View Javadoc

1   /**
2    * Licensed to the Apache Software Foundation (ASF) under one
3    * or more contributor license agreements.  See the NOTICE file
4    * distributed with this work for additional information
5    * regarding copyright ownership.  The ASF licenses this file
6    * to you under the Apache License, Version 2.0 (the
7    * "License"); you may not use this file except in compliance
8    * with the License.  You may obtain a copy of the License at
9    *
10   *     http://www.apache.org/licenses/LICENSE-2.0
11   *
12   * Unless required by applicable law or agreed to in writing, software
13   * distributed under the License is distributed on an "AS IS" BASIS,
14   * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15   * See the License for the specific language governing permissions and
16   * limitations under the License.
17   */
18  
19  package org.apache.hadoop.hbase;
20  
21  import java.security.InvalidParameterException;
22  import java.util.Map;
23  import java.util.Set;
24  import java.util.TreeMap;
25  import java.util.concurrent.atomic.AtomicLong;
26  
27  import org.apache.commons.logging.Log;
28  import org.apache.commons.logging.LogFactory;
29  import org.apache.hadoop.conf.Configuration;
30  import org.apache.hadoop.hbase.client.HTable;
31  import org.apache.hadoop.hbase.client.Result;
32  import org.apache.hadoop.hbase.client.ResultScanner;
33  import org.apache.hadoop.hbase.client.Scan;
34  import org.apache.hadoop.hbase.filter.CompareFilter;
35  import org.apache.hadoop.hbase.filter.Filter;
36  import org.apache.hadoop.hbase.filter.SingleColumnValueFilter;
37  import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding;
38  import org.apache.hadoop.hbase.util.Bytes;
39  import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
40  import org.apache.hadoop.hbase.util.MultiThreadedWriter;
41  import org.apache.hadoop.hbase.util.RegionSplitter;
42  import org.apache.hadoop.hbase.util.test.LoadTestDataGenerator;
43  import org.apache.hadoop.hbase.util.test.LoadTestKVGenerator;
44  import org.junit.After;
45  import org.junit.Assert;
46  import org.junit.Before;
47  import org.junit.Test;
48  import org.junit.experimental.categories.Category;
49  
50  /**
51   * Integration test that verifies lazy CF loading during scans by doing repeated scans
52   * with this feature while multiple threads are continuously writing values; and
53   * verifying the result.
54   */
55  @Category(IntegrationTests.class)
56  public class IntegrationTestLazyCfLoading {
57    private static final TableName TABLE_NAME =
58        TableName.valueOf(IntegrationTestLazyCfLoading.class.getSimpleName());
59    private static final String TIMEOUT_KEY = "hbase.%s.timeout";
60    private static final String ENCODING_KEY = "hbase.%s.datablock.encoding";
61  
62    /** A soft test timeout; duration of the test, as such, depends on number of keys to put. */
63    private static final int DEFAULT_TIMEOUT_MINUTES = 10;
64  
65    private static final int NUM_SERVERS = 1;
66    /** Set regions per server low to ensure splits happen during test */
67    private static final int REGIONS_PER_SERVER = 3;
68    private static final int KEYS_TO_WRITE_PER_SERVER = 20000;
69    private static final int WRITER_THREADS = 10;
70    private static final int WAIT_BETWEEN_SCANS_MS = 1000;
71  
72    private static final Log LOG = LogFactory.getLog(IntegrationTestLazyCfLoading.class);
73    private IntegrationTestingUtility util = new IntegrationTestingUtility();
74    private final DataGenerator dataGen = new DataGenerator();
75  
76    /** Custom LoadTestDataGenerator. Uses key generation and verification from
77     * LoadTestKVGenerator. Creates 3 column families; one with an integer column to
78     * filter on, the 2nd one with an integer column that matches the first integer column (for
79     * test-specific verification), and byte[] value that is used for general verification; and
80     * the third one with just the value.
