"></a>
+<span class="sourceLineNo">033</span>import org.apache.hadoop.hbase.replication.HBaseReplicationEndpoint;<a name="line.33"></a>
+<span class="sourceLineNo">034</span>import com.google.common.collect.Lists;<a name="line.34"></a>
+<span class="sourceLineNo">035</span>import com.google.common.collect.Maps;<a name="line.35"></a>
+<span class="sourceLineNo">036</span><a name="line.36"></a>
+<span class="sourceLineNo">037</span>/**<a name="line.37"></a>
+<span class="sourceLineNo">038</span> * Maintains a collection of peers to replicate to, and randomly selects a<a name="line.38"></a>
+<span class="sourceLineNo">039</span> * single peer to replicate to per set of data to replicate. Also handles<a name="line.39"></a>
+<span class="sourceLineNo">040</span> * keeping track of peer availability.<a name="line.40"></a>
+<span class="sourceLineNo">041</span> */<a name="line.41"></a>
+<span class="sourceLineNo">042</span>public class ReplicationSinkManager {<a name="line.42"></a>
+<span class="sourceLineNo">043</span><a name="line.43"></a>
+<span class="sourceLineNo">044</span> private static final Log LOG = LogFactory.getLog(ReplicationSinkManager.class);<a name="line.44"></a>
+<span class="sourceLineNo">045</span><a name="line.45"></a>
+<span class="sourceLineNo">046</span> /**<a name="line.46"></a>
+<span class="sourceLineNo">047</span> * Default maximum number of times a replication sink can be reported as bad before<a name="line.47"></a>
+<span class="sourceLineNo">048</span> * it will no longer be provided as a sink for replication without the pool of<a name="line.48"></a>
+<span class="sourceLineNo">049</span> * replication sinks being refreshed.<a name="line.49"></a>
+<span class="sourceLineNo">050</span> */<a name="line.50"></a>
+<span class="sourceLineNo">051</span> static final int DEFAULT_BAD_SINK_THRESHOLD = 3;<a name="line.51"></a>
+<span class="sourceLineNo">052</span><a name="line.52"></a>
+<span class="sourceLineNo">053</span> /**<a name="line.53"></a>
+<span class="sourceLineNo">054</span> * Default ratio of the total number of peer cluster region servers to consider<a name="line.54"></a>
+<span class="sourceLineNo">055</span> * replicating to.<a name="line.55"></a>
+<span class="sourceLineNo">056</span> */<a name="line.56"></a>
+<span class="sourceLineNo">057</span> static final float DEFAULT_REPLICATION_SOURCE_RATIO = 0.1f;<a name="line.57"></a>
<span class="sourceLineNo">058</span><a name="line.58"></a>
-<span class="sourceLineNo">059</span> private final HConnection conn;<a name="line.59"></a>
-<span class="sourceLineNo">060</span><a name="line.60"></a>
-<span class="sourceLineNo">061</span> private final String peerClusterId;<a name="line.61"></a>
-<span class="sourceLineNo">062</span><a name="line.62"></a>
-<span class="sourceLineNo">063</span> private final HBaseReplicationEndpoint endpoint;<a name="line.63"></a>
-<span class="sourceLineNo">064</span><a name="line.64"></a>
-<span class="sourceLineNo">065</span> // Count of "bad replication sink" reports per peer sink<a name="line.65"></a>
-<span class="sourceLineNo">066</span> private final Map<ServerName, Integer> badReportCounts;<a name="line.66"></a>
-<span class="sourceLineNo">067</span><a name="line.67"></a>
-<span class="sourceLineNo">068</span> // Ratio of total number of potential peer region servers to be used<a name="line.68"></a>
-<span class="sourceLineNo">069</span> private final float ratio;<a name="line.69"></a>
-<span class="sourceLineNo">070</span><a name="line.70"></a>
-<span class="sourceLineNo">071</span> // Maximum number of times a sink can be reported as bad before the pool of<a name="line.71"></a>
-<span class="sourceLineNo">072</span> // replication sinks is refreshed<a name="line.72"></a>
