diff --git a/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenarios.java b/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenarios.java index efe9fff780..16965a3eb0 100644 --- a/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenarios.java +++ b/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenarios.java @@ -98,8 +98,11 @@ import static org.apache.hadoop.hive.metastore.ReplChangeManager.SOURCE_OF_REPLICATION; import static org.apache.hadoop.hive.metastore.Warehouse.DEFAULT_CATALOG_NAME; +import static org.apache.hadoop.hive.ql.exec.repl.ReplAck.LOAD_ACKNOWLEDGEMENT; +import static org.apache.hadoop.hive.ql.exec.repl.ReplAck.DUMP_ACKNOWLEDGEMENT; import static org.junit.Assert.assertEquals; import static org.junit.Assert.assertFalse; +import static org.junit.Assert.assertNotEquals; import static org.junit.Assert.assertNotNull; import static org.junit.Assert.assertNull; import static org.junit.Assert.assertTrue; @@ -315,8 +318,8 @@ public void testBasic() throws IOException { FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); Path dumpPath = new Path(bootstrapDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); - assertTrue(fs.exists(new Path(dumpPath, ReplUtils.DUMP_ACKNOWLEDGEMENT))); - assertTrue(fs.exists(new Path(dumpPath, ReplUtils.LOAD_ACKNOWLEDGEMENT))); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + assertTrue(fs.exists(new Path(dumpPath, LOAD_ACKNOWLEDGEMENT.toString()))); verifyRun("SELECT * from " + replicatedDbName + ".unptned", unptn_data, driverMirror); verifyRun("SELECT a from " + replicatedDbName + ".ptned WHERE b=1", ptn_data_1, driverMirror); @@ -367,8 +370,8 @@ public void testBootstrapFailedDump() throws IOException { advanceDumpDir(); FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); Path dumpPath = new Path(bootstrapDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); - assertTrue(fs.exists(new Path(dumpPath, ReplUtils.DUMP_ACKNOWLEDGEMENT))); - assertTrue(fs.exists(new Path(dumpPath, ReplUtils.LOAD_ACKNOWLEDGEMENT))); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + assertTrue(fs.exists(new Path(dumpPath, LOAD_ACKNOWLEDGEMENT.toString()))); verifyRun("SELECT * from " + replicatedDbName + ".unptned", unptnData, driverMirror); verifyRun("SELECT a from " + replicatedDbName + ".ptned WHERE b=1", ptnData1, driverMirror); @@ -452,7 +455,7 @@ public void testTaskCreationOptimization() throws Throwable { Path loadPath = new Path(dump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); //delete load ack to reload the same dump - loadPath.getFileSystem(hconf).delete(new Path(loadPath, ReplUtils.LOAD_ACKNOWLEDGEMENT), true); + loadPath.getFileSystem(hconf).delete(new Path(loadPath, LOAD_ACKNOWLEDGEMENT.toString()), true); loadAndVerify(dbNameReplica, dbName, dump.lastReplId); run("insert into table " + dbName + ".t2 partition(country='india') values ('delhi')", driver); @@ -466,7 +469,7 @@ public void testTaskCreationOptimization() throws Throwable { loadPath = new Path(dump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); //delete load ack to reload the same dump - loadPath.getFileSystem(hconf).delete(new Path(loadPath, ReplUtils.LOAD_ACKNOWLEDGEMENT), true); + loadPath.getFileSystem(hconf).delete(new Path(loadPath, LOAD_ACKNOWLEDGEMENT.toString()), true); loadAndVerify(dbNameReplica, dbName, dump.lastReplId); run("insert into table " + dbName + ".t2 partition(country='us') values ('sf')", driver); @@ -902,8 +905,8 @@ public void testIncrementalAdds() throws IOException { Tuple incrementalDump = incrementalLoadAndVerify(dbName, replDbName); FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); Path dumpPath = new Path(incrementalDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); - assertTrue(fs.exists(new Path(dumpPath, ReplUtils.DUMP_ACKNOWLEDGEMENT))); - assertTrue(fs.exists(new Path(dumpPath, ReplUtils.LOAD_ACKNOWLEDGEMENT))); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + assertTrue(fs.exists(new Path(dumpPath, LOAD_ACKNOWLEDGEMENT.toString()))); // VERIFY tables and partitions on destination for equivalence. verifyRun("SELECT * from " + replDbName + ".unptned_empty", empty, driverMirror); @@ -1723,7 +1726,8 @@ public void testIncrementalLoadWithOneFailedDump() throws IOException { Tuple incrementalDump = replDumpDb(dbName); //Remove the dump ack file, so that dump is treated as an invalid dump. - String ackFileRelativePath = ReplUtils.REPL_HIVE_BASE_DIR + File.separator + ReplUtils.DUMP_ACKNOWLEDGEMENT; + String ackFileRelativePath = ReplUtils.REPL_HIVE_BASE_DIR + File.separator + + DUMP_ACKNOWLEDGEMENT.toString(); Path dumpFinishedAckFilePath = new Path(incrementalDump.dumpLocation, ackFileRelativePath); Path tmpDumpFinishedAckFilePath = new