http://git-wip-us.apache.org/repos/asf/hbase-site/blob/2b3f2bee/apidocs/src-html/org/apache/hadoop/hbase/exceptions/RegionInRecoveryException.html ---------------------------------------------------------------------- diff --git a/apidocs/src-html/org/apache/hadoop/hbase/exceptions/RegionInRecoveryException.html b/apidocs/src-html/org/apache/hadoop/hbase/exceptions/RegionInRecoveryException.html deleted file mode 100644 index e324a9e..0000000 --- a/apidocs/src-html/org/apache/hadoop/hbase/exceptions/RegionInRecoveryException.html +++ /dev/null @@ -1,116 +0,0 @@ -<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd";> -<html lang="en"> -<head> -<title>Source code</title> -<link rel="stylesheet" type="text/css" href="../../../../../../stylesheet.css" title="Style"> -</head> -<body> -<div class="sourceContainer"> -<pre><span class="sourceLineNo">001</span>/**<a name="line.1"></a> -<span class="sourceLineNo">002</span> *<a name="line.2"></a> -<span class="sourceLineNo">003</span> * Licensed to the Apache Software Foundation (ASF) under one<a name="line.3"></a> -<span class="sourceLineNo">004</span> * or more contributor license agreements. See the NOTICE file<a name="line.4"></a> -<span class="sourceLineNo">005</span> * distributed with this work for additional information<a name="line.5"></a> -<span class="sourceLineNo">006</span> * regarding copyright ownership. The ASF licenses this file<a name="line.6"></a> -<span class="sourceLineNo">007</span> * to you under the Apache License, Version 2.0 (the<a name="line.7"></a> -<span class="sourceLineNo">008</span> * "License"); you may not use this file except in compliance<a name="line.8"></a> -<span class="sourceLineNo">009</span> * with the License. You may obtain a copy of the License at<a name="line.9"></a> -<span class="sourceLineNo">010</span> *<a name="line.10"></a> -<span class="sourceLineNo">011</span> * http://www.apache.org/licenses/LICENSE-2.0<a name="line.11"></a> -<span class="sourceLineNo">012</span> *<a name="line.12"></a> -<span class="sourceLineNo">013</span> * Unless required by applicable law or agreed to in writing, software<a name="line.13"></a> -<span class="sourceLineNo">014</span> * distributed under the License is distributed on an "AS IS" BASIS,<a name="line.14"></a> -<span class="sourceLineNo">015</span> * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.<a name="line.15"></a> -<span class="sourceLineNo">016</span> * See the License for the specific language governing permissions and<a name="line.16"></a> -<span class="sourceLineNo">017</span> * limitations under the License.<a name="line.17"></a> -<span class="sourceLineNo">018</span> */<a name="line.18"></a> -<span class="sourceLineNo">019</span>package org.apache.hadoop.hbase.exceptions;<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.NotServingRegionException;<a name="line.21"></a> -<span class="sourceLineNo">022</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.22"></a> -<span class="sourceLineNo">023</span><a name="line.23"></a> -<span class="sourceLineNo">024</span>/**<a name="line.24"></a> -<span class="sourceLineNo">025</span> * Thrown when a read request issued against a region which is in recovering state.<a name="line.25"></a> -<span class="sourceLineNo">026</span> */<a name="line.26"></a> -<span class="sourceLineNo">027</span>@InterfaceAudience.Public<a name="line.27"></a> -<span class="sourceLineNo">028</span>public class RegionInRecoveryException extends NotServingRegionException {<a name="line.28"></a> -<span class="sourceLineNo">029</span> private static final long serialVersionUID = 327302071153799L;<a name="line.29"></a> -<span class="sourceLineNo">030</span><a name="line.30"></a> -<span class="sourceLineNo">031</span> /** default constructor */<a name="line.31"></a> -<span class="sourceLineNo">032</span> public RegionInRecoveryException() {<a name="line.32"></a> -<span class="sourceLineNo">033</span> super();<a name="line.33"></a> -<span class="sourceLineNo">034</span> }<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> * Constructor<a name="line.37"></a> -<span class="sourceLineNo">038</span> * @param s message<a name="line.38"></a> -<span class="sourceLineNo">039</span> */<a name="line.39"></a> -<span class="sourceLineNo">040</span> public RegionInRecoveryException(String s) {<a name="line.40"></a> -<span class="sourceLineNo">041</span> super(s);<a name="line.41"></a> -<span class="sourceLineNo">042</span> }<a name="line.42"></a> -<span class="sourceLineNo">043</span><a name="line.43"></a> -<span class="sourceLineNo">044</span>}<a name="line.44"></a> - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -</pre> -</div> -</body> -</html>
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/2b3f2bee/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html ---------------------------------------------------------------------- diff --git a/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html b/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html index 898d59a..be26fad 100644 --- a/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html +++ b/apidocs/src-html/org/apache/hadoop/hbase/mapreduce/TableInputFormatBase.html @@ -276,391 +276,392 @@ <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> //The default value of "hbase.mapreduce.input.autobalance" is false.<a name="line.270"></a> -<span class="sourceLineNo">271</span> if (context.getConfiguration().getBoolean(MAPREDUCE_INPUT_AUTOBALANCE, false) != false) {<a name="line.271"></a> -<span class="sourceLineNo">272</span> long maxAveRegionSize = context.getConfiguration().getInt(MAX_AVERAGE_REGION_SIZE, 8*1073741824);<a name="line.272"></a> -<span class="sourceLineNo">273</span> return calculateAutoBalancedSplits(splits, maxAveRegionSize);<a name="line.273"></a> -<span class="sourceLineNo">274</span> }<a name="line.274"></a> -<span class="sourceLineNo">275</span><a name="line.275"></a> -<span class="sourceLineNo">276</span> // return one mapper per region<a name="line.276"></a> -<span class="sourceLineNo">277</span> return splits;<a name="line.277"></a> -<span class="sourceLineNo">278</span> } finally {<a name="line.278"></a> -<span class="sourceLineNo">279</span> if (closeOnFinish) {<a name="line.279"></a> -<span class="sourceLineNo">280</span> closeTable();<a name="line.280"></a> -<span class="sourceLineNo">281</span> }<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> /**<a name="line.285"></a> -<span class="sourceLineNo">286</span> * Create one InputSplit per region<a name="line.286"></a> -<span class="sourceLineNo">287</span> *<a name="line.287"></a> -<span class="sourceLineNo">288</span> * @return The list of InputSplit for all the regions<a name="line.288"></a> -<span class="sourceLineNo">289</span> * @throws IOException<a name="line.289"></a> -<span class="sourceLineNo">290</span> */<a name="line.290"></a> -<span class="sourceLineNo">291</span> private List<InputSplit> oneInputSplitPerRegion() throws IOException {<a name="line.291"></a> -<span class="sourceLineNo">292</span> RegionSizeCalculator sizeCalculator =<a name="line.292"></a> -<span class="sourceLineNo">293</span> new RegionSizeCalculator(getRegionLocator(), getAdmin());<a name="line.293"></a> -<span class="sourceLineNo">294</span><a name="line.294"></a> -<span class="sourceLineNo">295</span> TableName tableName = getTable().getName();<a name="line.295"></a> -<span class="sourceLineNo">296</span><a name="line.296"></a> -<span class="sourceLineNo">297</span> Pair<byte[][], byte[][]> keys = getStartEndKeys();<a name="line.297"></a> -<span class="sourceLineNo">298</span> if (keys == null || keys.getFirst() == null ||<a name="line.298"></a> -<span class="sourceLineNo">299</span> keys.getFirst().length == 0) {<a name="line.299"></a> -<span class="sourceLineNo">300</span> HRegionLocation regLoc =<a name="line.300"></a> -<span class="sourceLineNo">301</span> getRegionLocator().getRegionLocation(HConstants.EMPTY_BYTE_ARRAY, false);<a name="line.301"></a> -<span class="sourceLineNo">302</span> if (null == regLoc) {<a name="line.302"></a> -<span class="sourceLineNo">303</span> throw new IOException("Expecting at least one region.");<a name="line.303"></a> -<span class="sourceLineNo">304</span> }<a name="line.304"></a> -<span class="sourceLineNo">305</span> List<InputSplit> splits = new ArrayList<>(1);<a name="line.305"></a> -<span class="sourceLineNo">306</span> long regionSize = sizeCalculator.getRegionSize(regLoc.getRegionInfo().getRegionName());<a name="line.306"></a> -<span class="sourceLineNo">307</span> TableSplit split = new TableSplit(tableName, scan,<a name="line.307"></a> -<span class="sourceLineNo">308</span> HConstants.EMPTY_BYTE_ARRAY, HConstants.EMPTY_BYTE_ARRAY, regLoc<a name="line.308"></a> -<span class="sourceLineNo">309</span> .getHostnamePort().split(Addressing.HOSTNAME_PORT_SEPARATOR)[0], regionSize);<a name="line.309"></a> -<span class="sourceLineNo">310</span> splits.add(split);<a name="line.310"></a> -<span class="sourceLineNo">311</span> return splits;<a name="line.311"></a> -<span class="sourceLineNo">312</span> }<a name="line.312"></a> -<span class="sourceLineNo">313</span> List<InputSplit> splits = new ArrayList<>(keys.getFirst().length);<a name="line.313"></a> -<span class="sourceLineNo">314</span> for (int i = 0; i < keys.getFirst().length; i++) {<a name="line.314"></a> -<span class="sourceLineNo">315</span> if (!includeRegionInSplit(keys.getFirst()[i], keys.getSecond()[i])) {<a name="line.315"></a> -<span class="sourceLineNo">316</span> continue;<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> byte[] startRow = scan.getStartRow();<a name="line.319"></a> -<span class="sourceLineNo">320</span> byte[] stopRow = scan.getStopRow();<a name="line.320"></a> -<span class="sourceLineNo">321</span> // determine if the given start an stop key fall into the region<a name="line.321"></a> -<span class="sourceLineNo">322</span> if ((startRow.length == 0 || keys.getSecond()[i].length == 0 ||<a name="line.322"></a> -<span class="sourceLineNo">323</span> Bytes.compareTo(startRow, keys.getSecond()[i]) < 0) &&<a name="line.323"></a> -<span class="sourceLineNo">324</span> (stopRow.length == 0 ||<a name="line.324"></a> -<span class="sourceLineNo">325</span> Bytes.compareTo(stopRow, keys.getFirst()[i]) > 0)) {<a name="line.325"></a> -<span class="sourceLineNo">326</span> byte[] splitStart = startRow.length == 0 ||<a name="line.326"></a> -<span class="sourceLineNo">327</span> Bytes.compareTo(keys.getFirst()[i], startRow) >= 0 ?