-<span class="sourceLineNo">073</span> private final int badSinkThreshold;<a name="line.73"></a>
-<span class="sourceLineNo">074</span><a name="line.74"></a>
-<span class="sourceLineNo">075</span> private final Random random;<a name="line.75"></a>
-<span class="sourceLineNo">076</span><a name="line.76"></a>
-<span class="sourceLineNo">077</span> // A timestamp of the last time the list of replication peers changed<a name="line.77"></a>
-<span class="sourceLineNo">078</span> private long lastUpdateToPeers;<a name="line.78"></a>
-<span class="sourceLineNo">079</span><a name="line.79"></a>
-<span class="sourceLineNo">080</span> // The current pool of sinks to which replication can be performed<a name="line.80"></a>
-<span class="sourceLineNo">081</span> private List<ServerName> sinks = Lists.newArrayList();<a name="line.81"></a>
-<span class="sourceLineNo">082</span><a name="line.82"></a>
-<span class="sourceLineNo">083</span> /**<a name="line.83"></a>
-<span class="sourceLineNo">084</span> * Instantiate for a single replication peer cluster.<a name="line.84"></a>
-<span class="sourceLineNo">085</span> * @param conn connection to the peer cluster<a name="line.85"></a>
-<span class="sourceLineNo">086</span> * @param peerClusterId identifier of the peer cluster<a name="line.86"></a>
-<span class="sourceLineNo">087</span> * @param endpoint replication endpoint for inter cluster replication<a name="line.87"></a>
-<span class="sourceLineNo">088</span> * @param conf HBase configuration, used for determining replication source ratio and bad peer<a name="line.88"></a>
-<span class="sourceLineNo">089</span> * threshold<a name="line.89"></a>
-<span class="sourceLineNo">090</span> */<a name="line.90"></a>
-<span class="sourceLineNo">091</span> public ReplicationSinkManager(HConnection conn, String peerClusterId,<a name="line.91"></a>
-<span class="sourceLineNo">092</span> HBaseReplicationEndpoint endpoint, Configuration conf) {<a name="line.92"></a>
-<span class="sourceLineNo">093</span> this.conn = conn;<a name="line.93"></a>
-<span class="sourceLineNo">094</span> this.peerClusterId = peerClusterId;<a name="line.94"></a>
-<span class="sourceLineNo">095</span> this.endpoint = endpoint;<a name="line.95"></a>
-<span class="sourceLineNo">096</span> this.badReportCounts = Maps.newHashMap();<a name="line.96"></a>
-<span class="sourceLineNo">097</span> this.ratio = conf.getFloat("replication.source.ratio", DEFAULT_REPLICATION_SOURCE_RATIO);<a name="line.97"></a>
-<span class="sourceLineNo">098</span> this.badSinkThreshold = conf.getInt("replication.bad.sink.threshold",<a name="line.98"></a>
-<span class="sourceLineNo">099</span> DEFAULT_BAD_SINK_THRESHOLD);<a name="line.99"></a>
-<span class="sourceLineNo">100</span> this.random = new Random();<a name="line.100"></a>
-<span class="sourceLineNo">101</span> }<a name="line.101"></a>
-<span class="sourceLineNo">102</span><a name="line.102"></a>
-<span class="sourceLineNo">103</span> /**<a name="line.103"></a>
-<span class="sourceLineNo">104</span> * Get a randomly-chosen replication sink to replicate to.<a name="line.104"></a>
-<span class="sourceLineNo">105</span> *<a name="line.105"></a>
-<span class="sourceLineNo">106</span> * @return a replication sink to replicate to<a name="line.106"></a>
-<span class="sourceLineNo">107</span> */<a name="line.107"></a>
-<span class="sourceLineNo">108</span> public SinkPeer getReplicationSink() throws IOException {<a name="line.108"></a>
-<span class="sourceLineNo">109</span> if (endpoint.getLastRegionServerUpdate() > this.lastUpdateToPeers || sinks.isEmpty()) {<a name="line.109"></a>
-<span class="sourceLineNo">110</span> LOG.info("Current list of sinks is out of date or empty, updating");<a name="line.110"></a>