Path(dumpFinishedAckFilePath.getParent(), "old_" + dumpFinishedAckFilePath.getName()); @@ -1809,7 +1813,7 @@ public void testIncrementalLoadWithPreviousDumpDeleteFailed() throws IOException FileSystem fs = FileSystem.get(fileToDelete.toUri(), hconf); fs.delete(fileToDelete, true); assertTrue(fs.exists(bootstrapDumpDir)); - assertTrue(fs.exists(new Path(bootstrapDumpDir, ReplUtils.DUMP_ACKNOWLEDGEMENT))); + assertTrue(fs.exists(new Path(bootstrapDumpDir, DUMP_ACKNOWLEDGEMENT.toString()))); loadAndVerify(replDbName, dbName, incrDump.lastReplId); @@ -3835,4 +3839,225 @@ private static void createTestDataFile(String filename, String[] lines) throws I } } } + + @Test + public void testCheckPointingDataDumpFailure() throws IOException { + String testname = testName.getMethodName(); + String dbName = createDB(testname, driver); + run("CREATE TABLE " + dbName + ".t1(a string) STORED AS TEXTFILE", driver); + run("CREATE TABLE " + dbName + ".t2(a string) STORED AS TEXTFILE", driver); + + run("insert into "+ dbName +".t1 values (1)", driver); + run("insert into "+ dbName +".t1 values (2)", driver); + run("insert into "+ dbName +".t1 values (3)", driver); + run("insert into "+ dbName +".t2 values (11)", driver); + run("insert into "+ dbName +".t2 values (21)", driver); + + String replicatedDbName = dbName + "_dupe"; + Tuple bootstrapDump = replDumpDb(dbName); + + FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); + Path dumpPath = new Path(bootstrapDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + Path dbPath = new Path(dumpPath, dbName); + Path metadataPath = new Path(dbPath, EximUtil.METADATA_PATH_NAME); + long modifiedTimeMetadata = fs.getFileStatus(metadataPath).getModificationTime(); + Path dataPath = new Path(dbPath, EximUtil.DATA_PATH_NAME); + Path tablet1Path = new Path(dataPath, "t1"); + Path tablet2Path = new Path(dataPath, "t2"); + long modifiedTimeTable1 = fs.getFileStatus(tablet1Path).getModificationTime(); + long modifiedTimeTable2 = fs.getFileStatus(tablet2Path).getModificationTime(); + //Delete dump ack and t2 data, metadata should be rewritten, data should be same for t1 but rewritten for t2 + fs.delete(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()), true); + assertFalse(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + FileStatus[] statuses = fs.listStatus(tablet2Path); + //Delete t2 data + fs.delete(statuses[0].getPath(), true); + //Do another dump. It should only dump table t2. Modification time of table t1 should be same while t2 is greater + Tuple nextDump = incrementalLoadAndVerify(dbName, replicatedDbName); + assertEquals(nextDump.dumpLocation, bootstrapDump.dumpLocation); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + assertEquals(modifiedTimeTable1, fs.getFileStatus(tablet1Path).getModificationTime()); + assertTrue(modifiedTimeTable2 < fs.getFileStatus(tablet2Path).getModificationTime()); + assertTrue(modifiedTimeMetadata < fs.getFileStatus(metadataPath).getModificationTime()); + verifySetup("select * from " + replicatedDbName + ".t1", new String[]{"1", "2", "3"}, driver); + verifySetup("select * from " + replicatedDbName + ".t2", new String[]{"11", "21"}, driver); + } + + @Test + public void testCheckPointingWithSourceTableDataInserted() throws IOException { + String testname = testName.getMethodName(); + String dbName = createDB(testname, driver); + run("CREATE TABLE " + dbName + ".t1(a string) STORED AS TEXTFILE", driver); + run("CREATE TABLE " + dbName + ".t2(a string) STORED AS TEXTFILE", driver); + + run("insert into "+ dbName +".t1 values (1)", driver); + run("insert into "+ dbName +".t1 values (2)", driver); + run("insert into "+ dbName +".t1 values (3)", driver); + run("insert into "+ dbName +".t2 values (11)", driver); + run("insert into "+ dbName +".t2 values (21)", driver); + + String replicatedDbName = dbName + "_dupe"; + Tuple bootstrapDump = replDumpDb(dbName); + + FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); + Path dumpPath = new Path(bootstrapDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + Path dbPath = new Path(dumpPath, dbName); + Path dataPath = new Path(dbPath, EximUtil.DATA_PATH_NAME); + Path tablet1Path = new Path(dataPath, "t1"); + Path tablet2Path = new Path(dataPath, "t2"); + long modifiedTimeTable1 = fs.getFileStatus(tablet1Path).getModificationTime(); + long modifiedTimeTable2 = fs.getFileStatus(tablet2Path).getModificationTime(); + //Delete table 2 data + FileStatus[] statuses = fs.listStatus(tablet2Path); + //Delete t2 data + fs.delete(statuses[0].getPath(), true); + fs.delete(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()), true); + //Do a load. It shouldn't contain any data + run("REPL LOAD " + dbName + " INTO " + replicatedDbName, driverMirror); + verifySetup("select tables in " + replicatedDbName, new