<a name="line.327"></a> -<span class="sourceLineNo">328</span> keys.getFirst()[i] : startRow;<a name="line.328"></a> -<span class="sourceLineNo">329</span> byte[] splitStop = (stopRow.length == 0 ||<a name="line.329"></a> -<span class="sourceLineNo">330</span> Bytes.compareTo(keys.getSecond()[i], stopRow) <= 0) &&<a name="line.330"></a> -<span class="sourceLineNo">331</span> keys.getSecond()[i].length > 0 ?<a name="line.331"></a> -<span class="sourceLineNo">332</span> keys.getSecond()[i] : stopRow;<a name="line.332"></a> -<span class="sourceLineNo">333</span><a name="line.333"></a> -<span class="sourceLineNo">334</span> HRegionLocation location = getRegionLocator().getRegionLocation(keys.getFirst()[i], false);<a name="line.334"></a> -<span class="sourceLineNo">335</span> // The below InetSocketAddress creation does a name resolution.<a name="line.335"></a> -<span class="sourceLineNo">336</span> InetSocketAddress isa = new InetSocketAddress(location.getHostname(), location.getPort());<a name="line.336"></a> -<span class="sourceLineNo">337</span> if (isa.isUnresolved()) {<a name="line.337"></a> -<span class="sourceLineNo">338</span> LOG.warn("Failed resolve " + isa);<a name="line.338"></a> -<span class="sourceLineNo">339</span> }<a name="line.339"></a> -<span class="sourceLineNo">340</span> InetAddress regionAddress = isa.getAddress();<a name="line.340"></a> -<span class="sourceLineNo">341</span> String regionLocation;<a name="line.341"></a> -<span class="sourceLineNo">342</span> regionLocation = reverseDNS(regionAddress);<a name="line.342"></a> -<span class="sourceLineNo">343</span><a name="line.343"></a> -<span class="sourceLineNo">344</span> byte[] regionName = location.getRegionInfo().getRegionName();<a name="line.344"></a> -<span class="sourceLineNo">345</span> String encodedRegionName = location.getRegionInfo().getEncodedName();<a name="line.345"></a> -<span class="sourceLineNo">346</span> long regionSize = sizeCalculator.getRegionSize(regionName);<a name="line.346"></a> -<span class="sourceLineNo">347</span> TableSplit split = new TableSplit(tableName, scan,<a name="line.347"></a> -<span class="sourceLineNo">348</span> splitStart, splitStop, regionLocation, encodedRegionName, regionSize);<a name="line.348"></a> -<span class="sourceLineNo">349</span> splits.add(split);<a name="line.349"></a> -<span class="sourceLineNo">350</span> if (LOG.isDebugEnabled()) {<a name="line.350"></a> -<span class="sourceLineNo">351</span> LOG.debug("getSplits: split -> " + i + " -> " + split);<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">354</span> }<a name="line.354"></a> -<span class="sourceLineNo">355</span> return splits;<a name="line.355"></a> -<span class="sourceLineNo">356</span> }<a name="line.356"></a> -<span class="sourceLineNo">357</span><a name="line.357"></a> -<span class="sourceLineNo">358</span> /**<a name="line.358"></a> -<span class="sourceLineNo">359</span> * Create n splits for one InputSplit, For now only support uniform distribution<a name="line.359"></a> -<span class="sourceLineNo">360</span> * @param split A TableSplit corresponding to a range of rowkeys<a name="line.360"></a> -<span class="sourceLineNo">361</span> * @param n Number of ranges after splitting. Pass 1 means no split for the range<a name="line.361"></a> -<span class="sourceLineNo">362</span> * Pass 2 if you want to split the range in two;<a name="line.362"></a> -<span class="sourceLineNo">363</span> * @return A list of TableSplit, the size of the list is n<a name="line.363"></a> -<span class="sourceLineNo">364</span> * @throws IllegalArgumentIOException<a name="line.364"></a> -<span class="sourceLineNo">365</span> */<a name="line.365"></a> -<span class="sourceLineNo">366</span> protected List<InputSplit> createNInputSplitsUniform(InputSplit split, int n)<a name="line.366"></a> -<span class="sourceLineNo">367</span> throws IllegalArgumentIOException {<a name="line.367"></a> -<span class="sourceLineNo">368</span> if (split == null || !(split instanceof TableSplit)) {<a name="line.368"></a> -<span class="sourceLineNo">369</span> throw new IllegalArgumentIOException(<a name="line.369"></a> -<span class="sourceLineNo">370</span> "InputSplit for CreateNSplitsPerRegion can not be null + "<a name="line.370"></a> -<span class="sourceLineNo">371</span> + "and should be instance of TableSplit");<a name="line.371"></a> -<span class="sourceLineNo">372</span> }<a name="line.372"></a> -<span class="sourceLineNo">373</span> //if n < 1, then still continue using n = 1<a name="line.373"></a> -<span class="sourceLineNo">374</span> n = n < 1 ? 1 : n;<a name="line.374"></a> -<span class="sourceLineNo">375</span> List<InputSplit> res = new ArrayList<>(n);<a name="line.375"></a> -<span class="sourceLineNo">376</span> if (n == 1) {<a name="line.376"></a> -<span class="sourceLineNo">377</span> res.add(split);<a name="line.377"></a> -<span class="sourceLineNo">378</span> return res;<a name="line.378"></a> -<span class="sourceLineNo">379</span> }<a name="line.379"></a> -<span class="sourceLineNo">380</span><a name="line.380"></a> -<span class="sourceLineNo">381</span> // Collect Region related information<a name="line.381"></a> -<span class="sourceLineNo">382</span> TableSplit ts = (TableSplit) split;<a name="line.382"></a> -<span class="sourceLineNo">383</span> TableName tableName = ts.getTable();<a name="line.383"></a> -<span class="sourceLineNo">384</span> String regionLocation = ts.getRegionLocation();<a name="line.384"></a> -<span class="sourceLineNo">385</span> String encodedRegionName = ts.getEncodedRegionName();<a name="line.385"></a> -<span class="sourceLineNo">386</span> long regionSize = ts.getLength();<a name="line.386"></a> -<span class="sourceLineNo">387</span> byte[] startRow = ts.getStartRow();<a name="line.387"></a> -<span class="sourceLineNo">388</span> byte[] endRow = ts.getEndRow();<a name="line.388"></a> -<span class="sourceLineNo">389</span><a name="line.389"></a> -<span class="sourceLineNo">390</span> // For special case: startRow or endRow is empty<a name="line.390"></a> -<span class="sourceLineNo">391</span> if (startRow.length == 0 && endRow.length == 0){<a name="line.391"></a> -<span class="sourceLineNo">392</span> startRow = new byte[1];<a name="line.392"></a> -<span class="sourceLineNo">393</span> endRow = new byte[1];<a name="line.393"></a> -<span class="sourceLineNo">394</span> startRow[0] = 0;<a name="line.394"></a> -<span class="sourceLineNo">395</span> endRow[0] = -1;<a name="line.395"></a> -<span class="sourceLineNo">396</span> }<a name="line.396"></a> -<span class="sourceLineNo">397</span> if (startRow.length == 0 && endRow.length != 0){<a name="line.397"></a> -<span class="sourceLineNo">398</span> startRow = new byte[1];<a name="line.398"></a> -<span class="sourceLineNo">399</span> startRow[0] = 0;<a name="line.399"></a> -<span class="sourceLineNo">400</span> }<a name="line.400"></a> -<span class="sourceLineNo">401</span> if (startRow.length != 0 && endRow.length == 0){<a name="line.401"></a> -<span class="sourceLineNo">402</span> endRow =new byte[startRow.length];<a name="line.402"></a> -<span class="sourceLineNo">403</span> for (int k = 0; k < startRow.length; k++){<a name="line.403"></a> -<span class="sourceLineNo">404</span> endRow[k] = -1;<a name="line.404"></a> -<span class="sourceLineNo">405</span> }<a name="line.405"></a> -<span class="sourceLineNo">406</span> }<a name="line.406"></a> -<span class="sourceLineNo">407</span><a name="line.407"></a> -<span class="sourceLineNo">408</span> // Split Region into n chunks evenly<a name="line.408"></a> -<span class="sourceLineNo">409</span> byte[][] splitKeys = Bytes.split(startRow, endRow, true, n-1);<a name="line.409"></a> -<span class="sourceLineNo">410</span> for (int i = 0; i < splitKeys.length - 1; i++) {<a name="line.410"></a> -<span class="sourceLineNo">411</span> //notice that the regionSize parameter may be not very accurate<a name="line.411"></a> -<span class="sourceLineNo">412</span> TableSplit tsplit =<a name="line.412"></a> -<span class="sourceLineNo">413</span> new TableSplit(tableName, scan, splitKeys[i], splitKeys[i + 1], regionLocation,<a name="line.413"></a> -<span class="sourceLineNo">414</span> encodedRegionName, regionSize / n);<a name="line.414"></a> -<span class="sourceLineNo">415</span> res.add(tsplit);<a name="line.415"></a> -<span class="sourceLineNo">416</span> }<a name="line.416"></a> -<span class="sourceLineNo">417</span> return res;<a name="line.417"></a> -<span class="sourceLineNo">418</span> }<a name="line.418"></a> -<span class="sourceLineNo">419</span> /**<a name="line.419"></a> -<span class="sourceLineNo">420</span> * Calculates the number of MapReduce input splits for the map tasks. The number of<a name="line.420"></a> -<span class="sourceLineNo">421</span> * MapReduce input splits depends on the average region size.<a name="line.421"></a> -<span class="sourceLineNo">422</span> * Make it 'public' for testing<a name="line.422"></a> -<span class="sourceLineNo">423</span> *<a name="line.423"></a> -<span class="sourceLineNo">424</span> * @param splits The list of input splits before balance.<a name="line.424"></a> -<span class="sourceLineNo">425</span> * @param maxAverageRegionSize max Average region size for one mapper<a name="line.425"></a> -<span class="sourceLineNo">426</span> * @return The list of input splits.<a name="line.426"></a> -<span class="sourceLineNo">427</span> * @throws IOException When creating the list of splits fails.<a name="line.427"></a> -<span class="sourceLineNo">428</span> * @see org.apache.hadoop.mapreduce.InputFormat#getSplits(<a name="line.428"></a> -<span class="sourceLineNo">429</span> *org.apache.hadoop.mapreduce.JobContext)<a name="line.429"></a> -<span class="sourceLineNo">430</span> */<a name="line.430"></a> -<span class="sourceLineNo">431</span> public List<InputSplit> calculateAutoBalancedSplits(List<InputSplit> splits, long maxAverageRegionSize)<a name="line.431"></a> -<span class="sourceLineNo">432</span> throws IOException {<a name="line.432"></a> -<span class="sourceLineNo">433</span> if (splits.size() == 0) {<a name="line.433"></a> -<span class="sourceLineNo">434</span> return splits;<a name="line.434"></a> -<span class="sourceLineNo">435</span> }<a name="line.435"></a> -<span class="sourceLineNo">436</span> List<InputSplit> resultList = new ArrayList<>();<a name="line.436"></a> -<span class="sourceLineNo">437</span> long totalRegionSize = 0;<a name="line.437"></a> -<span class="sourceLineNo">438</span> for (int i = 0; i < splits.size(); i++) {<a name="line.438"></a> -<span class="sourceLineNo">439</span> TableSplit ts = (TableSplit) splits.get(i);<a name="line.439"></a> -<span class="sourceLineNo">440</span> totalRegionSize += ts.getLength();<a name="line.440"></a> -<span class="sourceLineNo">441</span> }<a name="line.441"></a> -<span class="sourceLineNo">442</span> long averageRegionSize = totalRegionSize / splits.size();<a name="line.442"></a> -<span class="sourceLineNo">443</span> // totalRegionSize might be overflow, and the averageRegionSize must be positive.