-<span class="sourceLineNo">111</span> chooseSinks();<a name="line.111"></a>
-<span class="sourceLineNo">112</span> }<a name="line.112"></a>
-<span class="sourceLineNo">113</span><a name="line.113"></a>
-<span class="sourceLineNo">114</span> if (sinks.isEmpty()) {<a name="line.114"></a>
-<span class="sourceLineNo">115</span> throw new IOException("No replication sinks are available");<a name="line.115"></a>
-<span class="sourceLineNo">116</span> }<a name="line.116"></a>
-<span class="sourceLineNo">117</span> ServerName serverName = sinks.get(random.nextInt(sinks.size()));<a name="line.117"></a>
-<span class="sourceLineNo">118</span> return new SinkPeer(serverName, conn.getAdmin(serverName));<a name="line.118"></a>
-<span class="sourceLineNo">119</span> }<a name="line.119"></a>
-<span class="sourceLineNo">120</span><a name="line.120"></a>
-<span class="sourceLineNo">121</span> /**<a name="line.121"></a>
-<span class="sourceLineNo">122</span> * Report a {@code SinkPeer} as being bad (i.e. an attempt to replicate to it<a name="line.122"></a>
-<span class="sourceLineNo">123</span> * failed). If a single SinkPeer is reported as bad more than<a name="line.123"></a>
-<span class="sourceLineNo">124</span> * replication.bad.sink.threshold times, it will be removed<a name="line.124"></a>
-<span class="sourceLineNo">125</span> * from the pool of potential replication targets.<a name="line.125"></a>
-<span class="sourceLineNo">126</span> *<a name="line.126"></a>
-<span class="sourceLineNo">127</span> * @param sinkPeer<a name="line.127"></a>
-<span class="sourceLineNo">128</span> * The SinkPeer that had a failed replication attempt on it<a name="line.128"></a>
-<span class="sourceLineNo">129</span> */<a name="line.129"></a>
-<span class="sourceLineNo">130</span> public void reportBadSink(SinkPeer sinkPeer) {<a name="line.130"></a>
-<span class="sourceLineNo">131</span> ServerName serverName = sinkPeer.getServerName();<a name="line.131"></a>
-<span class="sourceLineNo">132</span> int badReportCount = (badReportCounts.containsKey(serverName)<a name="line.132"></a>
-<span class="sourceLineNo">133</span> ? badReportCounts.get(serverName) : 0) + 1;<a name="line.133"></a>
-<span class="sourceLineNo">134</span> badReportCounts.put(serverName, badReportCount);<a name="line.134"></a>
-<span class="sourceLineNo">135</span> if (badReportCount > badSinkThreshold) {<a name="line.135"></a>
-<span class="sourceLineNo">136</span> this.sinks.remove(serverName);<a name="line.136"></a>
-<span class="sourceLineNo">137</span> if (sinks.isEmpty()) {<a name="line.137"></a>
-<span class="sourceLineNo">138</span> chooseSinks();<a name="line.138"></a>
-<span class="sourceLineNo">139</span> }<a name="line.139"></a>
-<span class="sourceLineNo">140</span> }<a name="line.140"></a>
-<span class="sourceLineNo">141</span> }<a name="line.141"></a>
-<span class="sourceLineNo">142</span><a name="line.142"></a>
-<span class="sourceLineNo">143</span> /**<a name="line.143"></a>
-<span class="sourceLineNo">144</span> * Report that a {@code SinkPeer} successfully replicated a chunk of data.<a name="line.144"></a>
-<span class="sourceLineNo">145</span> *<a name="line.145"></a>
-<span class="sourceLineNo">146</span> * @param sinkPeer<a name="line.146"></a>
-<span class="sourceLineNo">147</span> * The SinkPeer that had a failed replication attempt on it<a name="line.147"></a>
-<span class="sourceLineNo">148</span> */<a name="line.148"></a>
-<span class="sourceLineNo">149</span> public void reportSinkSuccess(SinkPeer sinkPeer) {<a name="line.149"></a>
-<span class="sourceLineNo">150</span> badReportCounts.remove(sinkPeer.getServerName());<a name="line.150"></a>
-<span class="sourceLineNo">151</span> }<a name="line.151"></a>