String[]{}, driver); + + //Do another dump. It should only dump table t2. Also insert new data. New data should be there in target + run("insert into "+ dbName +".t2 values (13)", driver); + run("insert into "+ dbName +".t2 values (24)", driver); + + incrementalLoadAndVerify(dbName, replicatedDbName); + assertEquals(modifiedTimeTable1, fs.getFileStatus(tablet1Path).getModificationTime()); + assertTrue(modifiedTimeTable2 < fs.getFileStatus(tablet2Path).getModificationTime()); + String[] t1Data = new String[]{"1", "2", "3"}; + String[] t2Data = new String[]{"11", "21", "13", "24"}; + verifySetup("select * from " + replicatedDbName + ".t1", t1Data, driver); + verifySetup("select * from " + replicatedDbName + ".t2", t2Data, driver); + } + + @Test + public void testCheckPointingWithNewTablesAdded() throws IOException { + String testname = testName.getMethodName(); + String dbName = createDB(testname, driver); + run("CREATE TABLE " + dbName + ".t1(a string) STORED AS TEXTFILE", driver); + run("CREATE TABLE " + dbName + ".t2(a string) STORED AS TEXTFILE", driver); + + run("insert into "+ dbName +".t1 values (1)", driver); + run("insert into "+ dbName +".t1 values (2)", driver); + run("insert into "+ dbName +".t1 values (3)", driver); + run("insert into "+ dbName +".t2 values (11)", driver); + run("insert into "+ dbName +".t2 values (21)", driver); + + String replicatedDbName = dbName + "_dupe"; + Tuple bootstrapDump = replDumpDb(dbName); + + FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); + Path dumpPath = new Path(bootstrapDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + Path dbPath = new Path(dumpPath, dbName); + Path dataPath = new Path(dbPath, EximUtil.DATA_PATH_NAME); + Path tablet1Path = new Path(dataPath, "t1"); + Path tablet2Path = new Path(dataPath, "t2"); + long modifiedTimeTable1 = fs.getFileStatus(tablet1Path).getModificationTime(); + long modifiedTimeTable2 = fs.getFileStatus(tablet2Path).getModificationTime(); + //Delete table 2 data + FileStatus[] statuses = fs.listStatus(tablet2Path); + fs.delete(statuses[0].getPath(), true); + fs.delete(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()), true); + //Do a load. It shouldn't contain any data + run("REPL LOAD " + dbName + " INTO " + replicatedDbName, driverMirror); + verifySetup("select tables in " + replicatedDbName, new String[]{}, driver); + + //Do another dump. It should only dump table t2. Also insert new data. New data will be there in target + run("insert into "+ dbName +".t2 values (13)", driver); + run("insert into "+ dbName +".t2 values (24)", driver); + run("CREATE TABLE " + dbName + ".t3(a string) STORED AS TEXTFILE", driver); + run("insert into "+ dbName +".t3 values (1)", driver); + run("insert into "+ dbName +".t3 values (2)", driver); + + incrementalLoadAndVerify(dbName, replicatedDbName); + assertEquals(modifiedTimeTable1, fs.getFileStatus(tablet1Path).getModificationTime()); + assertTrue(modifiedTimeTable2 < fs.getFileStatus(tablet2Path).getModificationTime()); + String[] t1Data = new String[]{"1", "2", "3"}; + String[] t2Data = new String[]{"11", "21", "13", "24"}; + String[] tableList = new String[]{"t1", "t2", "t3"}; + String[] t3Data = new String[]{"1", "2", "3"}; + verifySetup("select * from " + replicatedDbName + ".t1", t1Data, driver); + verifySetup("select * from " + replicatedDbName + ".t2", t2Data, driver); + verifySetup("show tables in " + replicatedDbName, tableList, driver); + verifySetup("select * from " + replicatedDbName + ".t3", t3Data, driver); + } + + @Test + public void testCheckPointingWithSourceTableDeleted() throws IOException { + String testname = testName.getMethodName(); + String dbName = createDB(testname, driver); + run("CREATE TABLE " + dbName + ".t1(a string) STORED AS TEXTFILE", driver); + run("CREATE TABLE " + dbName + ".t2(a string) STORED AS TEXTFILE", driver); + + run("insert into "+ dbName +".t1 values (1)", driver); + run("insert into "+ dbName +".t1 values (2)", driver); + run("insert into "+ dbName +".t1 values (3)", driver); + run("insert into "+ dbName +".t2 values (11)", driver); + run("insert into "+ dbName +".t2 values (21)", driver); + + String replicatedDbName = dbName + "_dupe"; + Tuple bootstrapDump = replDumpDb(dbName); + FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); + Path dumpPath = new Path(bootstrapDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + Path dbPath = new Path(dumpPath, dbName); + + //Delete dump ack and t2 data, Also drop table. New data will be there in target + fs.delete(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()), true); + assertFalse(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + Path dataPath = new Path(dbPath, EximUtil.DATA_PATH_NAME); + Path tablet2Path = new