Details
-
Bug
-
Status: Closed
-
Critical
-
Resolution: Fixed
-
4.7.0
-
None
-
None
Description
When I join a salted table (which has been split after creation) with another table in my business system ,I meet following error,even though I clear the salted table 's TableRegionCache:
org.apache.phoenix.exception.PhoenixIOException: org.apache.hadoop.hbase.DoNotRetryIOException: Could not find hash cache for joinId: %�����2. The cache might have expired and have been removed. at org.apache.phoenix.coprocessor.HashJoinRegionScanner.<init>(HashJoinRegionScanner.java:98) at org.apache.phoenix.coprocessor.ScanRegionObserver.doPostScannerOpen(ScanRegionObserver.java:218) at org.apache.phoenix.coprocessor.BaseScannerRegionObserver.postScannerOpen(BaseScannerRegionObserver.java:201) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost$52.call(RegionCoprocessorHost.java:1203) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost$RegionOperation.call(RegionCoprocessorHost.java:1517) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost.execOperation(RegionCoprocessorHost.java:1592) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost.execOperationWithResult(RegionCoprocessorHost.java:1556) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost.postScannerOpen(RegionCoprocessorHost.java:1198) at org.apache.hadoop.hbase.regionserver.HRegionServer.scan(HRegionServer.java:3178) at org.apache.hadoop.hbase.protobuf.generated.ClientProtos$ClientService$2.callBlockingMethod(ClientProtos.java:29925) at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2031) at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108) at org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:116) at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:96) at java.lang.Thread.run(Thread.java:745) at org.apache.phoenix.util.ServerUtil.parseServerException(ServerUtil.java:111) at org.apache.phoenix.iterate.TableResultIterator.initScanner(TableResultIterator.java:127) at org.apache.phoenix.iterate.ParallelIterators$1.call(ParallelIterators.java:108) at org.apache.phoenix.iterate.ParallelIterators$1.call(ParallelIterators.java:103) at java.util.concurrent.FutureTask.run(FutureTask.java:262) at org.apache.phoenix.job.JobManager$InstrumentedJobFutureTask.run(JobManager.java:183) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) at java.lang.Thread.run(Thread.java:745) Caused by: org.apache.hadoop.hbase.DoNotRetryIOException: org.apache.hadoop.hbase.DoNotRetryIOException: Could not find hash cache for joinId: %�����2. The cache might have expired and have been removed. at org.apache.phoenix.coprocessor.HashJoinRegionScanner.<init>(HashJoinRegionScanner.java:98) at org.apache.phoenix.coprocessor.ScanRegionObserver.doPostScannerOpen(ScanRegionObserver.java:218) at org.apache.phoenix.coprocessor.BaseScannerRegionObserver.postScannerOpen(BaseScannerRegionObserver.java:201) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost$52.call(RegionCoprocessorHost.java:1203) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost$RegionOperation.call(RegionCoprocessorHost.java:1517) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost.execOperation(RegionCoprocessorHost.java:1592) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost.execOperationWithResult(RegionCoprocessorHost.java:1556) at org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost.postScannerOpen(RegionCoprocessorHost.java:1198) at org.apache.hadoop.hbase.regionserver.HRegionServer.scan(HRegionServer.java:3178) at org.apache.hadoop.hbase.protobuf.generated.ClientProtos$ClientService$2.callBlockingMethod(ClientProtos.java:29925) at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2031) at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108) at org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:116) at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:96) at java.lang.Thread.run(Thread.java:745) at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method) at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:57) at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45) at java.lang.reflect.Constructor.newInstance(Constructor.java:526) at org.apache.hadoop.ipc.RemoteException.instantiateException(RemoteException.java:106) at org.apache.hadoop.ipc.RemoteException.unwrapRemoteException(RemoteException.java:95) at org.apache.hadoop.hbase.protobuf.ProtobufUtil.getRemoteException(ProtobufUtil.java:304) at org.apache.hadoop.hbase.client.ScannerCallable.openScanner(ScannerCallable.java:316) at org.apache.hadoop.hbase.client.ScannerCallable.call(ScannerCallable.java:164) at org.apache.hadoop.hbase.client.ScannerCallable.call(ScannerCallable.java:59) at org.apache.hadoop.hbase.client.RpcRetryingCaller.callWithRetries(RpcRetryingCaller.java:114) at org.apache.hadoop.hbase.client.RpcRetryingCaller.callWithRetries(RpcRetryingCaller.java:90) at org.apache.hadoop.hbase.client.ClientScanner.nextScanner(ClientScanner.java:283) at org.apache.hadoop.hbase.client.ClientScanner.initializeScannerInConstruction(ClientScanner.java:188) at org.apache.hadoop.hbase.client.ClientScanner.<init>(ClientScanner.java:183) at org.apache.hadoop.hbase.client.ClientScanner.<init>(ClientScanner.java:110) at org.apache.hadoop.hbase.client.HTable.getScanner(HTable.java:739) at org.apache.phoenix.iterate.TableResultIterator.initScanner(TableResultIterator.java:123) ... 7 more