81     */
82    private static class DataGenerator extends LoadTestDataGenerator {
83      private static final int MIN_DATA_SIZE = 4096;
84      private static final int MAX_DATA_SIZE = 65536;
85      public static final byte[] ESSENTIAL_CF = Bytes.toBytes("essential");
86      public static final byte[] JOINED_CF1 = Bytes.toBytes("joined");
87      public static final byte[] JOINED_CF2 = Bytes.toBytes("joined2");
88      public static final byte[] FILTER_COLUMN = Bytes.toBytes("filter");
89      public static final byte[] VALUE_COLUMN = Bytes.toBytes("val");
90      public static final long ACCEPTED_VALUE = 1L;
91  
92      private static final Map<byte[], byte[][]> columnMap = new TreeMap<byte[], byte[][]>(
93          Bytes.BYTES_COMPARATOR);
94  
95      private final AtomicLong expectedNumberOfKeys = new AtomicLong(0);
96      private final AtomicLong totalNumberOfKeys = new AtomicLong(0);
97  
98      public DataGenerator() {
99        super(MIN_DATA_SIZE, MAX_DATA_SIZE);
100       columnMap.put(ESSENTIAL_CF, new byte[][] { FILTER_COLUMN });
101       columnMap.put(JOINED_CF1, new byte[][] { FILTER_COLUMN, VALUE_COLUMN });
102       columnMap.put(JOINED_CF2, new byte[][] { VALUE_COLUMN });
103     }
104 
105     public long getExpectedNumberOfKeys() {
106       return expectedNumberOfKeys.get();
107     }
108 
109     public long getTotalNumberOfKeys() {
110       return totalNumberOfKeys.get();
111     }
112 
113     @Override
114     public byte[] getDeterministicUniqueKey(long keyBase) {
115       return LoadTestKVGenerator.md5PrefixedKey(keyBase).getBytes();
116     }
117 
118     @Override
119     public byte[][] getColumnFamilies() {
120       return columnMap.keySet().toArray(new byte[columnMap.size()][]);
121     }
122 
123     @Override
124     public byte[][] generateColumnsForCf(byte[] rowKey, byte[] cf) {
125       return columnMap.get(cf);
126     }
127 
128     @Override
129     public byte[] generateValue(byte[] rowKey, byte[] cf, byte[] column) {
130       if (Bytes.BYTES_COMPARATOR.compare(column, FILTER_COLUMN) == 0) {
131         // Random deterministic way to make some values "on" and others "off" for filters.
132         long value = Long.parseLong(Bytes.toString(rowKey, 0, 4), 16) & ACCEPTED_VALUE;
133         if (Bytes.BYTES_COMPARATOR.compare(cf, ESSENTIAL_CF) == 0) {
134           totalNumberOfKeys.incrementAndGet();
135           if (value == ACCEPTED_VALUE) {
136             expectedNumberOfKeys.incrementAndGet();
137           }
138         }
139         return Bytes.toBytes(value);
140       } else if (Bytes.BYTES_COMPARATOR.compare(column, VALUE_COLUMN) == 0) {
141         return kvGenerator.generateRandomSizeValue(rowKey, cf, column);
142       }
143       String error = "Unknown column " + Bytes.toString(column);
144       assert false : error;
145       throw new InvalidParameterException(error);
146     }
147 
148     @Override
149     public boolean verify(byte[] rowKey, byte[] cf, byte[] column, byte[] value) {
150       if (Bytes.BYTES_COMPARATOR.compare(column, FILTER_COLUMN) == 0) {
151         // Relies on the filter from getScanFilter being used.
152         return Bytes.toLong(value) == ACCEPTED_VALUE;
153       } else if (Bytes.BYTES_COMPARATOR.compare(column, VALUE_COLUMN) == 0) {
154         return LoadTestKVGenerator.verify(value, rowKey, cf, column);
155       }
156       return false; // some bogus value from read, we don't expect any such thing.
157     }
158 
159     @Override
160     public boolean verify(byte[] rowKey, byte[] cf, Set<byte[]> columnSet) {
161       return columnMap.get(cf).length == columnSet.size();
162     }
163 
164     public Filter getScanFilter() {
165       SingleColumnValueFilter scf = new SingleColumnValueFilter(ESSENTIAL_CF, FILTER_COLUMN,
166           CompareFilter.CompareOp.EQUAL, Bytes.toBytes(ACCEPTED_VALUE));
167       scf.setFilterIfMissing(true);
168       return scf;
169     }
170   }
171 
172   @Before
173   public void setUp() throws Exception {
174     LOG.info("Initializing cluster with " + NUM_SERVERS + " servers");
175     util.initializeCluster(NUM_SERVERS);
176     LOG.info("Done initializing cluster");
177     createTable();
178     // after table creation, ACLs need time to be propagated to RSs in a secure deployment
179     // so we sleep a little bit because we don't have a good way to know when permissions
180     // are received by RSs
181     Thread.sleep(3000);
182   }
183 
184   private void createTable() throws Exception {
185     deleteTable();
186     LOG.info("Creating table");
187     Configuration conf = util.getConfiguration();
188     String encodingKey = String.format(ENCODING_KEY, this.getClass().getSimpleName());
189     DataBlockEncoding blockEncoding = DataBlockEncoding.valueOf(conf.get(encodingKey, "FAST_DIFF"));
190     HTableDescriptor htd = new HTableDescriptor(TABLE_NAME);