-<span class="sourceLineNo">152</span><a name="line.152"></a>
-<span class="sourceLineNo">153</span> void chooseSinks() {<a name="line.153"></a>
-<span class="sourceLineNo">154</span> List<ServerName> slaveAddresses = endpoint.getRegionServers();<a name="line.154"></a>
-<span class="sourceLineNo">155</span> Collections.shuffle(slaveAddresses, random);<a name="line.155"></a>
-<span class="sourceLineNo">156</span> int numSinks = (int) Math.ceil(slaveAddresses.size() * ratio);<a name="line.156"></a>
-<span class="sourceLineNo">157</span> sinks = slaveAddresses.subList(0, numSinks);<a name="line.157"></a>
-<span class="sourceLineNo">158</span> lastUpdateToPeers = System.currentTimeMillis();<a name="line.158"></a>
-<span class="sourceLineNo">159</span> badReportCounts.clear();<a name="line.159"></a>
-<span class="sourceLineNo">160</span> }<a name="line.160"></a>
-<span class="sourceLineNo">161</span><a name="line.161"></a>
-<span class="sourceLineNo">162</span> List<ServerName> getSinks() {<a name="line.162"></a>
-<span class="sourceLineNo">163</span> return sinks;<a name="line.163"></a>
+<span class="sourceLineNo">059</span><a name="line.59"></a>
+<span class="sourceLineNo">060</span> private final HConnection conn;<a name="line.60"></a>
+<span class="sourceLineNo">061</span><a name="line.61"></a>
+<span class="sourceLineNo">062</span> private final String peerClusterId;<a name="line.62"></a>
+<span class="sourceLineNo">063</span><a name="line.63"></a>
+<span class="sourceLineNo">064</span> private final HBaseReplicationEndpoint endpoint;<a name="line.64"></a>
+<span class="sourceLineNo">065</span><a name="line.65"></a>
+<span class="sourceLineNo">066</span> // Count of "bad replication sink" reports per peer sink<a name="line.66"></a>
+<span class="sourceLineNo">067</span> private final Map<ServerName, Integer> badReportCounts;<a name="line.67"></a>
+<span class="sourceLineNo">068</span><a name="line.68"></a>
+<span class="sourceLineNo">069</span> // Ratio of total number of potential peer region servers to be used<a name="line.69"></a>
+<span class="sourceLineNo">070</span> private final float ratio;<a name="line.70"></a>
+<span class="sourceLineNo">071</span><a name="line.71"></a>
+<span class="sourceLineNo">072</span> // Maximum number of times a sink can be reported as bad before the pool of<a name="line.72"></a>
+<span class="sourceLineNo">073</span> // replication sinks is refreshed<a name="line.73"></a>
+<span class="sourceLineNo">074</span> private final int badSinkThreshold;<a name="line.74"></a>
+<span class="sourceLineNo">075</span><a name="line.75"></a>
+<span class="sourceLineNo">076</span> private final Random random;<a name="line.76"></a>
+<span class="sourceLineNo">077</span><a name="line.77"></a>
+<span class="sourceLineNo">078</span> // A timestamp of the last time the list of replication peers changed<a name="line.78"></a>
+<span class="sourceLineNo">079</span> private long lastUpdateToPeers;<a name="line.79"></a>
+<span class="sourceLineNo">080</span><a name="line.80"></a>
+<span class="sourceLineNo">081</span> // The current pool of sinks to which replication can be performed<a name="line.81"></a>
+<span class="sourceLineNo">082</span> private List<ServerName> sinks = Lists.newArrayList();<a name="line.82"></a>
+<span class="sourceLineNo">083</span><a name="line.83"></a>
+<span class="sourceLineNo">084</span> /**<a name="line.84"></a>
+<span class="sourceLineNo">085</span> * Instantiate for a single replication peer cluster.<a name="line.85"></a>
+<span class="sourceLineNo">086</span> * @param conn connection to the peer cluster<a name="line.86"></a>
+<span class="sourceLineNo">087</span> * @param peerClusterId identifier of the peer cluster<a name="line.87"></a>