Path(dataPath, "t2"); + FileStatus[] statuses = fs.listStatus(tablet2Path); + //Delete t2 data. + fs.delete(statuses[0].getPath(), true); + //Drop table t1. Target shouldn't have t1 table as metadata dump is rewritten + run("drop "+ dbName +".t1", driver); + + incrementalLoadAndVerify(dbName, replicatedDbName); + String[] tableList = new String[]{"t2"}; + String[] t2Data = new String[]{"11", "21"}; + verifySetup("show tables in " + replicatedDbName, tableList, driver); + verifySetup("select * from " + replicatedDbName + ".t2", t2Data, driver); + } + + @Test + public void testCheckPointingMetadataDumpFailure() throws IOException { + String testname = testName.getMethodName(); + String dbName = createDB(testname, driver); + run("CREATE TABLE " + dbName + ".t1(a string) STORED AS TEXTFILE", driver); + run("CREATE TABLE " + dbName + ".t2(a string) STORED AS TEXTFILE", driver); + + run("insert into "+ dbName +".t1 values (1)", driver); + run("insert into "+ dbName +".t1 values (2)", driver); + run("insert into "+ dbName +".t1 values (3)", driver); + run("insert into "+ dbName +".t2 values (11)", driver); + run("insert into "+ dbName +".t2 values (21)", driver); + + String replicatedDbName = dbName + "_dupe"; + Tuple bootstrapDump = replDumpDb(dbName); + FileSystem fs = new Path(bootstrapDump.dumpLocation).getFileSystem(hconf); + Path dumpPath = new Path(bootstrapDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + Path dbPath = new Path(dumpPath, dbName); + + //Delete dump ack and metadata ack, everything should be rewritten in a new dump dir + fs.delete(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()), true); + fs.delete(new Path(dumpPath, "_dumpmetadata"), true); + assertFalse(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + //Insert new data + run("insert into "+ dbName +".t1 values (12)", driver); + run("insert into "+ dbName +".t1 values (13)", driver); + //Do another dump. It should dump everything in a new dump dir + // checkpointing will not be used + Tuple nextDump = incrementalLoadAndVerify(dbName, replicatedDbName); + assertNotEquals(nextDump.dumpLocation, bootstrapDump.dumpLocation); + dumpPath = new Path(nextDump.dumpLocation, ReplUtils.REPL_HIVE_BASE_DIR); + assertTrue(fs.exists(new Path(dumpPath, DUMP_ACKNOWLEDGEMENT.toString()))); + String[] t1Data = new String[]{"1", "2", "3", "12", "13"}; + String[] t2Data = new String[]{"11", "21"}; + verifySetup("select * from " + replicatedDbName + ".t2", t2Data, driver); + verifySetup("select * from " + replicatedDbName + ".t1", t1Data, driver); + } } diff --git a/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenariosAcrossInstances.java b/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenariosAcrossInstances.java index 56b27a555e..33124c8f76 100644 --- a/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenariosAcrossInstances.java +++ b/itests/hive-unit/src/test/java/org/apache/hadoop/hive/ql/parse/TestReplicationScenariosAcrossInstances.java @@ -55,6 +55,7 @@ import java.util.stream.Collectors; import static org.apache.hadoop.hive.metastore.ReplChangeManager.SOURCE_OF_REPLICATION; +import static org.apache.hadoop.hive.ql.exec.repl.ReplAck.LOAD_ACKNOWLEDGEMENT; import static org.hamcrest.CoreMatchers.equalTo; import static org.hamcrest.CoreMatchers.is; import static org.hamcrest.MatcherAssert.assertThat; @@ -1093,13 +1094,13 @@ public void testIfBootstrapReplLoadFailWhenRetryAfterBootstrapComplete() throws // To retry with same dump delete the load ack new Path(tuple.dumpLocation).getFileSystem(conf).delete(new Path( - hiveDumpLocation, ReplUtils.LOAD_ACKNOWLEDGEMENT), true); + hiveDumpLocation, LOAD_ACKNOWLEDGEMENT.toString()), true); // Retry with same dump with which it was already loaded also fails. replica.loadFailure(replicatedDbName, primaryDbName); // To retry with same dump delete the load ack new Path(tuple.dumpLocation).getFileSystem(conf).delete(new Path( - hiveDumpLocation, ReplUtils.LOAD_ACKNOWLEDGEMENT), true); + hiveDumpLocation, LOAD_ACKNOWLEDGEMENT.toString()), true); // Retry from same dump when the database is empty is also not allowed. replica.run("drop table t1") .run("drop table t2") @@ -1344,7 +1345,7 @@ public void testMoveOptimizationIncrementalFailureAfterCopyReplace() throws Thro //delete load ack to reuse the dump new Path(tuple.dumpLocation).getFileSystem(conf).delete(new Path(tuple.dumpLocation + Path.SEPARATOR + ReplUtils.REPL_HIVE_BASE_DIR + Path.SEPARATOR - + ReplUtils.LOAD_ACKNOWLEDGEMENT), true); + + LOAD_ACKNOWLEDGEMENT.toString()), true); replica.load(replicatedDbName_CM, primaryDbName, withConfigs); replica.run("alter database " + replicatedDbName + " set DBPROPERTIES ('" + SOURCE_OF_REPLICATION + "' = '1,2,3')") @@ -1370,7 +1371,7 @@ public void