<a name="line.443"></a> -<span class="sourceLineNo">444</span> if (averageRegionSize <= 0) {<a name="line.444"></a> -<span class="sourceLineNo">445</span> LOG.warn("The averageRegionSize is not positive: " + averageRegionSize + ", " +<a name="line.445"></a> -<span class="sourceLineNo">446</span> "set it to Long.MAX_VALUE " + splits.size());<a name="line.446"></a> -<span class="sourceLineNo">447</span> averageRegionSize = Long.MAX_VALUE / splits.size();<a name="line.447"></a> -<span class="sourceLineNo">448</span> }<a name="line.448"></a> -<span class="sourceLineNo">449</span> //if averageRegionSize is too big, change it to default as 1 GB,<a name="line.449"></a> -<span class="sourceLineNo">450</span> if (averageRegionSize > maxAverageRegionSize) {<a name="line.450"></a> -<span class="sourceLineNo">451</span> averageRegionSize = maxAverageRegionSize;<a name="line.451"></a> -<span class="sourceLineNo">452</span> }<a name="line.452"></a> -<span class="sourceLineNo">453</span> // if averageRegionSize is too small, we do not need to allocate more mappers for those 'large' region<a name="line.453"></a> -<span class="sourceLineNo">454</span> // set default as 16M = (default hdfs block size) / 4;<a name="line.454"></a> -<span class="sourceLineNo">455</span> if (averageRegionSize < 16 * 1048576) {<a name="line.455"></a> -<span class="sourceLineNo">456</span> return splits;<a name="line.456"></a> -<span class="sourceLineNo">457</span> }<a name="line.457"></a> -<span class="sourceLineNo">458</span> for (int i = 0; i < splits.size(); i++) {<a name="line.458"></a> -<span class="sourceLineNo">459</span> TableSplit ts = (TableSplit) splits.get(i);<a name="line.459"></a> -<span class="sourceLineNo">460</span> TableName tableName = ts.getTable();<a name="line.460"></a> -<span class="sourceLineNo">461</span> String regionLocation = ts.getRegionLocation();<a name="line.461"></a> -<span class="sourceLineNo">462</span> String encodedRegionName = ts.getEncodedRegionName();<a name="line.462"></a> -<span class="sourceLineNo">463</span> long regionSize = ts.getLength();<a name="line.463"></a> -<span class="sourceLineNo">464</span><a name="line.464"></a> -<span class="sourceLineNo">465</span> if (regionSize >= averageRegionSize) {<a name="line.465"></a> -<span class="sourceLineNo">466</span> // make this region as multiple MapReduce input split.<a name="line.466"></a> -<span class="sourceLineNo">467</span> int n = (int) Math.round(Math.log(((double) regionSize) / ((double) averageRegionSize)) + 1.0);<a name="line.467"></a> -<span class="sourceLineNo">468</span> List<InputSplit> temp = createNInputSplitsUniform(ts, n);<a name="line.468"></a> -<span class="sourceLineNo">469</span> resultList.addAll(temp);<a name="line.469"></a> -<span class="sourceLineNo">470</span> } else {<a name="line.470"></a> -<span class="sourceLineNo">471</span> // if the total size of several small continuous regions less than the average region size,<a name="line.471"></a> -<span class="sourceLineNo">472</span> // combine them into one MapReduce input split.<a name="line.472"></a> -<span class="sourceLineNo">473</span> long totalSize = regionSize;<a name="line.473"></a> -<span class="sourceLineNo">474</span> byte[] splitStartKey = ts.getStartRow();<a name="line.474"></a> -<span class="sourceLineNo">475</span> byte[] splitEndKey = ts.getEndRow();<a name="line.475"></a> -<span class="sourceLineNo">476</span> int j = i + 1;<a name="line.476"></a> -<span class="sourceLineNo">477</span> while (j < splits.size()) {<a name="line.477"></a> -<span class="sourceLineNo">478</span> TableSplit nextRegion = (TableSplit) splits.get(j);<a name="line.478"></a> -<span class="sourceLineNo">479</span> long nextRegionSize = nextRegion.getLength();<a name="line.479"></a> -<span class="sourceLineNo">480</span> if (totalSize + nextRegionSize <= averageRegionSize) {<a name="line.480"></a> -<span class="sourceLineNo">481</span> totalSize = totalSize + nextRegionSize;<a name="line.481"></a> -<span class="sourceLineNo">482</span> splitEndKey = nextRegion.getEndRow();<a name="line.482"></a> -<span class="sourceLineNo">483</span> j++;<a name="line.483"></a> -<span class="sourceLineNo">484</span> } else {<a name="line.484"></a> -<span class="sourceLineNo">485</span> break;<a name="line.485"></a> -<span class="sourceLineNo">486</span> }<a name="line.486"></a> -<span class="sourceLineNo">487</span> }<a name="line.487"></a> -<span class="sourceLineNo">488</span> i = j - 1;<a name="line.488"></a> -<span class="sourceLineNo">489</span> TableSplit t = new TableSplit(tableName, scan, splitStartKey, splitEndKey, regionLocation,<a name="line.489"></a> -<span class="sourceLineNo">490</span> encodedRegionName, totalSize);<a name="line.490"></a> -<span class="sourceLineNo">491</span> resultList.add(t);<a name="line.491"></a> -<span class="sourceLineNo">492</span> }<a name="line.492"></a> -<span class="sourceLineNo">493</span> }<a name="line.493"></a> -<span class="sourceLineNo">494</span> return resultList;<a name="line.494"></a> -<span class="sourceLineNo">495</span> }<a name="line.495"></a> -<span class="sourceLineNo">496</span><a name="line.496"></a> -<span class="sourceLineNo">497</span> String reverseDNS(InetAddress ipAddress) throws UnknownHostException {<a name="line.497"></a> -<span class="sourceLineNo">498</span> String hostName = this.reverseDNSCacheMap.get(ipAddress);<a name="line.498"></a> -<span class="sourceLineNo">499</span> if (hostName == null) {<a name="line.499"></a> -<span class="sourceLineNo">500</span> String ipAddressString = null;<a name="line.500"></a> -<span class="sourceLineNo">501</span> try {<a name="line.501"></a> -<span class="sourceLineNo">502</span> ipAddressString = DNS.reverseDns(ipAddress, null);<a name="line.502"></a> -<span class="sourceLineNo">503</span> } catch (Exception e) {<a name="line.503"></a> -<span class="sourceLineNo">504</span> // We can use InetAddress in case the jndi failed to pull up the reverse DNS entry from the<a name="line.504"></a> -<span class="sourceLineNo">505</span> // name service. Also, in case of ipv6, we need to use the InetAddress since resolving<a name="line.505"></a> -<span class="sourceLineNo">506</span> // reverse DNS using jndi doesn't work well with ipv6 addresses.<a name="line.506"></a> -<span class="sourceLineNo">507</span> ipAddressString = InetAddress.getByName(ipAddress.getHostAddress()).getHostName();<a name="line.507"></a> -<span class="sourceLineNo">508</span> }<a name="line.508"></a> -<span class="sourceLineNo">509</span> if (ipAddressString == null) throw new UnknownHostException("No host found for " + ipAddress);<a name="line.509"></a> -<span class="sourceLineNo">510</span> hostName = Strings.domainNamePointerToHostName(ipAddressString);<a name="line.510"></a> -<span class="sourceLineNo">511</span> this.reverseDNSCacheMap.put(ipAddress, hostName);<a name="line.511"></a> -<span class="sourceLineNo">512</span> }<a name="line.512"></a> -<span class="sourceLineNo">513</span> return hostName;<a name="line.513"></a> -<span class="sourceLineNo">514</span> }<a name="line.514"></a> -<span class="sourceLineNo">515</span><a name="line.515"></a> -<span class="sourceLineNo">516</span> /**<a name="line.516"></a> -<span class="sourceLineNo">517</span> * Test if the given region is to be included in the InputSplit while splitting<a name="line.517"></a> -<span class="sourceLineNo">518</span> * the regions of a table.<a name="line.518"></a> -<span class="sourceLineNo">519</span> * <p><a name="line.519"></a> -<span class="sourceLineNo">520</span> * This optimization is effective when there is a specific reasoning to exclude an entire region from the M-R job,<a name="line.520"></a> -<span class="sourceLineNo">521</span> * (and hence, not contributing to the InputSplit), given the start and end keys of the same. <br><a name="line.521"></a> -<span class="sourceLineNo">522</span> * Useful when we need to remember the last-processed top record and revisit the [last, current) interval for M-R processing,<a name="line.522"></a> -<span class="sourceLineNo">523</span> * continuously. In addition to reducing InputSplits, reduces the load on the region server as well, due to the ordering of the keys.<a name="line.523"></a> -<span class="sourceLineNo">524</span> * <br><a name="line.524"></a> +<span class="sourceLineNo">271</span> if (context.getConfiguration().getBoolean(MAPREDUCE_INPUT_AUTOBALANCE, false)) {<a name="line.271"></a> +<span class="sourceLineNo">272</span> long maxAveRegionSize = context.getConfiguration()<a name="line.272"></a> +<span class="sourceLineNo">273</span> .getLong(MAX_AVERAGE_REGION_SIZE, 8L*1073741824); //8GB<a name="line.273"></a> +<span class="sourceLineNo">274</span> return calculateAutoBalancedSplits(splits, maxAveRegionSize);<a name="line.274"></a> +<span class="sourceLineNo">275</span> }<a name="line.275"></a> +<span class="sourceLineNo">276</span><a name="line.276"></a> +<span class="sourceLineNo">277</span> // return one mapper per region<a name="line.277"></a> +<span class="sourceLineNo">278</span> return splits;<a name="line.278"></a> +<span class="sourceLineNo">279</span> } finally {<a name="line.279"></a> +<span class="sourceLineNo">280</span> if (closeOnFinish) {<a name="line.280"></a> +<span class="sourceLineNo">281</span> closeTable();<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><a name="line.285"></a> +<span class="sourceLineNo">286</span> /**<a name="line.286"></a> +<span class="sourceLineNo">287</span> * Create one InputSplit per region<a name="line.287"></a> +<span class="sourceLineNo">288</span> *<a name="line.288"></a> +<span class="sourceLineNo">289</span> * @return The list of InputSplit for all the regions<a name="line.289"></a> +<span class="sourceLineNo">290</span> * @throws IOException<a name="line.290"></a> +<span class="sourceLineNo">291</span> */<a name="line.291"></a> +<span class="sourceLineNo">292</span> private List<InputSplit> oneInputSplitPerRegion() throws IOException {<a name="line.292"></a> +<span class="sourceLineNo">293</span> RegionSizeCalculator