+<span class="sourceLineNo">088</span> * @param endpoint replication endpoint for inter cluster replication<a name="line.88"></a>
+<span class="sourceLineNo">089</span> * @param conf HBase configuration, used for determining replication source ratio and bad peer<a name="line.89"></a>
+<span class="sourceLineNo">090</span> * threshold<a name="line.90"></a>
+<span class="sourceLineNo">091</span> */<a name="line.91"></a>
+<span class="sourceLineNo">092</span> public ReplicationSinkManager(HConnection conn, String peerClusterId,<a name="line.92"></a>
+<span class="sourceLineNo">093</span> HBaseReplicationEndpoint endpoint, Configuration conf) {<a name="line.93"></a>
+<span class="sourceLineNo">094</span> this.conn = conn;<a name="line.94"></a>
+<span class="sourceLineNo">095</span> this.peerClusterId = peerClusterId;<a name="line.95"></a>
+<span class="sourceLineNo">096</span> this.endpoint = endpoint;<a name="line.96"></a>
+<span class="sourceLineNo">097</span> this.badReportCounts = Maps.newHashMap();<a name="line.97"></a>
+<span class="sourceLineNo">098</span> this.ratio = conf.getFloat("replication.source.ratio", DEFAULT_REPLICATION_SOURCE_RATIO);<a name="line.98"></a>
+<span class="sourceLineNo">099</span> this.badSinkThreshold = conf.getInt("replication.bad.sink.threshold",<a name="line.99"></a>
+<span class="sourceLineNo">100</span> DEFAULT_BAD_SINK_THRESHOLD);<a name="line.100"></a>
+<span class="sourceLineNo">101</span> this.random = new Random();<a name="line.101"></a>
+<span class="sourceLineNo">102</span> }<a name="line.102"></a>
+<span class="sourceLineNo">103</span><a name="line.103"></a>
+<span class="sourceLineNo">104</span> /**<a name="line.104"></a>
+<span class="sourceLineNo">105</span> * Get a randomly-chosen replication sink to replicate to.<a name="line.105"></a>
+<span class="sourceLineNo">106</span> *<a name="line.106"></a>
+<span class="sourceLineNo">107</span> * @return a replication sink to replicate to<a name="line.107"></a>
+<span class="sourceLineNo">108</span> */<a name="line.108"></a>
+<span class="sourceLineNo">109</span> public synchronized SinkPeer getReplicationSink() throws IOException {<a name="line.109"></a>
+<span class="sourceLineNo">110</span> if (endpoint.getLastRegionServerUpdate() > this.lastUpdateToPeers || sinks.isEmpty()) {<a name="line.110"></a>
+<span class="sourceLineNo">111</span> LOG.info("Current list of sinks is out of date or empty, updating");<a name="line.111"></a>
+<span class="sourceLineNo">112</span> chooseSinks();<a name="line.112"></a>
+<span class="sourceLineNo">113</span> }<a name="line.113"></a>
+<span class="sourceLineNo">114</span><a name="line.114"></a>
+<span class="sourceLineNo">115</span> if (sinks.isEmpty()) {<a name="line.115"></a>
+<span class="sourceLineNo">116</span> throw new IOException("No replication sinks are available");<a name="line.116"></a>
+<span class="sourceLineNo">117</span> }<a name="line.117"></a>
+<span class="sourceLineNo">118</span> ServerName serverName = sinks.get(random.nextInt(sinks.size()));<a name="line.118"></a>
+<span class="sourceLineNo">119</span> return new SinkPeer(serverName, conn.getAdmin(serverName));<a name="line.119"></a>
+<span class="sourceLineNo">120</span> }<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> * Report a {@code SinkPeer} as being bad (i.e. an attempt to replicate to it<a name="line.123"></a>
+<span class="sourceLineNo">124</span> * failed). If a single SinkPeer is reported as bad more than<a name="line.124"></a>
+<span class="sourceLineNo">125</span> * replication.bad.sink.threshold times, it will be removed<a name="line.125"></a>
+<span class="sourceLineNo">126</span> * from the pool of potential replication targets.<a name="line.126"></a>