testMoveOptimizationIncrementalFailureAfterCopy() throws Throwable { replica.load(replicatedDbName, primaryDbName, withConfigs); //delete load ack to reuse the dump new Path(bootstrapDump.dumpLocation).getFileSystem(conf).delete(new Path(bootstrapDump.dumpLocation - + Path.SEPARATOR + ReplUtils.REPL_HIVE_BASE_DIR + Path.SEPARATOR + ReplUtils.LOAD_ACKNOWLEDGEMENT), true); + + Path.SEPARATOR + ReplUtils.REPL_HIVE_BASE_DIR + Path.SEPARATOR + LOAD_ACKNOWLEDGEMENT.toString()), true); replica.load(replicatedDbName_CM, primaryDbName, withConfigs); replica.run("alter database " + replicatedDbName + " set DBPROPERTIES ('" + SOURCE_OF_REPLICATION + "' = '1,2,3')") .run("alter database " + replicatedDbName_CM + " set DBPROPERTIES ('" + SOURCE_OF_REPLICATION + "' = '1,2,3')"); @@ -1423,7 +1424,7 @@ public Boolean apply(NotificationEvent entry) { //delete load ack to reuse the dump new Path(tuple.dumpLocation).getFileSystem(conf).delete(new Path(tuple.dumpLocation + Path.SEPARATOR + ReplUtils.REPL_HIVE_BASE_DIR + Path.SEPARATOR - + ReplUtils.LOAD_ACKNOWLEDGEMENT), true); + + LOAD_ACKNOWLEDGEMENT.toString()), true); InjectableBehaviourObjectStore.setAddNotificationModifier(callerVerifier); diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/ExportTask.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/ExportTask.java index 56f0c9308e..d3e94134ab 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/ExportTask.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/ExportTask.java @@ -46,7 +46,7 @@ public int execute() { TableExport.Paths exportPaths = new TableExport.Paths( work.getAstRepresentationForErrorMsg(), work.getExportRootDir(), conf, false); Hive db = getHive(); - LOG.debug("Exporting data to: {}", exportPaths.exportRootDir()); + LOG.debug("Exporting data to: {}", exportPaths.metadataExportRootDir()); work.acidPostProcess(db); TableExport tableExport = new TableExport(exportPaths, work.getTableSpec(), work.getReplicationSpec(), db, null, conf, work.getMmContext()); diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/DirCopyWork.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/DirCopyWork.java index efef052268..46f9bb3add 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/DirCopyWork.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/DirCopyWork.java @@ -28,6 +28,7 @@ Explain.Level.DEFAULT, Explain.Level.EXTENDED }) public class DirCopyWork implements Serializable { + private static final long serialVersionUID = 1L; private final Path fullyQualifiedSourcePath; private final Path fullyQualifiedTargetPath; diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplAck.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplAck.java new file mode 100644 index 0000000000..db8db5f8e7 --- /dev/null +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplAck.java @@ -0,0 +1,35 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.hadoop.hive.ql.exec.repl; + +/** + * ReplAck, used for repl acknowledgement constants. + */ +public enum ReplAck { + DUMP_ACKNOWLEDGEMENT("_finished_dump"), + LOAD_ACKNOWLEDGEMENT("_finished_load"); + private String ack; + ReplAck(String ack) { + this.ack = ack; + } + + @Override + public String toString() { + return ack; + } +} diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplDumpTask.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplDumpTask.java index 69f6ffef5a..7500d4d887 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplDumpTask.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplDumpTask.java @@ -93,6 +93,7 @@ import java.util.ArrayList; import java.util.concurrent.TimeUnit; import static org.apache.hadoop.hive.ql.exec.repl.ReplExternalTables.Writer; +import static org.apache.hadoop.hive.ql.exec.repl.ReplAck.LOAD_ACKNOWLEDGEMENT; public class ReplDumpTask extends Task implements Serializable { private static final long serialVersionUID = 1L; @@ -135,20 +136,20 @@ public int execute() { Path dumpRoot = new Path(conf.getVar(HiveConf.ConfVars.REPLDIR), Base64.getEncoder().encodeToString(work.dbNameOrPattern.toLowerCase() .getBytes(StandardCharsets.UTF_8.name()))); - Path previousHiveDumpPath = getPreviousDumpMetadataPath(dumpRoot); + Path previousValidHiveDumpPath = getPreviousValidDumpMetadataPath(dumpRoot); //If no previous dump is present or previous dump is already loaded, proceed with the dump operation. - if (shouldDump(previousHiveDumpPath)) { - Path currentDumpPath = new Path(dumpRoot, getNextDumpDir()); + if (shouldDump(previousValidHiveDumpPath)) { + Path currentDumpPath = getCurrentDumpPath(dumpRoot); Path hiveDumpRoot = new Path(currentDumpPath, ReplUtils.REPL_HIVE_BASE_DIR); DumpMetaData dmd = new DumpMetaData(hiveDumpRoot, conf); // Initialize ReplChangeManager instance since we will require it to encode file