sizeCalculator =<a name="line.293"></a> +<span class="sourceLineNo">294</span> new RegionSizeCalculator(getRegionLocator(), getAdmin());<a name="line.294"></a> +<span class="sourceLineNo">295</span><a name="line.295"></a> +<span class="sourceLineNo">296</span> TableName tableName = getTable().getName();<a name="line.296"></a> +<span class="sourceLineNo">297</span><a name="line.297"></a> +<span class="sourceLineNo">298</span> Pair<byte[][], byte[][]> keys = getStartEndKeys();<a name="line.298"></a> +<span class="sourceLineNo">299</span> if (keys == null || keys.getFirst() == null ||<a name="line.299"></a> +<span class="sourceLineNo">300</span> keys.getFirst().length == 0) {<a name="line.300"></a> +<span class="sourceLineNo">301</span> HRegionLocation regLoc =<a name="line.301"></a> +<span class="sourceLineNo">302</span> getRegionLocator().getRegionLocation(HConstants.EMPTY_BYTE_ARRAY, false);<a name="line.302"></a> +<span class="sourceLineNo">303</span> if (null == regLoc) {<a name="line.303"></a> +<span class="sourceLineNo">304</span> throw new IOException("Expecting at least one region.");<a name="line.304"></a> +<span class="sourceLineNo">305</span> }<a name="line.305"></a> +<span class="sourceLineNo">306</span> List<InputSplit> splits = new ArrayList<>(1);<a name="line.306"></a> +<span class="sourceLineNo">307</span> long regionSize = sizeCalculator.getRegionSize(regLoc.getRegionInfo().getRegionName());<a name="line.307"></a> +<span class="sourceLineNo">308</span> TableSplit split = new TableSplit(tableName, scan,<a name="line.308"></a> +<span class="sourceLineNo">309</span> HConstants.EMPTY_BYTE_ARRAY, HConstants.EMPTY_BYTE_ARRAY, regLoc<a name="line.309"></a> +<span class="sourceLineNo">310</span> .getHostnamePort().split(Addressing.HOSTNAME_PORT_SEPARATOR)[0], regionSize);<a name="line.310"></a> +<span class="sourceLineNo">311</span> splits.add(split);<a name="line.311"></a> +<span class="sourceLineNo">312</span> return splits;<a name="line.312"></a> +<span class="sourceLineNo">313</span> }<a name="line.313"></a> +<span class="sourceLineNo">314</span> List<InputSplit> splits = new ArrayList<>(keys.getFirst().length);<a name="line.314"></a> +<span class="sourceLineNo">315</span> for (int i = 0; i < keys.getFirst().length; i++) {<a name="line.315"></a> +<span class="sourceLineNo">316</span> if (!includeRegionInSplit(keys.getFirst()[i], keys.getSecond()[i])) {<a name="line.316"></a> +<span class="sourceLineNo">317</span> continue;<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> byte[] startRow = scan.getStartRow();<a name="line.320"></a> +<span class="sourceLineNo">321</span> byte[] stopRow = scan.getStopRow();<a name="line.321"></a> +<span class="sourceLineNo">322</span> // determine if the given start an stop key fall into the region<a name="line.322"></a> +<span class="sourceLineNo">323</span> if ((startRow.length == 0 || keys.getSecond()[i].length == 0 ||<a name="line.323"></a> +<span class="sourceLineNo">324</span> Bytes.compareTo(startRow, keys.getSecond()[i]) < 0) &&<a name="line.324"></a> +<span class="sourceLineNo">325</span> (stopRow.length == 0 ||<a name="line.325"></a> +<span class="sourceLineNo">326</span> Bytes.compareTo(stopRow, keys.getFirst()[i]) > 0)) {<a name="line.326"></a> +<span class="sourceLineNo">327</span> byte[] splitStart = startRow.length == 0 ||<a name="line.327"></a> +<span class="sourceLineNo">328</span> Bytes.compareTo(keys.getFirst()[i], startRow) >= 0 ?<a name="line.328"></a> +<span class="sourceLineNo">329</span> keys.getFirst()[i] : startRow;<a name="line.329"></a> +<span class="sourceLineNo">330</span> byte[] splitStop = (stopRow.length == 0 ||<a name="line.330"></a> +<span class="sourceLineNo">331</span> Bytes.compareTo(keys.getSecond()[i], stopRow) <= 0) &&<a name="line.331"></a> +<span class="sourceLineNo">332</span> keys.getSecond()[i].length > 0 ?<a name="line.332"></a> +<span class="sourceLineNo">333</span> keys.getSecond()[i] : stopRow;<a name="line.333"></a> +<span class="sourceLineNo">334</span><a name="line.334"></a> +<span class="sourceLineNo">335</span> HRegionLocation location = getRegionLocator().getRegionLocation(keys.getFirst()[i], false);<a name="line.335"></a> +<span class="sourceLineNo">336</span> // The below InetSocketAddress creation does a name resolution.<a name="line.336"></a> +<span class="sourceLineNo">337</span> InetSocketAddress isa = new InetSocketAddress(location.getHostname(), location.getPort());<a name="line.337"></a> +<span class="sourceLineNo">338</span> if (isa.isUnresolved()) {<a name="line.338"></a> +<span class="sourceLineNo">339</span> LOG.warn("Failed resolve " + isa);<a name="line.339"></a> +<span class="sourceLineNo">340</span> }<a name="line.340"></a> +<span class="sourceLineNo">341</span> InetAddress regionAddress = isa.getAddress();<a name="line.341"></a> +<span class="sourceLineNo">342</span> String regionLocation;<a name="line.342"></a> +<span class="sourceLineNo">343</span> regionLocation = reverseDNS(regionAddress);<a name="line.343"></a> +<span class="sourceLineNo">344</span><a name="line.344"></a> +<span class="sourceLineNo">345</span> byte[] regionName = location.getRegionInfo().getRegionName();<a name="line.345"></a> +<span class="sourceLineNo">346</span> String encodedRegionName = location.getRegionInfo().getEncodedName();<a name="line.346"></a> +<span class="sourceLineNo">347</span> long regionSize = sizeCalculator.getRegionSize(regionName);<a name="line.347"></a> +<span class="sourceLineNo">348</span> TableSplit split = new TableSplit(tableName, scan,<a name="line.348"></a> +<span class="sourceLineNo">349</span> splitStart, splitStop, regionLocation, encodedRegionName, regionSize);<a name="line.349"></a> +<span class="sourceLineNo">350</span> splits.add(split);<a name="line.350"></a> +<span class="sourceLineNo">351</span> if (LOG.isDebugEnabled()) {<a name="line.351"></a> +<span class="sourceLineNo">352</span> LOG.debug("getSplits: split -> " + i + " -> " + split);<a name="line.352"></a> +<span class="sourceLineNo">353</span> }<a name="line.353"></a> +<span class="sourceLineNo">354</span> }<a name="line.354"></a> +<span class="sourceLineNo">355</span> }<a name="line.355"></a> +<span class="sourceLineNo">356</span> return splits;<a name="line.356"></a> +<span class="sourceLineNo">357</span> }<a name="line.357"></a> +<span class="sourceLineNo">358</span><a name="line.358"></a> +<span class="sourceLineNo">359</span> /**<a name="line.359"></a> +<span class="sourceLineNo">360</span> * Create n splits for one InputSplit, For now only support uniform distribution<a name="line.360"></a> +<span class="sourceLineNo">361</span> * @param split A TableSplit corresponding to a range of rowkeys<a name="line.361"></a> +<span class="sourceLineNo">362</span> * @param n Number of ranges after splitting. Pass 1 means no split for the range<a name="line.362"></a> +<span class="sourceLineNo">363</span> * Pass 2 if you want to split the range in two;<a name="line.363"></a> +<span class="sourceLineNo">364</span> * @return A list of TableSplit, the size of the list is n<a name="line.364"></a> +<span class="sourceLineNo">365</span> * @throws IllegalArgumentIOException<a name="line.365"></a> +<span class="sourceLineNo">366</span> */<a name="line.366"></a> +<span class="sourceLineNo">367</span> protected List<InputSplit> createNInputSplitsUniform(InputSplit split, int n)<a name="line.367"></a> +<span class="sourceLineNo">368</span> throws IllegalArgumentIOException {<a name="line.368"></a> +<span class="sourceLineNo">369</span> if (split == null || !(split instanceof TableSplit)) {<a name="line.369"></a> +<span class="sourceLineNo">370</span> throw new IllegalArgumentIOException(<a name="line.370"></a> +<span class="sourceLineNo">371</span> "InputSplit for CreateNSplitsPerRegion can not be null + "<a name="line.371"></a> +<span class="sourceLineNo">372</span> + "and should be instance of TableSplit");<a name="line.372"></a> +<span class="sourceLineNo">373</span> }<a name="line.373"></a> +<span class="sourceLineNo">374</span> //if n < 1, then still continue using n = 1<a name="line.374"></a> +<span class="sourceLineNo">375</span> n = n < 1 ? 1 : n;<a name="line.375"></a> +<span class="sourceLineNo">376</span> List<InputSplit> res = new ArrayList<>(n);<a name="line.376"></a> +<span class="sourceLineNo">377</span> if (n == 1) {<a name="line.377"></a> +<span class="sourceLineNo">378</span> res.add(split);<a name="line.378"></a> +<span class="sourceLineNo">379</span> return res;<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> // Collect Region related information<a name="line.382"></a> +<span class="sourceLineNo">383</span> TableSplit ts = (TableSplit) split;<a name="line.383"></a> +<span class="sourceLineNo">384</span> TableName tableName = ts.getTable();<a name="line.384"></a> +<span class="sourceLineNo">385</span> String regionLocation = ts.getRegionLocation();<a name="line.385"></a> +<span class="sourceLineNo">386</span> String encodedRegionName = ts.getEncodedRegionName();<a name="line.386"></a> +<span class="sourceLineNo">387</span> long regionSize = ts.getLength();<a name="line.387"></a> +<span class="sourceLineNo">388</span> byte[] startRow = ts.getStartRow();<a name="line.388"></a> +<span class="sourceLineNo">389</span> byte[] endRow = ts.getEndRow();<a name="line.389"></a> +<span class="sourceLineNo">390</span><a name="line.390"></a> +<span class="sourceLineNo">391</span> // For special case: startRow or endRow is empty<a name="line.391"></a> +<span class="sourceLineNo">392</span> if (startRow.length == 0 && endRow.length == 0){<a name="line.392"></a> +<span class="sourceLineNo">393</span> startRow = new byte[1];<a name="line.393"></a> +<span class="sourceLineNo">394</span> endRow = new byte[1];<a name="line.394"></a> +<span class="sourceLineNo">395</span> startRow[0] = 0;<a name="line.395"></a> +<span class="sourceLineNo">396</span> endRow[0] = -1;<a name="line.396"></a> +<span class="sourceLineNo">397</span> }<a name="line.397"></a> +<span class="sourceLineNo">398</span> if (startRow.length == 0 && endRow.length != 0){<a name="line.398"></a> +<span class="sourceLineNo">399</span> startRow = new byte[1];<a name="line.399"></a> +<span class="sourceLineNo">400</span> startRow[0] = 0;<a name="line.400"></a> +<span class="sourceLineNo">401</span> }<a name="line.401"></a> +<span class="sourceLineNo">402</span> if (startRow.length != 0 && endRow.length == 0){<a name="line.402"></a> +<span class="sourceLineNo">403</span> endRow =new byte[startRow.length];<a