+<span class="sourceLineNo">127</span> *<a name="line.127"></a>
+<span class="sourceLineNo">128</span> * @param sinkPeer<a name="line.128"></a>
+<span class="sourceLineNo">129</span> * The SinkPeer that had a failed replication attempt on it<a name="line.129"></a>
+<span class="sourceLineNo">130</span> */<a name="line.130"></a>
+<span class="sourceLineNo">131</span> public synchronized void reportBadSink(SinkPeer sinkPeer) {<a name="line.131"></a>
+<span class="sourceLineNo">132</span> ServerName serverName = sinkPeer.getServerName();<a name="line.132"></a>
+<span class="sourceLineNo">133</span> int badReportCount = (badReportCounts.containsKey(serverName)<a name="line.133"></a>
+<span class="sourceLineNo">134</span> ? badReportCounts.get(serverName) : 0) + 1;<a name="line.134"></a>
+<span class="sourceLineNo">135</span> badReportCounts.put(serverName, badReportCount);<a name="line.135"></a>
+<span class="sourceLineNo">136</span> if (badReportCount > badSinkThreshold) {<a name="line.136"></a>
+<span class="sourceLineNo">137</span> this.sinks.remove(serverName);<a name="line.137"></a>
+<span class="sourceLineNo">138</span> if (sinks.isEmpty()) {<a name="line.138"></a>
+<span class="sourceLineNo">139</span> chooseSinks();<a name="line.139"></a>
+<span class="sourceLineNo">140</span> }<a name="line.140"></a>
+<span class="sourceLineNo">141</span> }<a name="line.141"></a>
+<span class="sourceLineNo">142</span> }<a name="line.142"></a>
+<span class="sourceLineNo">143</span><a name="line.143"></a>
+<span class="sourceLineNo">144</span> /**<a name="line.144"></a>
+<span class="sourceLineNo">145</span> * Report that a {@code SinkPeer} successfully replicated a chunk of data.<a name="line.145"></a>
+<span class="sourceLineNo">146</span> *<a name="line.146"></a>
+<span class="sourceLineNo">147</span> * @param sinkPeer<a name="line.147"></a>
+<span class="sourceLineNo">148</span> * The SinkPeer that had a failed replication attempt on it<a name="line.148"></a>
+<span class="sourceLineNo">149</span> */<a name="line.149"></a>
+<span class="sourceLineNo">150</span> public synchronized void reportSinkSuccess(SinkPeer sinkPeer) {<a name="line.150"></a>
+<span class="sourceLineNo">151</span> badReportCounts.remove(sinkPeer.getServerName());<a name="line.151"></a>
+<span class="sourceLineNo">152</span> }<a name="line.152"></a>
+<span class="sourceLineNo">153</span><a name="line.153"></a>
+<span class="sourceLineNo">154</span> /**<a name="line.154"></a>
+<span class="sourceLineNo">155</span> * Refresh the list of sinks.<a name="line.155"></a>
+<span class="sourceLineNo">156</span> */<a name="line.156"></a>
+<span class="sourceLineNo">157</span> public synchronized void chooseSinks() {<a name="line.157"></a>
+<span class="sourceLineNo">158</span> List<ServerName> slaveAddresses = endpoint.getRegionServers();<a name="line.158"></a>
+<span class="sourceLineNo">159</span> Collections.shuffle(slaveAddresses, random);<a name="line.159"></a>
+<span class="sourceLineNo">160</span> int numSinks = (int) Math.ceil(slaveAddresses.size() * ratio);<a name="line.160"></a>
+<span class="sourceLineNo">161</span> sinks = slaveAddresses.subList(0, numSinks);<a name="line.161"></a>
+<span class="sourceLineNo">162</span> lastUpdateToPeers = System.currentTimeMillis();<a name="line.162"></a>
+<span class="sourceLineNo">163</span> badReportCounts.clear();<a name="line.163"></a>
<span class="sourceLineNo">164</span> }<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> * Wraps a replication region server sink to provide the ability to identify<a name="line.167"></a>
-<span class="sourceLineNo">168</span> * it.<a name="line.168"></a>