I write a unit test to reproduce this error, just as follows:
public void testSaltTableJoinError() throws Exception { //1.create LHS SALT_TEST salted table this.jdbcTemplate.update("drop table if exists SALT_TEST "); this.jdbcTemplate.update( "create table SALT_TEST"+ "("+ "id UNSIGNED_INT not null primary key,"+ "appId VARCHAR"+ ")SALT_BUCKETS=2"); this.jdbcTemplate.update("upsert into SALT_TEST(id,appId) values(1,'app1')"); this.jdbcTemplate.update("upsert into SALT_TEST(id,appId) values(2,'app2')"); this.jdbcTemplate.update("upsert into SALT_TEST(id,appId) values(3,'app3')"); this.jdbcTemplate.update("upsert into SALT_TEST(id,appId) values(4,'app4')"); this.jdbcTemplate.update("upsert into SALT_TEST(id,appId) values(5,'app5')"); this.jdbcTemplate.update("upsert into SALT_TEST(id,appId) values(6,'app6')"); //2.split SALT_TEST at rowkey3,i.e.,split the first region byte[] id3=Bytes.toBytes(3); byte[] rowKey3=new byte[1+4]; System.arraycopy(id3, 0, rowKey3, 1, 4); byte salt3=SaltingUtil.getSaltingByte(rowKey3, 1, rowKey3.length-1, 2); rowKey3[0]=salt3; HBaseAdmin hbaseAdmin=this.getHBaseAdmin(); hbaseAdmin.split(Bytes.toBytes("SALT_TEST"), rowKey3); //3.wait the SALT_TEST split complele while(hbaseAdmin.getTableRegions(Bytes.toBytes("SALT_TEST")).size() < 3) { Thread.sleep(1000); } //4.we should make sure region0 and region1 is not on same region server,not share the GlobalCache HRegionInfo region1Info=hbaseAdmin.getTableRegions(Bytes.toBytes("SALT_TEST")).get(1); String region1EncodedName=region1Info.getEncodedName(); System.out.println(region1EncodedName); hbaseAdmin.move(Bytes.toBytes(region1EncodedName), null); //5.create RHS RIGHT_TEST table this.jdbcTemplate.update("drop table if exists RIGHT_TEST "); this.jdbcTemplate.update( "create table RIGHT_TEST"+ "("+ "appId VARCHAR not null primary key,"+ "createTime VARCHAR"+ ")"); this.jdbcTemplate.update("upsert into RIGHT_TEST(appId,createTime) values('app2','201601')"); this.jdbcTemplate.update("upsert into RIGHT_TEST(appId,createTime) values('app3','201602')"); this.jdbcTemplate.update("upsert into RIGHT_TEST(appId,createTime) values('app4','201603')"); this.jdbcTemplate.update("upsert into RIGHT_TEST(appId,createTime) values('app5','201604')"); //6.clear SALT_TEST's TableRegionCache,let join know SALT_TEST's newest 3 region ((PhoenixConnection)this.dataSource.getConnection()) .getQueryServices().clearTableRegionCache(Bytes.toBytes("SALT_TEST")); //7. join the salted table SALT_TEST with RIGHT_TEST,throw exception String sql="select * from SALT_TEST a inner join RIGHT_TEST b on a.appId=b.appId where a.id>=3 and a.id<=5"; List<Map<String,Object>> result=this.jdbcTemplate.queryForList(sql, new Object[0]); assertTrue(result.size()==3); }
I debug the source code,and find something error in the addServerCache method of org.apache.phoenix.cache.ServerCacheClient class, just as the following code. In line 166, we get all three regions of SALT_TEST table, but in line 176,the keyRanges is [[[\x00], [[\x00\x00\x00\x03 - \x00\x00\x00\x05]]]], so only the second region [\x00\x00\x00\x00\x03,\x01\x00\x00\x00\x00) can pass the keyRanges's intersects method in line 176. As a result, the RHS is only sent to the second region,However,we know SALT_TEST is a salted table, the third region [\x01\x00\x00\x00\x00,+∞) also has records matching the where condition,and the RHS should also be sent to the third region. Therefore, the correct keyRanges should be [[[\x00-\x01], [[\x00\x00\x00\x03 - \x00\x00\x00\x05]]]],not [[[\x00], [[\x00\x00\x00\x03 - \x00\x00\x00\x05]]]]
164 try { 165 final PTable cacheUsingTable = cacheUsingTableRef.getTable(); 166 List<HRegionLocation> locations = services.getAllTableRegions(cacheUsingTable.getPhysicalName().getBytes()); 167 int nRegions = locations.size(); 168 // Size these based on worst case 169 futures = new ArrayList<Future<Boolean>>(nRegions); 170 Set<HRegionLocation> servers = new HashSet<HRegionLocation>(nRegions); 171 for (HRegionLocation entry : locations) { 172 // Keep track of servers we've sent to and only send once 173 byte[] regionStartKey = entry.getRegionInfo().getStartKey(); 174 byte[] regionEndKey = entry.getRegionInfo().getEndKey(); 175 if ( ! servers.contains(entry) && 176 keyRanges.intersects(regionStartKey, regionEndKey, 177 cacheUsingTable.getIndexType() == IndexType.LOCAL ? 178 ScanUtil.getRowKeyOffset(regionStartKey, regionEndKey) : 0, true)) { 179 // Call RPC once per server 180 servers.add(entry); 181 if (LOG.isDebugEnabled()) {LOG.debug(addCustomAnnotations("Adding cache entry to be sent for " + entry, connection));} 182 final byte[] key = entry.getRegionInfo().getStartKey(); 183 final HTableInterface htable = services.getTable(cacheUsingTableRef.getTable().getPhysicalName().getBytes()); 184 closeables.add(htable); 185 futures.add(executor.submit(new JobCallable<Boolean>() { 186 187 @Override 188 public Boolean call() throws Exception { 189 final Map<byte[], AddServerCacheResponse> results; 190 try { 191 results = htable.coprocessorService(ServerCachingService.class, key, key,
I think the incorrect keyRanges problem is caused by the following code in org.apache.phoenix.compile.ScanRanges's ctor (around the 178 line), In my opinion ,the if statement should be if (isSalted && !isPointLookup).