name="line.403"></a> +<span class="sourceLineNo">404</span> for (int k = 0; k < startRow.length; k++){<a name="line.404"></a> +<span class="sourceLineNo">405</span> endRow[k] = -1;<a name="line.405"></a> +<span class="sourceLineNo">406</span> }<a name="line.406"></a> +<span class="sourceLineNo">407</span> }<a name="line.407"></a> +<span class="sourceLineNo">408</span><a name="line.408"></a> +<span class="sourceLineNo">409</span> // Split Region into n chunks evenly<a name="line.409"></a> +<span class="sourceLineNo">410</span> byte[][] splitKeys = Bytes.split(startRow, endRow, true, n-1);<a name="line.410"></a> +<span class="sourceLineNo">411</span> for (int i = 0; i < splitKeys.length - 1; i++) {<a name="line.411"></a> +<span class="sourceLineNo">412</span> //notice that the regionSize parameter may be not very accurate<a name="line.412"></a> +<span class="sourceLineNo">413</span> TableSplit tsplit =<a name="line.413"></a> +<span class="sourceLineNo">414</span> new TableSplit(tableName, scan, splitKeys[i], splitKeys[i + 1], regionLocation,<a name="line.414"></a> +<span class="sourceLineNo">415</span> encodedRegionName, regionSize / n);<a name="line.415"></a> +<span class="sourceLineNo">416</span> res.add(tsplit);<a name="line.416"></a> +<span class="sourceLineNo">417</span> }<a name="line.417"></a> +<span class="sourceLineNo">418</span> return res;<a name="line.418"></a> +<span class="sourceLineNo">419</span> }<a name="line.419"></a> +<span class="sourceLineNo">420</span> /**<a name="line.420"></a> +<span class="sourceLineNo">421</span> * Calculates the number of MapReduce input splits for the map tasks. The number of<a name="line.421"></a> +<span class="sourceLineNo">422</span> * MapReduce input splits depends on the average region size.<a name="line.422"></a> +<span class="sourceLineNo">423</span> * Make it 'public' for testing<a name="line.423"></a> +<span class="sourceLineNo">424</span> *<a name="line.424"></a> +<span class="sourceLineNo">425</span> * @param splits The list of input splits before balance.<a name="line.425"></a> +<span class="sourceLineNo">426</span> * @param maxAverageRegionSize max Average region size for one mapper<a name="line.426"></a> +<span class="sourceLineNo">427</span> * @return The list of input splits.<a name="line.427"></a> +<span class="sourceLineNo">428</span> * @throws IOException When creating the list of splits fails.<a name="line.428"></a> +<span class="sourceLineNo">429</span> * @see org.apache.hadoop.mapreduce.InputFormat#getSplits(<a name="line.429"></a> +<span class="sourceLineNo">430</span> *org.apache.hadoop.mapreduce.JobContext)<a name="line.430"></a> +<span class="sourceLineNo">431</span> */<a name="line.431"></a> +<span class="sourceLineNo">432</span> public List<InputSplit> calculateAutoBalancedSplits(List<InputSplit> splits, long maxAverageRegionSize)<a name="line.432"></a> +<span class="sourceLineNo">433</span> throws IOException {<a name="line.433"></a> +<span class="sourceLineNo">434</span> if (splits.size() == 0) {<a name="line.434"></a> +<span class="sourceLineNo">435</span> return splits;<a name="line.435"></a> +<span class="sourceLineNo">436</span> }<a name="line.436"></a> +<span class="sourceLineNo">437</span> List<InputSplit> resultList = new ArrayList<>();<a name="line.437"></a> +<span class="sourceLineNo">438</span> long totalRegionSize = 0;<a name="line.438"></a> +<span class="sourceLineNo">439</span> for (int i = 0; i < splits.size(); i++) {<a name="line.439"></a> +<span class="sourceLineNo">440</span> TableSplit ts = (TableSplit) splits.get(i);<a name="line.440"></a> +<span class="sourceLineNo">441</span> totalRegionSize += ts.getLength();<a name="line.441"></a> +<span class="sourceLineNo">442</span> }<a name="line.442"></a> +<span class="sourceLineNo">443</span> long averageRegionSize = totalRegionSize / splits.size();<a name="line.443"></a> +<span class="sourceLineNo">444</span> // totalRegionSize might be overflow, and the averageRegionSize must be positive.<a name="line.444"></a> +<span class="sourceLineNo">445</span> if (averageRegionSize <= 0) {<a name="line.445"></a> +<span class="sourceLineNo">446</span> LOG.warn("The averageRegionSize is not positive: " + averageRegionSize + ", " +<a name="line.446"></a> +<span class="sourceLineNo">447</span> "set it to Long.MAX_VALUE " + splits.size());<a name="line.447"></a> +<span class="sourceLineNo">448</span> averageRegionSize = Long.MAX_VALUE / splits.size();<a name="line.448"></a> +<span class="sourceLineNo">449</span> }<a name="line.449"></a> +<span class="sourceLineNo">450</span> //if averageRegionSize is too big, change it to default as 1 GB,<a name="line.450"></a> +<span class="sourceLineNo">451</span> if (averageRegionSize > maxAverageRegionSize) {<a name="line.451"></a> +<span class="sourceLineNo">452</span> averageRegionSize = maxAverageRegionSize;<a name="line.452"></a> +<span class="sourceLineNo">453</span> }<a name="line.453"></a> +<span class="sourceLineNo">454</span> // if averageRegionSize is too small, we do not need to allocate more mappers for those 'large' region<a name="line.454"></a> +<span class="sourceLineNo">455</span> // set default as 16M = (default hdfs block size) / 4;<a name="line.455"></a> +<span class="sourceLineNo">456</span> if (averageRegionSize < 16 * 1048576) {<a name="line.456"></a> +<span class="sourceLineNo">457</span> return splits;<a name="line.457"></a> +<span class="sourceLineNo">458</span> }<a name="line.458"></a> +<span class="sourceLineNo">459</span> for (int i = 0; i < splits.size(); i++) {<a name="line.459"></a> +<span class="sourceLineNo">460</span> TableSplit ts = (TableSplit) splits.get(i);<a name="line.460"></a> +<span class="sourceLineNo">461</span> TableName tableName = ts.getTable();<a name="line.461"></a> +<span class="sourceLineNo">462</span> String regionLocation = ts.getRegionLocation();<a name="line.462"></a> +<span class="sourceLineNo">463</span> String encodedRegionName = ts.getEncodedRegionName();<a name="line.463"></a> +<span class="sourceLineNo">464</span> long regionSize = ts.getLength();<a name="line.464"></a> +<span class="sourceLineNo">465</span><a name="line.465"></a> +<span class="sourceLineNo">466</span> if (regionSize >= averageRegionSize) {<a name="line.466"></a> +<span class="sourceLineNo">467</span> // make this region as multiple MapReduce input split.<a name="line.467"></a> +<span class="sourceLineNo">468</span> int n = (int) Math.round(Math.log(((double) regionSize) / ((double) averageRegionSize)) + 1.0);<a name="line.468"></a> +<span class="sourceLineNo">469</span> List<InputSplit> temp = createNInputSplitsUniform(ts, n);<a name="line.469"></a> +<span class="sourceLineNo">470</span> resultList.addAll(temp);<a name="line.470"></a> +<span class="sourceLineNo">471</span> } else {<a name="line.471"></a> +<span class="sourceLineNo">472</span> // if the total size of several small continuous regions less than the average region size,<a name="line.472"></a> +<span class="sourceLineNo">473</span> // combine them into one MapReduce input split.<a name="line.473"></a> +<span class="sourceLineNo">474</span> long totalSize = regionSize;<a name="line.474"></a> +<span class="sourceLineNo">475</span> byte[] splitStartKey = ts.getStartRow();<a name="line.475"></a> +<span class="sourceLineNo">476</span> byte[] splitEndKey = ts.getEndRow();<a name="line.476"></a> +<span class="sourceLineNo">477</span> int j = i + 1;<a name="line.477"></a> +<span class="sourceLineNo">478</span> while (j < splits.size()) {<a name="line.478"></a> +<span class="sourceLineNo">479</span> TableSplit nextRegion = (TableSplit) splits.get(j);<a name="line.479"></a> +<span class="sourceLineNo">480</span> long nextRegionSize = nextRegion.getLength();<a name="line.480"></a> +<span class="sourceLineNo">481</span> if (totalSize + nextRegionSize <= averageRegionSize) {<a name="line.481"></a> +<span class="sourceLineNo">482</span> totalSize = totalSize + nextRegionSize;<a name="line.482"></a> +<span class="sourceLineNo">483</span> splitEndKey = nextRegion.getEndRow();<a name="line.483"></a> +<span class="sourceLineNo">484</span> j++;<a name="line.484"></a> +<span class="sourceLineNo">485</span> } else {<a name="line.485"></a> +<span class="sourceLineNo">486</span> break;<a name="line.486"></a> +<span class="sourceLineNo">487</span> }<a name="line.487"></a> +<span class="sourceLineNo">488</span> }<a name="line.488"></a> +<span class="sourceLineNo">489</span> i = j - 1;<a name="line.489"></a> +<span class="sourceLineNo">490</span> TableSplit t = new TableSplit(tableName, scan, splitStartKey, splitEndKey, regionLocation,<a name="line.490"></a> +<span class="sourceLineNo">491</span> encodedRegionName, totalSize);<a name="line.491"></a> +<span class="sourceLineNo">492</span> resultList.add(t);<a name="line.492"></a> +<span class="sourceLineNo">493</span> }<a name="line.493"></a> +<span class="sourceLineNo">494</span> }<a name="line.494"></a> +<span class="sourceLineNo">495</span> return resultList;<a name="line.495"></a> +<span class="sourceLineNo">496</span> }<a name="line.496"></a> +<span class="sourceLineNo">497</span><a name="line.497"></a> +<span class="sourceLineNo">498</span> String reverseDNS(InetAddress ipAddress) throws UnknownHostException {<a name="line.498"></a> +<span class="sourceLineNo">499</span> String hostName = this.reverseDNSCacheMap.get(ipAddress);<a name="line.499"></a> +<span class="sourceLineNo">500</span> if (hostName == null) {<a name="line.500"></a> +<span class="sourceLineNo">501</span> String ipAddressString = null;<a name="line.501"></a> +<span class="sourceLineNo">502</span> try {<a name="line.502"></a> +<span class="sourceLineNo">503</span> ipAddressString = DNS.reverseDns(ipAddress, null);<a name="line.503"></a> +<span class="sourceLineNo">504</span> } catch (Exception e) {<a name="line.504"></a> +<span class="sourceLineNo">505</span> // We can use InetAddress in case the jndi failed to pull up the reverse DNS entry from the<a name="line.505"></a> +<span class="sourceLineNo">506</span> // name service. Also, in case of ipv6, we need to use the InetAddress since resolving<a name="line.506"></a> +<span class="sourceLineNo">507</span> // reverse DNS using jndi doesn't work well with ipv6 addresses.