-<span class="sourceLineNo">169</span> */<a name="line.169"></a>
-<span class="sourceLineNo">170</span> public static class SinkPeer {<a name="line.170"></a>
-<span class="sourceLineNo">171</span> private ServerName serverName;<a name="line.171"></a>
-<span class="sourceLineNo">172</span> private AdminService.BlockingInterface regionServer;<a name="line.172"></a>
-<span class="sourceLineNo">173</span><a name="line.173"></a>
-<span class="sourceLineNo">174</span> public SinkPeer(ServerName serverName, AdminService.BlockingInterface regionServer) {<a name="line.174"></a>
-<span class="sourceLineNo">175</span> this.serverName = serverName;<a name="line.175"></a>
-<span class="sourceLineNo">176</span> this.regionServer = regionServer;<a name="line.176"></a>
-<span class="sourceLineNo">177</span> }<a name="line.177"></a>
-<span class="sourceLineNo">178</span><a name="line.178"></a>
-<span class="sourceLineNo">179</span> ServerName getServerName() {<a name="line.179"></a>
-<span class="sourceLineNo">180</span> return serverName;<a name="line.180"></a>
-<span class="sourceLineNo">181</span> }<a name="line.181"></a>
+<span class="sourceLineNo">166</span> public synchronized int getNumSinks() {<a name="line.166"></a>
+<span class="sourceLineNo">167</span> return sinks.size();<a name="line.167"></a>
+<span class="sourceLineNo">168</span> }<a name="line.168"></a>
+<span class="sourceLineNo">169</span><a name="line.169"></a>
+<span class="sourceLineNo">170</span> @VisibleForTesting<a name="line.170"></a>
+<span class="sourceLineNo">171</span> protected List<ServerName> getSinksForTesting() {<a name="line.171"></a>
+<span class="sourceLineNo">172</span> return Collections.unmodifiableList(sinks);<a name="line.172"></a>
+<span class="sourceLineNo">173</span> }<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> * Wraps a replication region server sink to provide the ability to identify<a name="line.176"></a>
+<span class="sourceLineNo">177</span> * it.<a name="line.177"></a>
+<span class="sourceLineNo">178</span> */<a name="line.178"></a>
+<span class="sourceLineNo">179</span> public static class SinkPeer {<a name="line.179"></a>
+<span class="sourceLineNo">180</span> private ServerName serverName;<a name="line.180"></a>
+<span class="sourceLineNo">181</span> private AdminService.BlockingInterface regionServer;<a name="line.181"></a>
<span class="sourceLineNo">182</span><a name="line.182"></a>
-<span class="sourceLineNo">183</span> public AdminService.BlockingInterface getRegionServer() {<a name="line.183"></a>
-<span class="sourceLineNo">184</span> return regionServer;<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><a name="line.188"></a>
-<span class="sourceLineNo">189</span>}<a name="line.189"></a>
+<span class="sourceLineNo">183</span> public SinkPeer(ServerName serverName, AdminService.BlockingInterface regionServer) {<a name="line.183"></a>
+<span class="sourceLineNo">184</span> this.serverName = serverName;<a name="line.184"></a>
+<span class="sourceLineNo">185</span> this.regionServer = regionServer;<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> ServerName getServerName() {<a name="line.188"></a>
+<span class="sourceLineNo">189</span> return serverName;<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> public AdminService.BlockingInterface getRegionServer() {<a name="line.192"></a>
+<span class="sourceLineNo">193</span> return regionServer;<a name="line.193"></a>
+<span class="sourceLineNo">194</span> }<a name="line.194"></a>
+<span class="sourceLineNo">195</span><a name="line.195"></a>
+<span class="sourceLineNo">196</span> }<a name="line.196"></a>
+<span class="sourceLineNo">197</span><a name="line.197"></a>
+<span class="sourceLineNo">198</span>}<a name="line.198"></a>