URI. ReplChangeManager.getInstance(conf); Path cmRoot = new Path(conf.getVar(HiveConf.ConfVars.REPLCMDIR)); Long lastReplId; - if (previousHiveDumpPath == null) { + if (previousValidHiveDumpPath == null) { lastReplId = bootStrapDump(hiveDumpRoot, dmd, cmRoot, hiveDb); } else { - work.setEventFrom(getEventFromPreviousDumpMetadata(previousHiveDumpPath)); + work.setEventFrom(getEventFromPreviousDumpMetadata(previousValidHiveDumpPath)); lastReplId = incrementalDump(hiveDumpRoot, dmd, cmRoot, hiveDb); } work.setResultValues(Arrays.asList(currentDumpPath.toUri().toString(), String.valueOf(lastReplId))); @@ -166,6 +167,16 @@ public int execute() { return 0; } + private Path getCurrentDumpPath(Path dumpRoot) throws IOException { + Path previousDumpPath = getPreviousDumpPath(dumpRoot); + if (previousDumpPath != null && !validDump(previousDumpPath) && shouldResumePreviousDump(previousDumpPath)) { + //Resume previous dump + return previousDumpPath; + } else { + return new Path(dumpRoot, getNextDumpDir()); + } + } + private void initiateDataCopyTasks() throws SemanticException, IOException { TaskTracker taskTracker = new TaskTracker(conf.getIntVar(HiveConf.ConfVars.REPL_APPROX_MAX_LOAD_TASKS)); List> childTasks = new ArrayList<>(); @@ -183,7 +194,8 @@ private void initiateDataCopyTasks() throws SemanticException, IOException { private void finishRemainingTasks() throws SemanticException, IOException { prepareReturnValues(work.getResultValues()); Path dumpAckFile = new Path(work.getCurrentDumpPath(), - ReplUtils.REPL_HIVE_BASE_DIR + File.separator + ReplUtils.DUMP_ACKNOWLEDGEMENT); + ReplUtils.REPL_HIVE_BASE_DIR + File.separator + + ReplAck.DUMP_ACKNOWLEDGEMENT.toString()); Utils.create(dumpAckFile, conf); deleteAllPreviousDumpMeta(work.getCurrentDumpPath()); } @@ -233,7 +245,7 @@ private Long getEventFromPreviousDumpMetadata(Path previousDumpPath) throws Sema return 0L; } - private Path getPreviousDumpMetadataPath(Path dumpRoot) throws IOException { + private Path getPreviousValidDumpMetadataPath(Path dumpRoot) throws IOException { FileStatus latestValidStatus = null; FileSystem fs = dumpRoot.getFileSystem(conf); if (fs.exists(dumpRoot)) { @@ -241,8 +253,8 @@ private Path getPreviousDumpMetadataPath(Path dumpRoot) throws IOException { for (FileStatus status : statuses) { LOG.info("Evaluating previous dump dir path:{}", status.getPath()); if (latestValidStatus == null) { - latestValidStatus = validDump(fs, status.getPath()) ? status : null; - } else if (validDump(fs, status.getPath()) + latestValidStatus = validDump(status.getPath()) ? status : null; + } else if (validDump(status.getPath()) && status.getModificationTime() > latestValidStatus.getModificationTime()) { latestValidStatus = status; } @@ -254,10 +266,14 @@ private Path getPreviousDumpMetadataPath(Path dumpRoot) throws IOException { return latestDumpDir; } - private boolean validDump(FileSystem fs, Path dumpDir) throws IOException { + private boolean validDump(Path dumpDir) throws IOException { //Check if it was a successful dump - Path hiveDumpDir = new Path(dumpDir, ReplUtils.REPL_HIVE_BASE_DIR); - return fs.exists(new Path(hiveDumpDir, ReplUtils.DUMP_ACKNOWLEDGEMENT)); + if (dumpDir != null) { + FileSystem fs = dumpDir.getFileSystem(conf); + Path hiveDumpDir = new Path(dumpDir, ReplUtils.REPL_HIVE_BASE_DIR); + return fs.exists(new Path(hiveDumpDir, ReplAck.DUMP_ACKNOWLEDGEMENT.toString())); + } + return false; } private boolean shouldDump(Path previousDumpPath) throws IOException { @@ -267,7 +283,7 @@ private boolean shouldDump(Path previousDumpPath) throws IOException { return true; } else { FileSystem fs = previousDumpPath.getFileSystem(conf); - return fs.exists(new Path(previousDumpPath, ReplUtils.LOAD_ACKNOWLEDGEMENT)); + return fs.exists(new Path(previousDumpPath, LOAD_ACKNOWLEDGEMENT.toString())); } } @@ -472,7 +488,7 @@ private Long incrementalDump(Path dumpRoot, DumpMetaData dmd, Path cmRoot, Hive if (shouldBootstrapDumpTable(table)) { HiveWrapper.Tuple tableTuple = new HiveWrapper(hiveDb, dbName).table(table); managedTableCopyPaths.addAll( - dumpTable(dbName, tableName, validTxnList, dbRoot, dumpRoot, bootDumpBeginReplId, + dumpTable(dbName, tableName, validTxnList, dbRoot, bootDumpBeginReplId, hiveDb, tableTuple)); } if (tableList != null && isTableSatifiesConfig(table)) { @@ -611,16 +627,21 @@ Long bootStrapDump(Path dumpRoot, DumpMetaData dmd, Path cmRoot, Hive hiveDb) List tableList; LOG.info("Bootstrap Dump for db {}", work.dbNameOrPattern); + List extTableCopyWorks = new ArrayList<>(); + List managedTableCopyPaths = new ArrayList<>(); long timeoutInMs = HiveConf.getTimeVar(conf, HiveConf.ConfVars.REPL_BOOTSTRAP_DUMP_OPEN_TXN_TIMEOUT, TimeUnit.MILLISECONDS); long waitUntilTime = System.currentTimeMillis() + timeoutInMs; - String