<a name="line.507"></a> +<span class="sourceLineNo">508</span> ipAddressString = InetAddress.getByName(ipAddress.getHostAddress()).getHostName();<a name="line.508"></a> +<span class="sourceLineNo">509</span> }<a name="line.509"></a> +<span class="sourceLineNo">510</span> if (ipAddressString == null) throw new UnknownHostException("No host found for " + ipAddress);<a name="line.510"></a> +<span class="sourceLineNo">511</span> hostName = Strings.domainNamePointerToHostName(ipAddressString);<a name="line.511"></a> +<span class="sourceLineNo">512</span> this.reverseDNSCacheMap.put(ipAddress, hostName);<a name="line.512"></a> +<span class="sourceLineNo">513</span> }<a name="line.513"></a> +<span class="sourceLineNo">514</span> return hostName;<a name="line.514"></a> +<span class="sourceLineNo">515</span> }<a name="line.515"></a> +<span class="sourceLineNo">516</span><a name="line.516"></a> +<span class="sourceLineNo">517</span> /**<a name="line.517"></a> +<span class="sourceLineNo">518</span> * Test if the given region is to be included in the InputSplit while splitting<a name="line.518"></a> +<span class="sourceLineNo">519</span> * the regions of a table.<a name="line.519"></a> +<span class="sourceLineNo">520</span> * <p><a name="line.520"></a> +<span class="sourceLineNo">521</span> * This optimization is effective when there is a specific reasoning to exclude an entire region from the M-R job,<a name="line.521"></a> +<span class="sourceLineNo">522</span> * (and hence, not contributing to the InputSplit), given the start and end keys of the same. <br><a name="line.522"></a> +<span class="sourceLineNo">523</span> * Useful when we need to remember the last-processed top record and revisit the [last, current) interval for M-R processing,<a name="line.523"></a> +<span class="sourceLineNo">524</span> * continuously. In addition to reducing InputSplits, reduces the load on the region server as well, due to the ordering of the keys.<a name="line.524"></a> <span class="sourceLineNo">525</span> * <br><a name="line.525"></a> -<span class="sourceLineNo">526</span> * Note: It is possible that <code>endKey.length() == 0 </code> , for the last (recent) region.<a name="line.526"></a> -<span class="sourceLineNo">527</span> * <br><a name="line.527"></a> -<span class="sourceLineNo">528</span> * Override this method, if you want to bulk exclude regions altogether from M-R. By default, no region is excluded( i.e. all regions are included).<a name="line.528"></a> -<span class="sourceLineNo">529</span> *<a name="line.529"></a> +<span class="sourceLineNo">526</span> * <br><a name="line.526"></a> +<span class="sourceLineNo">527</span> * Note: It is possible that <code>endKey.length() == 0 </code> , for the last (recent) region.<a name="line.527"></a> +<span class="sourceLineNo">528</span> * <br><a name="line.528"></a> +<span class="sourceLineNo">529</span> * Override this method, if you want to bulk exclude regions altogether from M-R. By default, no region is excluded( i.e. all regions are included).<a name="line.529"></a> <span class="sourceLineNo">530</span> *<a name="line.530"></a> -<span class="sourceLineNo">531</span> * @param startKey Start key of the region<a name="line.531"></a> -<span class="sourceLineNo">532</span> * @param endKey End key of the region<a name="line.532"></a> -<span class="sourceLineNo">533</span> * @return true, if this region needs to be included as part of the input (default).<a name="line.533"></a> -<span class="sourceLineNo">534</span> *<a name="line.534"></a> -<span class="sourceLineNo">535</span> */<a name="line.535"></a> -<span class="sourceLineNo">536</span> protected boolean includeRegionInSplit(final byte[] startKey, final byte [] endKey) {<a name="line.536"></a> -<span class="sourceLineNo">537</span> return true;<a name="line.537"></a> -<span class="sourceLineNo">538</span> }<a name="line.538"></a> -<span class="sourceLineNo">539</span><a name="line.539"></a> -<span class="sourceLineNo">540</span> /**<a name="line.540"></a> -<span class="sourceLineNo">541</span> * Allows subclasses to get the {@link RegionLocator}.<a name="line.541"></a> -<span class="sourceLineNo">542</span> */<a name="line.542"></a> -<span class="sourceLineNo">543</span> protected RegionLocator getRegionLocator() {<a name="line.543"></a> -<span class="sourceLineNo">544</span> if (regionLocator == null) {<a name="line.544"></a> -<span class="sourceLineNo">545</span> throw new IllegalStateException(NOT_INITIALIZED);<a name="line.545"></a> -<span class="sourceLineNo">546</span> }<a name="line.546"></a> -<span class="sourceLineNo">547</span> return regionLocator;<a name="line.547"></a> -<span class="sourceLineNo">548</span> }<a name="line.548"></a> -<span class="sourceLineNo">549</span><a name="line.549"></a> -<span class="sourceLineNo">550</span> /**<a name="line.550"></a> -<span class="sourceLineNo">551</span> * Allows subclasses to get the {@link Table}.<a name="line.551"></a> -<span class="sourceLineNo">552</span> */<a name="line.552"></a> -<span class="sourceLineNo">553</span> protected Table getTable() {<a name="line.553"></a> -<span class="sourceLineNo">554</span> if (table == null) {<a name="line.554"></a> -<span class="sourceLineNo">555</span> throw new IllegalStateException(NOT_INITIALIZED);<a name="line.555"></a> -<span class="sourceLineNo">556</span> }<a name="line.556"></a> -<span class="sourceLineNo">557</span> return table;<a name="line.557"></a> -<span class="sourceLineNo">558</span> }<a name="line.558"></a> -<span class="sourceLineNo">559</span><a name="line.559"></a> -<span class="sourceLineNo">560</span> /**<a name="line.560"></a> -<span class="sourceLineNo">561</span> * Allows subclasses to get the {@link Admin}.<a name="line.561"></a> -<span class="sourceLineNo">562</span> */<a name="line.562"></a> -<span class="sourceLineNo">563</span> protected Admin getAdmin() {<a name="line.563"></a> -<span class="sourceLineNo">564</span> if (admin == null) {<a name="line.564"></a> -<span class="sourceLineNo">565</span> throw new IllegalStateException(NOT_INITIALIZED);<a name="line.565"></a> -<span class="sourceLineNo">566</span> }<a name="line.566"></a> -<span class="sourceLineNo">567</span> return admin;<a name="line.567"></a> -<span class="sourceLineNo">568</span> }<a name="line.568"></a> -<span class="sourceLineNo">569</span><a name="line.569"></a> -<span class="sourceLineNo">570</span> /**<a name="line.570"></a> -<span class="sourceLineNo">571</span> * Allows subclasses to initialize the table information.<a name="line.571"></a> -<span class="sourceLineNo">572</span> *<a name="line.572"></a> -<span class="sourceLineNo">573</span> * @param connection The Connection to the HBase cluster. MUST be unmanaged. We will close.<a name="line.573"></a> -<span class="sourceLineNo">574</span> * @param tableName The {@link TableName} of the table to process.<a name="line.574"></a> -<span class="sourceLineNo">575</span> * @throws IOException<a name="line.575"></a> -<span class="sourceLineNo">576</span> */<a name="line.576"></a> -<span class="sourceLineNo">577</span> protected void initializeTable(Connection connection, TableName tableName) throws IOException {<a name="line.577"></a> -<span class="sourceLineNo">578</span> if (this.table != null || this.connection != null) {<a name="line.578"></a> -<span class="sourceLineNo">579</span> LOG.warn("initializeTable called multiple times. Overwriting connection and table " +<a name="line.579"></a> -<span class="sourceLineNo">580</span> "reference; TableInputFormatBase will not close these old references when done.");<a name="line.580"></a> -<span class="sourceLineNo">581</span> }<a name="line.581"></a> -<span class="sourceLineNo">582</span> this.table = connection.getTable(tableName);<a name="line.582"></a> -<span class="sourceLineNo">583</span> this.regionLocator = connection.getRegionLocator(tableName);<a name="line.583"></a> -<span class="sourceLineNo">584</span> this.admin = connection.getAdmin();<a name="line.584"></a> -<span class="sourceLineNo">585</span> this.connection = connection;<a name="line.585"></a> -<span class="sourceLineNo">586</span> }<a name="line.586"></a> -<span class="sourceLineNo">587</span><a name="line.587"></a> -<span class="sourceLineNo">588</span> /**<a name="line.588"></a> -<span class="sourceLineNo">589</span> * Gets the scan defining the actual details like columns etc.<a name="line.589"></a> -<span class="sourceLineNo">590</span> *<a name="line.590"></a> -<span class="sourceLineNo">591</span> * @return The internal scan instance.<a name="line.591"></a> -<span class="sourceLineNo">592</span> */<a name="line.592"></a> -<span class="sourceLineNo">593</span> public Scan getScan() {<a name="line.593"></a> -<span class="sourceLineNo">594</span> if (this.scan == null) this.scan = new Scan();<a name="line.594"></a> -<span class="sourceLineNo">595</span> return scan;<a name="line.595"></a> -<span class="sourceLineNo">596</span> }<a name="line.596"></a> -<span class="sourceLineNo">597</span><a name="line.597"></a> -<span class="sourceLineNo">598</span> /**<a name="line.598"></a> -<span class="sourceLineNo">599</span> * Sets the scan defining the actual details like columns etc.<a name="line.599"></a> -<span class="sourceLineNo">600</span> *<a name="line.600"></a> -<span class="sourceLineNo">601</span> * @param scan The scan to set.<a name="line.601"></a> -<span class="sourceLineNo">602</span> */<a name="line.602"></a> -<span class="sourceLineNo">603</span> public void setScan(Scan scan) {<a name="line.603"></a> -<span class="sourceLineNo">604</span> this.scan = scan;<a name="line.604"></a> -<span class="sourceLineNo">605</span> }<a name="line.605"></a> -<span class="sourceLineNo">606</span><a name="line.606"></a> -<span class="sourceLineNo">607</span> /**<a name="line.607"></a> -<span class="sourceLineNo">608</span> * Allows subclasses to set the {@link TableRecordReader}.<a name="line.608"></a> -<span class="sourceLineNo">609</span> *<a name="line.609"></a> -<span class="sourceLineNo">610</span> * @param tableRecordReader A different {@link TableRecordReader}<a name="line.610"></a> -<span class="sourceLineNo">611</span> * implementation.<a name="line.611"></a> -<span class="sourceLineNo">612</span> */<a name="line.612"></a> -<span class="sourceLineNo">613</span> protected void setTableRecordReader(TableRecordReader tableRecordReader) {<a name="line.613"></a> -<span class="sourceLineNo">614</span> this.tableRecordReader = tableRecordReader;<a name="line.614"></a> -<span class="sourceLineNo">615</span> }<a name="line.615"></a> -<span class="sourceLineNo">616</span><a name="line.616"></a> -<span class="sourceLineNo">617</span> /**<a name="line.617"></a> -<span class="sourceLineNo">618</span> * Handle subclass specific set up.