191     for (byte[] cf : dataGen.getColumnFamilies()) {
192       HColumnDescriptor hcd = new HColumnDescriptor(cf);
193       hcd.setDataBlockEncoding(blockEncoding);
194       htd.addFamily(hcd);
195     }
196     int serverCount = util.getHBaseClusterInterface().getClusterStatus().getServersSize();
197     byte[][] splits = new RegionSplitter.HexStringSplit().split(serverCount * REGIONS_PER_SERVER);
198     util.getHBaseAdmin().createTable(htd, splits);
199     LOG.info("Created table");
200   }
201 
202   private void deleteTable() throws Exception {
203     if (util.getHBaseAdmin().tableExists(TABLE_NAME)) {
204       LOG.info("Deleting table");
205       util.deleteTable(TABLE_NAME);
206       LOG.info("Deleted table");
207     }
208   }
209 
210   @After
211   public void tearDown() throws Exception {
212     deleteTable();
213     LOG.info("Restoring the cluster");
214     util.restoreCluster();
215     LOG.info("Done restoring the cluster");
216   }
217 
218   @Test
219   public void testReadersAndWriters() throws Exception {
220     Configuration conf = util.getConfiguration();
221     String timeoutKey = String.format(TIMEOUT_KEY, this.getClass().getSimpleName());
222     long maxRuntime = conf.getLong(timeoutKey, DEFAULT_TIMEOUT_MINUTES);
223     long serverCount = util.getHBaseClusterInterface().getClusterStatus().getServersSize();
224     long keysToWrite = serverCount * KEYS_TO_WRITE_PER_SERVER;
225     HTable table = new HTable(conf, TABLE_NAME);
226 
227     // Create multi-threaded writer and start it. We write multiple columns/CFs and verify
228     // their integrity, therefore multi-put is necessary.
229     MultiThreadedWriter writer =
230       new MultiThreadedWriter(dataGen, conf, TABLE_NAME);
231     writer.setMultiPut(true);
232 
233     LOG.info("Starting writer; the number of keys to write is " + keysToWrite);
234     // TODO : Need to see if tag support has to be given here in the integration test suite
235     writer.start(1, keysToWrite, WRITER_THREADS);
236 
237     // Now, do scans.
238     long now = EnvironmentEdgeManager.currentTimeMillis();
239     long timeLimit = now + (maxRuntime * 60000);
240     boolean isWriterDone = false;
241     while (now < timeLimit && !isWriterDone) {
242       LOG.info("Starting the scan; wrote approximately "
243         + dataGen.getTotalNumberOfKeys() + " keys");
244       isWriterDone = writer.isDone();
245       if (isWriterDone) {
246         LOG.info("Scanning full result, writer is done");
247       }
248       Scan scan = new Scan();
249       for (byte[] cf : dataGen.getColumnFamilies()) {
250         scan.addFamily(cf);
251       }
252       scan.setFilter(dataGen.getScanFilter());
253       scan.setLoadColumnFamiliesOnDemand(true);
254       // The number of keys we can expect from scan - lower bound (before scan).
255       // Not a strict lower bound - writer knows nothing about filters, so we report
256       // this from generator. Writer might have generated the value but not put it yet.
257       long onesGennedBeforeScan = dataGen.getExpectedNumberOfKeys();
258       long startTs = EnvironmentEdgeManager.currentTimeMillis();
259       ResultScanner results = table.getScanner(scan);
260       long resultCount = 0;
261       Result result = null;
262       // Verify and count the results.
263       while ((result = results.next()) != null) {
264         boolean isOk = writer.verifyResultAgainstDataGenerator(result, true, true);
265         Assert.assertTrue("Failed to verify [" + Bytes.toString(result.getRow())+ "]", isOk);
266         ++resultCount;
267       }
268       long timeTaken = EnvironmentEdgeManager.currentTimeMillis() - startTs;
269       // Verify the result count.
270       long onesGennedAfterScan = dataGen.getExpectedNumberOfKeys();
271       Assert.assertTrue("Read " + resultCount + " keys when at most " + onesGennedAfterScan
272         + " were generated ", onesGennedAfterScan >= resultCount);
273       if (isWriterDone) {
274         Assert.assertTrue("Read " + resultCount + " keys; the writer is done and "
275           + onesGennedAfterScan + " keys were generated", onesGennedAfterScan == resultCount);
276       } else if (onesGennedBeforeScan * 0.9 > resultCount) {
277         LOG.warn("Read way too few keys (" + resultCount + "/" + onesGennedBeforeScan
278           + ") - there might be a problem, or the writer might just be slow");
279       }
280       LOG.info("Scan took " + timeTaken + "ms");
281       if (!isWriterDone) {
282         Thread.sleep(WAIT_BETWEEN_SCANS_MS);
283         now = EnvironmentEdgeManager.currentTimeMillis();
284       }
285     }
286     Assert.assertEquals("There are write failures", 0, writer.getNumWriteFailures());
287     Assert.assertTrue("Writer is not done", isWriterDone);
288     // Assert.fail("Boom!");
289   }
290 }