if (useSkipScanFilter && isSalted && !isPointLookup) {
ranges.set(0, SaltingUtil.generateAllSaltingRanges(bucketNum));
}
Another problem is why when the salted table's first region is not split , the join sql can execute normally? I think it is caused by a subtle code error in the addServerCache method of org.apache.phoenix.cache.ServerCacheClient class:
171 for (HRegionLocation entry : locations) { 172 // Keep track of servers we've sent to and only send once 173 byte[] regionStartKey = entry.getRegionInfo().getStartKey(); 174 byte[] regionEndKey = entry.getRegionInfo().getEndKey(); 175 if ( ! servers.contains(entry) && 176 keyRanges.intersects(regionStartKey, regionEndKey, 177 cacheUsingTable.getIndexType() == IndexType.LOCAL ? 178 ScanUtil.getRowKeyOffset(regionStartKey, regionEndKey) : 0, true)) { 179 // Call RPC once per server 180 servers.add(entry); 181 if (LOG.isDebugEnabled()) {LOG.debug(addCustomAnnotations("Adding cache entry to be sent for " + entry, connection));} 182 final byte[] key = entry.getRegionInfo().getStartKey(); 183 final HTableInterface htable = services.getTable(cacheUsingTableRef.getTable().getPhysicalName().getBytes()); 184 closeables.add(htable); 185 futures.add(executor.submit(new JobCallable<Boolean>() { 186 187 @Override 188 public Boolean call() throws Exception { 189 final Map<byte[], AddServerCacheResponse> results; 190 try { 191 results = htable.coprocessorService(ServerCachingService.class, key, key,
In the above code, if the first region is not split, the first region (-∞,\x01\x00\x00\x00\x00) can always pass the if test in line 176, and the startKey of the first region is a byte[0](empty array), so in line 191, when we invoke the htable's coprocessorService method, the startKey and endKey parameters are both byte[0].
However, in the following org.apache.hadoop.hbase.client.HTable's getKeysAndRegionsInRange method, when the endKey parameter equals HConstants.EMPTY_END_ROW(which is also byte[0]), the bool variable endKeyIsEndOfTable is set to true. This causes all regions would be returned, which obviously is not as expected. Eventually, the RHS is sent to LHS 's all regions,and the join SQL certainly run normally :
private Pair<List<byte[]>, List<HRegionLocation>> getKeysAndRegionsInRange( final byte[] startKey, final byte[] endKey, final boolean includeEndKey, final boolean reload) throws IOException { final boolean endKeyIsEndOfTable = Bytes.equals(endKey,HConstants.EMPTY_END_ROW); if ((Bytes.compareTo(startKey, endKey) > 0) && !endKeyIsEndOfTable) { throw new IllegalArgumentException( "Invalid range: " + Bytes.toStringBinary(startKey) + " > " + Bytes.toStringBinary(endKey)); } List<byte[]> keysInRange = new ArrayList<byte[]>(); List<HRegionLocation> regionsInRange = new ArrayList<HRegionLocation>(); byte[] currentKey = startKey; do { HRegionLocation regionLocation = getRegionLocation(currentKey, reload); keysInRange.add(currentKey); regionsInRange.add(regionLocation); currentKey = regionLocation.getRegionInfo().getEndKey(); } while (!Bytes.equals(currentKey, HConstants.EMPTY_END_ROW) && (endKeyIsEndOfTable || Bytes.compareTo(currentKey, endKey) < 0 || (includeEndKey && Bytes.compareTo(currentKey, endKey) == 0))); return new Pair<List<byte[]>, List<HRegionLocation>>(keysInRange, regionsInRange); }
Attachments
Attachments
Issue Links
- is related to
-
PHOENIX-3199 ServerCacheClient sends cache to all regions unnecessarily
- Resolved
- links to