<a name="line.618"></a> -<span class="sourceLineNo">619</span> * Each of the entry points used by the MapReduce framework,<a name="line.619"></a> -<span class="sourceLineNo">620</span> * {@link #createRecordReader(InputSplit, TaskAttemptContext)} and {@link #getSplits(JobContext)},<a name="line.620"></a> -<span class="sourceLineNo">621</span> * will call {@link #initialize(JobContext)} as a convenient centralized location to handle<a name="line.621"></a> -<span class="sourceLineNo">622</span> * retrieving the necessary configuration information and calling<a name="line.622"></a> -<span class="sourceLineNo">623</span> * {@link #initializeTable(Connection, TableName)}.<a name="line.623"></a> -<span class="sourceLineNo">624</span> *<a name="line.624"></a> -<span class="sourceLineNo">625</span> * Subclasses should implement their initialize call such that it is safe to call multiple times.<a name="line.625"></a> -<span class="sourceLineNo">626</span> * The current TableInputFormatBase implementation relies on a non-null table reference to decide<a name="line.626"></a> -<span class="sourceLineNo">627</span> * if an initialize call is needed, but this behavior may change in the future. In particular,<a name="line.627"></a> -<span class="sourceLineNo">628</span> * it is critical that initializeTable not be called multiple times since this will leak<a name="line.628"></a> -<span class="sourceLineNo">629</span> * Connection instances.<a name="line.629"></a> -<span class="sourceLineNo">630</span> *<a name="line.630"></a> -<span class="sourceLineNo">631</span> */<a name="line.631"></a> -<span class="sourceLineNo">632</span> protected void initialize(JobContext context) throws IOException {<a name="line.632"></a> -<span class="sourceLineNo">633</span> }<a name="line.633"></a> -<span class="sourceLineNo">634</span><a name="line.634"></a> -<span class="sourceLineNo">635</span> /**<a name="line.635"></a> -<span class="sourceLineNo">636</span> * Close the Table and related objects that were initialized via<a name="line.636"></a> -<span class="sourceLineNo">637</span> * {@link #initializeTable(Connection, TableName)}.<a name="line.637"></a> -<span class="sourceLineNo">638</span> *<a name="line.638"></a> -<span class="sourceLineNo">639</span> * @throws IOException<a name="line.639"></a> -<span class="sourceLineNo">640</span> */<a name="line.640"></a> -<span class="sourceLineNo">641</span> protected void closeTable() throws IOException {<a name="line.641"></a> -<span class="sourceLineNo">642</span> close(admin, table, regionLocator, connection);<a name="line.642"></a> -<span class="sourceLineNo">643</span> admin = null;<a name="line.643"></a> -<span class="sourceLineNo">644</span> table = null;<a name="line.644"></a> -<span class="sourceLineNo">645</span> regionLocator = null;<a name="line.645"></a> -<span class="sourceLineNo">646</span> connection = null;<a name="line.646"></a> -<span class="sourceLineNo">647</span> }<a name="line.647"></a> -<span class="sourceLineNo">648</span><a name="line.648"></a> -<span class="sourceLineNo">649</span> private void close(Closeable... closables) throws IOException {<a name="line.649"></a> -<span class="sourceLineNo">650</span> for (Closeable c : closables) {<a name="line.650"></a> -<span class="sourceLineNo">651</span> if(c != null) { c.close(); }<a name="line.651"></a> -<span class="sourceLineNo">652</span> }<a name="line.652"></a> -<span class="sourceLineNo">653</span> }<a name="line.653"></a> -<span class="sourceLineNo">654</span><a name="line.654"></a> -<span class="sourceLineNo">655</span>}<a name="line.655"></a> +<span class="sourceLineNo">531</span> *<a name="line.531"></a> +<span class="sourceLineNo">532</span> * @param startKey Start key of the region<a name="line.532"></a> +<span class="sourceLineNo">533</span> * @param endKey End key of the region<a name="line.533"></a> +<span class="sourceLineNo">534</span> * @return true, if this region needs to be included as part of the input (default).<a name="line.534"></a> +<span class="sourceLineNo">535</span> *<a name="line.535"></a> +<span class="sourceLineNo">536</span> */<a name="line.536"></a> +<span class="sourceLineNo">537</span> protected boolean includeRegionInSplit(final byte[] startKey, final byte [] endKey) {<a name="line.537"></a> +<span class="sourceLineNo">538</span> return true;<a name="line.538"></a> +<span class="sourceLineNo">539</span> }<a name="line.539"></a> +<span class="sourceLineNo">540</span><a name="line.540"></a> +<span class="sourceLineNo">541</span> /**<a name="line.541"></a> +<span class="sourceLineNo">542</span> * Allows subclasses to get the {@link RegionLocator}.<a name="line.542"></a> +<span class="sourceLineNo">543</span> */<a name="line.543"></a> +<span class="sourceLineNo">544</span> protected RegionLocator getRegionLocator() {<a name="line.544"></a> +<span class="sourceLineNo">545</span> if (regionLocator == null) {<a name="line.545"></a> +<span class="sourceLineNo">546</span> throw new IllegalStateException(NOT_INITIALIZED);<a name="line.546"></a> +<span class="sourceLineNo">547</span> }<a name="line.547"></a> +<span class="sourceLineNo">548</span> return regionLocator;<a name="line.548"></a> +<span class="sourceLineNo">549</span> }<a name="line.549"></a> +<span class="sourceLineNo">550</span><a name="line.550"></a> +<span class="sourceLineNo">551</span> /**<a name="line.551"></a> +<span class="sourceLineNo">552</span> * Allows subclasses to get the {@link Table}.<a name="line.552"></a> +<span class="sourceLineNo">553</span> */<a name="line.553"></a> +<span class="sourceLineNo">554</span> protected Table getTable() {<a name="line.554"></a> +<span class="sourceLineNo">555</span> if (table == null) {<a name="line.555"></a> +<span class="sourceLineNo">556</span> throw new IllegalStateException(NOT_INITIALIZED);<a name="line.556"></a> +<span class="sourceLineNo">557</span> }<a name="line.557"></a> +<span class="sourceLineNo">558</span> return table;<a name="line.558"></a> +<span class="sourceLineNo">559</span> }<a name="line.559"></a> +<span class="sourceLineNo">560</span><a name="line.560"></a> +<span class="sourceLineNo">561</span> /**<a name="line.561"></a> +<span class="sourceLineNo">562</span> * Allows subclasses to get the {@link Admin}.<a name="line.562"></a> +<span class="sourceLineNo">563</span> */<a name="line.563"></a> +<span class="sourceLineNo">564</span> protected Admin getAdmin() {<a name="line.564"></a> +<span class="sourceLineNo">565</span> if (admin == null) {<a name="line.565"></a> +<span class="sourceLineNo">566</span> throw new IllegalStateException(NOT_INITIALIZED);<a name="line.566"></a> +<span class="sourceLineNo">567</span> }<a name="line.567"></a> +<span class="sourceLineNo">568</span> return admin;<a name="line.568"></a> +<span class="sourceLineNo">569</span> }<a name="line.569"></a> +<span class="sourceLineNo">570</span><a name="line.570"></a> +<span class="sourceLineNo">571</span> /**<a name="line.571"></a> +<span class="sourceLineNo">572</span> * Allows subclasses to initialize the table information.<a name="line.572"></a> +<span class="sourceLineNo">573</span> *<a name="line.573"></a> +<span class="sourceLineNo">574</span> * @param connection The Connection to the HBase cluster. MUST be unmanaged. We will close.<a name="line.574"></a> +<span class="sourceLineNo">575</span> * @param tableName The {@link TableName} of the table to process.<a name="line.575"></a> +<span class="sourceLineNo">576</span> * @throws IOException<a name="line.576"></a> +<span class="sourceLineNo">577</span> */<a name="line.577"></a> +<span class="sourceLineNo">578</span> protected void initializeTable(Connection connection, TableName tableName) throws IOException {<a name="line.578"></a> +<span class="sourceLineNo">579</span> if (this.table != null || this.connection != null) {<a name="line.579"></a> +<span class="sourceLineNo">580</span> LOG.warn("initializeTable called multiple times. Overwriting connection and table " +<a name="line.580"></a> +<span class="sourceLineNo">581</span> "reference; TableInputFormatBase will not close these old references when done.");<a name="line.581"></a> +<span class="sourceLineNo">582</span> }<a name="line.582"></a> +<span class="sourceLineNo">583</span> this.table = connection.getTable(tableName);<a name="line.583"></a> +<span class="sourceLineNo">584</span> this.regionLocator = connection.getRegionLocator(tableName);<a name="line.584"></a> +<span class="sourceLineNo">585</span> this.admin = connection.getAdmin();<a name="line.585"></a> +<span class="sourceLineNo">586</span> this.connection = connection;<a name="line.586"></a> +<span class="sourceLineNo">587</span> }<a name="line.587"></a> +<span class="sourceLineNo">588</span><a name="line.588"></a> +<span class="sourceLineNo">589</span> /**<a name="line.589"></a> +<span class="sourceLineNo">590</span> * Gets the scan defining the actual details like columns etc.<a name="line.590"></a> +<span class="sourceLineNo">591</span> *<a name="line.591"></a> +<span class="sourceLineNo">592</span> * @return The internal scan instance.<a name="line.592"></a> +<span class="sourceLineNo">593</span> */<a name="line.593"></a> +<span class="sourceLineNo">594</span> public Scan getScan() {<a name="line.594"></a> +<span class="sourceLineNo">595</span> if (this.scan == null) this.scan = new Scan();<a name="line.595"></a> +<span class="sourceLineNo">596</span> return scan;<a name="line.596"></a> +<span class="sourceLineNo">597</span> }<a name="line.597"></a> +<span class="sourceLineNo">598</span><a name="line.598"></a> +<span class="sourceLineNo">599</span> /**<a name="line.599"></a> +<span class="sourceLineNo">600</span> * Sets the scan defining the actual details like columns etc.<a name="line.600"></a> +<span class="sourceLineNo">601</span> *<a name="line.601"></a> +<span class="sourceLineNo">602</span> * @param scan The scan to set.<a name="line.602"></a> +<span class="sourceLineNo">603</span> */<a name="line.603"></a> +<span class="sourceLineNo">604</span> public void setScan(Scan scan) {<a name="line.604"></a> +<span class="sourceLineNo">605</span> this.scan = scan;<a name="line.605"></a> +<span class="sourceLineNo">606</span> }<a name="line.606"></a> +<span class="sourceLineNo">607</span><a name="line.607"></a> +<span class="sourceLineNo">608</span> /**<a name="line.608"></a> +<span class="sourceLineNo">609</span> * Allows subclasses to set the {@link TableRecordReader}.<a name="line.609"></a> +<span class="sourceLineNo">610</span> *<a name="line.610"></a> +<span class="sourceLineNo">611</span> * @param tableRecordReader A different {@link TableRecordReader}<a name="line.611"></a> +<span class="sourceLineNo">612</span> * implementation.