validTxnList = getValidTxnListForReplDump(hiveDb, waitUntilTime); - List extTableCopyWorks = new ArrayList<>(); - List managedTableCopyPaths = new ArrayList<>(); for (String dbName : Utils.matchesDb(hiveDb, work.dbNameOrPattern)) { LOG.debug("Dumping db: " + dbName); - + Path dbRoot = getBootstrapDbRoot(dumpRoot, dbName, false); + Path metadataPath = new Path(dbRoot, EximUtil.METADATA_PATH_NAME); + if (shouldResumePreviousDump(dmd)) { + //clear the metadata. We need to rewrite the metadata as the write id list will be changed + //We can't reuse the previous write id as it might be invalid due to compaction + dumpRoot.getFileSystem(conf).delete(metadataPath, true); + } // TODO : Currently we don't support separate table list for each database. tableList = work.replScope.includeAllTables() ? null : new ArrayList<>(); Database db = hiveDb.getDatabase(dbName); @@ -634,15 +655,15 @@ Long bootStrapDump(Path dumpRoot, DumpMetaData dmd, Path cmRoot, Hive hiveDb) Utils.getAllTables(hiveDb, dbName, work.replScope).size(), hiveDb.getAllFunctions().size()); replLogger.startLog(); - Path dbRoot = dumpDbMetadata(dbName, dumpRoot, bootDumpBeginReplId, hiveDb); - dumpFunctionMetadata(dbName, dumpRoot, hiveDb); + dumpDbMetadata(dbName, metadataPath, bootDumpBeginReplId, hiveDb); + dumpFunctionMetadata(dbName, metadataPath, hiveDb); String uniqueKey = Utils.setDbBootstrapDumpState(hiveDb, dbName); Exception caught = null; - try (Writer writer = new Writer(dbRoot, conf)) { + try (Writer writer = new Writer(metadataPath, conf)) { List extTableLocations = new LinkedList<>(); for (String tblName : Utils.matchesTbl(hiveDb, dbName, work.replScope)) { - LOG.debug("Dumping table: " + tblName + " to db root " + dbRoot.toUri()); + LOG.debug("Dumping table: " + tblName + " to db root " + metadataPath.toUri()); Table table = null; try { @@ -653,14 +674,15 @@ Long bootStrapDump(Path dumpRoot, DumpMetaData dmd, Path cmRoot, Hive hiveDb) LOG.debug("Adding table {} to external tables list", tblName); extTableLocations.addAll(writer.dataLocationDump(tableTuple.object)); } - managedTableCopyPaths.addAll(dumpTable(dbName, tblName, validTxnList, dbRoot, dumpRoot, bootDumpBeginReplId, + managedTableCopyPaths.addAll(dumpTable(dbName, tblName, validTxnList, dbRoot, + bootDumpBeginReplId, hiveDb, tableTuple)); } catch (InvalidTableException te) { // Bootstrap dump shouldn't fail if the table is dropped/renamed while dumping it. // Just log a debug message and skip it. LOG.debug(te.getMessage()); } - dumpConstraintMetadata(dbName, tblName, dbRoot, hiveDb); + dumpConstraintMetadata(dbName, tblName, metadataPath, hiveDb); if (tableList != null && isTableSatifiesConfig(table)) { tableList.add(tblName); } @@ -677,11 +699,11 @@ Long bootStrapDump(Path dumpRoot, DumpMetaData dmd, Path cmRoot, Hive hiveDb) throw e; } else { LOG.error("failed to reset the db state for " + uniqueKey - + " on failure of repl dump", e); + + " on failure of repl dump", e); throw caught; } } - if(caught != null) { + if (caught != null) { throw caught; } } @@ -689,29 +711,43 @@ Long bootStrapDump(Path dumpRoot, DumpMetaData dmd, Path cmRoot, Hive hiveDb) } Long bootDumpEndReplId = currentNotificationId(hiveDb); LOG.info("Preparing to return {},{}->{}", - dumpRoot.toUri(), bootDumpBeginReplId, bootDumpEndReplId); + dumpRoot.toUri(), bootDumpBeginReplId, bootDumpEndReplId); dmd.setDump(DumpType.BOOTSTRAP, bootDumpBeginReplId, bootDumpEndReplId, cmRoot); dmd.write(); + work.setDirCopyIterator(extTableCopyWorks.iterator()); work.setManagedTableCopyPathIterator(managedTableCopyPaths.iterator()); return bootDumpBeginReplId; } + private boolean shouldResumePreviousDump(DumpMetaData dumpMetaData) { + try { + return dumpMetaData.getEventFrom() != null; + } catch (Exception e) { + LOG.info("No previous dump present"); + return false; + } + } + + private boolean shouldResumePreviousDump(Path dumpPath) { + Path hiveDumpPath = new Path(dumpPath, ReplUtils.REPL_HIVE_BASE_DIR); + return shouldResumePreviousDump(new DumpMetaData(hiveDumpPath, conf)); + } + long currentNotificationId(Hive hiveDb) throws TException { return hiveDb.getMSC().getCurrentNotificationEventId().getEventId(); } - Path dumpDbMetadata(String dbName, Path dumpRoot, long lastReplId, Hive hiveDb) throws Exception { - Path dbRoot = getBootstrapDbRoot(dumpRoot, dbName, false); + Path dumpDbMetadata(String dbName, Path metadataRoot, long lastReplId, Hive hiveDb) throws Exception { // TODO : instantiating FS objects are generally costly. Refactor - FileSystem fs = dbRoot.getFileSystem(conf); - Path dumpPath = new Path(dbRoot, EximUtil.METADATA_NAME); + FileSystem fs = metadataRoot.getFileSystem(conf); + Path dumpPath = new Path(metadataRoot, EximUtil.METADATA_NAME); HiveWrapper.Tuple database = new HiveWrapper(hiveDb, dbName, lastReplId).database(); EximUtil.createDbExportDump(fs, dumpPath, database.object, database.replicationSpec); - return dbRoot; + return metadataRoot; } - List dumpTable(String dbName, String tblName, String validTxnList, Path dbRoot, Path dumproot, + List dumpTable(String dbName, String tblName, String validTxnList, Path dbRoot, long lastReplId, Hive hiveDb, HiveWrapper.Tuple