<a name="line.612"></a> +<span class="sourceLineNo">613</span> */<a name="line.613"></a> +<span class="sourceLineNo">614</span> protected void setTableRecordReader(TableRecordReader tableRecordReader) {<a name="line.614"></a> +<span class="sourceLineNo">615</span> this.tableRecordReader = tableRecordReader;<a name="line.615"></a> +<span class="sourceLineNo">616</span> }<a name="line.616"></a> +<span class="sourceLineNo">617</span><a name="line.617"></a> +<span class="sourceLineNo">618</span> /**<a name="line.618"></a> +<span class="sourceLineNo">619</span> * Handle subclass specific set up.<a name="line.619"></a> +<span class="sourceLineNo">620</span> * Each of the entry points used by the MapReduce framework,<a name="line.620"></a> +<span class="sourceLineNo">621</span> * {@link #createRecordReader(InputSplit, TaskAttemptContext)} and {@link #getSplits(JobContext)},<a name="line.621"></a> +<span class="sourceLineNo">622</span> * will call {@link #initialize(JobContext)} as a convenient centralized location to handle<a name="line.622"></a> +<span class="sourceLineNo">623</span> * retrieving the necessary configuration information and calling<a name="line.623"></a> +<span class="sourceLineNo">624</span> * {@link #initializeTable(Connection, TableName)}.<a name="line.624"></a> +<span class="sourceLineNo">625</span> *<a name="line.625"></a> +<span class="sourceLineNo">626</span> * Subclasses should implement their initialize call such that it is safe to call multiple times.<a name="line.626"></a> +<span class="sourceLineNo">627</span> * The current TableInputFormatBase implementation relies on a non-null table reference to decide<a name="line.627"></a> +<span class="sourceLineNo">628</span> * if an initialize call is needed, but this behavior may change in the future. In particular,<a name="line.628"></a> +<span class="sourceLineNo">629</span> * it is critical that initializeTable not be called multiple times since this will leak<a name="line.629"></a> +<span class="sourceLineNo">630</span> * Connection instances.<a name="line.630"></a> +<span class="sourceLineNo">631</span> *<a name="line.631"></a> +<span class="sourceLineNo">632</span> */<a name="line.632"></a> +<span class="sourceLineNo">633</span> protected void initialize(JobContext context) throws IOException {<a name="line.633"></a> +<span class="sourceLineNo">634</span> }<a name="line.634"></a> +<span class="sourceLineNo">635</span><a name="line.635"></a> +<span class="sourceLineNo">636</span> /**<a name="line.636"></a> +<span class="sourceLineNo">637</span> * Close the Table and related objects that were initialized via<a name="line.637"></a> +<span class="sourceLineNo">638</span> * {@link #initializeTable(Connection, TableName)}.<a name="line.638"></a> +<span class="sourceLineNo">639</span> *<a name="line.639"></a> +<span class="sourceLineNo">640</span> * @throws IOException<a name="line.640"></a> +<span class="sourceLineNo">641</span> */<a name="line.641"></a> +<span class="sourceLineNo">642</span> protected void closeTable() throws IOException {<a name="line.642"></a> +<span class="sourceLineNo">643</span> close(admin, table, regionLocator, connection);<a name="line.643"></a> +<span class="sourceLineNo">644</span> admin = null;<a name="line.644"></a> +<span class="sourceLineNo">645</span> table = null;<a name="line.645"></a> +<span class="sourceLineNo">646</span> regionLocator = null;<a name="line.646"></a> +<span class="sourceLineNo">647</span> connection = null;<a name="line.647"></a> +<span class="sourceLineNo">648</span> }<a name="line.648"></a> +<span class="sourceLineNo">649</span><a name="line.649"></a> +<span class="sourceLineNo">650</span> private void close(Closeable... closables) throws IOException {<a name="line.650"></a> +<span class="sourceLineNo">651</span> for (Closeable c : closables) {<a name="line.651"></a> +<span class="sourceLineNo">652</span> if(c != null) { c.close(); }<a name="line.652"></a> +<span class="sourceLineNo">653</span> }<a name="line.653"></a> +<span class="sourceLineNo">654</span> }<a name="line.654"></a> +<span class="sourceLineNo">655</span><a name="line.655"></a> +<span class="sourceLineNo">656</span>}<a name="line.656"></a> http://git-wip-us.apache.org/repos/asf/hbase-site/blob/2b3f2bee/book.html ---------------------------------------------------------------------- diff --git a/book.html b/book.html index ff26207..70fd530 100644 --- a/book.html +++ b/book.html @@ -3638,7 +3638,7 @@ Some configurations would only appear in source code; the only way to identify t <dd> <div class="paragraph"> <div class="title">Description</div> -<p>The HFile format version to use for new files. Version 3 adds support for tags in hfiles (See <a href="http://hbase.apache.org/book.html#hbase.tags"; class="bare">http://hbase.apache.org/book.html#hbase.tags</a>). Distributed Log Replay requires that tags are enabled. Also see the configuration 'hbase.replication.rpc.codec'.</p> +<p>The HFile format version to use for new files. Version 3 adds support for tags in hfiles (See <a href="http://hbase.apache.org/book.html#hbase.tags"; class="bare">http://hbase.apache.org/book.html#hbase.tags</a>). Also see the configuration 'hbase.replication.rpc.codec'.</p> </div> <div class="paragraph"> <div class="title">Default</div> @@ -4883,7 +4883,7 @@ Some configurations would only appear in source code; the only way to identify t <dd> <div class="paragraph"> <div class="title">Description</div> -<p>By default, in replication we can not make sure the order of operations in slave cluster is same as the order in master. If set REPLICATION_SCOPE to 2, we will push edits by the order of written. This configure is to set how long (in ms) we will wait before next checking if a log can not push right now because there are some logs written before it have not been pushed. A larger waiting will decrease the number of queries on hbase:meta but will enlarge the delay of replication. This feature relies on zk-less assignment, and conflicts with distributed log replay. So users must set hbase.assignment.usezk and hbase.master.distributed.log.replay to false to support it.</p> +<p>By default, in replication we can not make sure the order of operations in slave cluster is same as the order in master. If set REPLICATION_SCOPE to 2, we will push edits by the order of written. This configure is to set how long (in ms) we will wait before next checking if a log can not push right now because there are some logs written before it have not been pushed. A larger waiting will decrease the number of queries on hbase:meta but will enlarge the delay of replication. This feature relies on zk-less assignment, so users must set hbase.assignment.usezk to false to support it.</p> </div> <div class="paragraph"> <div class="title">Default</div> @@ -6421,10 +6421,6 @@ ports.</p> <div class="title">If you have your own customer filters.</div> <p>See the release notes on the issue <a href="https://issues.apache.org/jira/browse/HBASE-12068";>HBASE-12068 [Branch-1] Avoid need to always do KeyValueUtil#ensureKeyValue for Filter transformCell</a>; be sure to follow the recommendations therein.</p> </div> -<div id="dlr" class="paragraph"> -<div class="title">Distributed Log Replay</div> -<p><a href="#distributed.log.replay">Distributed Log Replay</a> is off by default in HBase 1.0.0. Enabling it can make a big difference improving HBase MTTR. Enable this feature if you are doing a clean stop/start when you are upgrading. You cannot rolling upgrade to this feature (caveat if you are running on a version of HBase in excess of HBase 0.98.4 — see <a href="https://issues.apache.org/jira/browse/HBASE-12577";>HBASE-12577 Disable distributed log replay by default</a> for more).</p> -</div> <div class="paragraph"> <div class="title">Mismatch Of <code>hbase.client.scanner.max.result.size</code> Between Client and Server</div> <p>If either the client or server version is lower than 0.98.11/1.0.0 and the server @@ -14834,41 +14830,6 @@ If none are found, it throws an exception so that the log splitting can be retri </li> </ol> </div> -<div class="sect5"> -<h6 id="distributed.log.replay"><a class="anchor" href="#distributed.log.replay"></a>Distributed Log Replay</h6> -<div class="paragraph"> -<p>After a RegionServer fails, its failed regions are assigned to another RegionServer, which are marked as "recovering" in ZooKeeper. -A split log worker directly replays edits from the WAL of the failed RegionServer to the regions at its new location. -When a region is in "recovering" state, it can accept writes but no reads (including Append and Increment), region splits or merges.</p> -</div> -<div class="paragraph"> -<p>Distributed Log Replay extends the <a href="#distributed.log.splitting">Enabling or Disabling Distributed Log Splitting</a> framework. -It works by directly replaying WAL edits to another RegionServer instead of creating <em>recovered.edits</em> files. -It provides the following advantages over distributed log splitting alone:</p> -</div> -<div class="ulist"> -<ul> -<li> -<p>It eliminates the overhead of writing and reading a large number of <em>recovered.edits</em> files. -It is not unusual for thousands of <em>recovered.edits</em> files to be created and written concurrently during a RegionServer recovery. -Many small random writes can degrade overall system performance.</p> -</li> -<li> -<p>It allows writes even when a region is in recovering state. -It only takes seconds for a recovering region to accept writes again.</p> -</li> -</ul> -</div> -<div class="paragraph"> -<div class="title">Enabling Distributed Log Replay</div> -<p>To enable distributed log replay, set <code>hbase.master.distributed.log.replay</code> to <code>true</code>. -This will be the default for HBase 0.99 (<a href="https://issues.apache.org/jira/browse/HBASE-10888";>HBASE-10888</a>).</p> -</div> -<div class="paragraph"> -<p>You must also enable HFile version 3 (which is the default HFile format starting in HBase 0.99. -See <a href="https://issues.apache.org/jira/browse/HBASE-10855";>HBASE-10855</a>). Distributed log replay is unsafe for rolling upgrades.</p> -</div> -</div> </div> </div> <div class="sect3"> @@ -35484,7 +35445,7 @@ The server will return cellblocks compressed using this same compressor as long <div id="footer"> <div id="footer-text"> Version 3.0.0-SNAPSHOT<br> -Last updated 2017-11-07 14:29:34 UTC +Last updated 2017-11-09 14:29:37 UTC </div> </div> </body>