tuple) throws Exception { LOG.info("Bootstrap Dump for table " + tblName); TableSpec tableSpec = new TableSpec(tuple.object); @@ -827,8 +863,26 @@ private String getNextDumpDir() { } } - void dumpFunctionMetadata(String dbName, Path dumpRoot, Hive hiveDb) throws Exception { - Path functionsRoot = new Path(new Path(dumpRoot, dbName), ReplUtils.FUNCTIONS_ROOT_DIR_NAME); + private Path getPreviousDumpPath(Path dumpRoot) throws IOException { + FileSystem fs = dumpRoot.getFileSystem(conf); + if (fs.exists(dumpRoot)) { + FileStatus[] statuses = fs.listStatus(dumpRoot); + if (statuses.length > 0) { + FileStatus latestValidStatus = statuses[0]; + for (FileStatus status : statuses) { + LOG.info("Evaluating previous dump dir path:{}", status.getPath()); + if (status.getModificationTime() > latestValidStatus.getModificationTime()) { + latestValidStatus = status; + } + } + return latestValidStatus.getPath(); + } + } + return null; + } + + void dumpFunctionMetadata(String dbName, Path metadataRoot, Hive hiveDb) throws Exception { + Path functionsRoot = new Path(metadataRoot, ReplUtils.FUNCTIONS_ROOT_DIR_NAME); List functionNames = hiveDb.getFunctions(dbName, "*"); for (String functionName : functionNames) { HiveWrapper.Tuple tuple = functionTuple(functionName, dbName, hiveDb); diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplLoadTask.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplLoadTask.java index 3427b59e67..a5935552c4 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplLoadTask.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/ReplLoadTask.java @@ -45,7 +45,6 @@ import org.apache.hadoop.hive.ql.exec.repl.bootstrap.load.util.Context; import org.apache.hadoop.hive.ql.exec.repl.incremental.IncrementalLoadTasksBuilder; import org.apache.hadoop.hive.ql.exec.repl.util.AddDependencyToLeaves; -import org.apache.hadoop.hive.ql.exec.repl.util.ReplUtils; import org.apache.hadoop.hive.ql.exec.repl.util.TaskTracker; import org.apache.hadoop.hive.ql.exec.util.DAGTraversal; import org.apache.hadoop.hive.ql.metadata.Hive; @@ -65,6 +64,7 @@ import java.util.Map; import static org.apache.hadoop.hive.ql.exec.repl.bootstrap.load.LoadDatabase.AlterDatabase; +import static org.apache.hadoop.hive.ql.exec.repl.ReplAck.LOAD_ACKNOWLEDGEMENT; public class ReplLoadTask extends Task implements Serializable { private final static int ZERO_TASKS = 0; @@ -316,7 +316,7 @@ private void createReplLoadCompleteAckTask() { || (!work.isIncrementalLoad() && !work.hasBootstrapLoadTasks())) { //All repl load tasks are executed and status is 0, create the task to add the acknowledgement AckWork replLoadAckWork = new AckWork( - new Path(work.dumpDirectory, ReplUtils.LOAD_ACKNOWLEDGEMENT)); + new Path(work.dumpDirectory, LOAD_ACKNOWLEDGEMENT.toString())); Task loadAckWorkTask = TaskFactory.get(replLoadAckWork, conf); if (this.childTasks.isEmpty()) { this.childTasks.add(loadAckWorkTask); diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/TableEvent.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/TableEvent.java index 10732b0f69..05ef274449 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/TableEvent.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/TableEvent.java @@ -42,4 +42,6 @@ Licensed to the Apache Software Foundation (ASF) under one * Exposing the FileSystem implementation outside which is what it should NOT do. */ Path metadataPath(); + + Path dataPath(); } diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/BootstrapEventsIterator.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/BootstrapEventsIterator.java index 1af6a4c637..50e334b398 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/BootstrapEventsIterator.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/BootstrapEventsIterator.java @@ -162,11 +162,12 @@ public ReplLogger replLogger() { private void initReplLogger() { try { Path dbDumpPath = currentDatabaseIterator.dbLevelPath(); + Path metadataPath = new Path(dbDumpPath, EximUtil.METADATA_PATH_NAME); FileSystem fs = dbDumpPath.getFileSystem(hiveConf); - long numTables = getSubDirs(fs, dbDumpPath).length; + long numTables = getSubDirs(fs, metadataPath).length; long numFunctions = 0; - Path funcPath = new Path(dbDumpPath, ReplUtils.FUNCTIONS_ROOT_DIR_NAME); + Path funcPath = new Path(metadataPath, ReplUtils.FUNCTIONS_ROOT_DIR_NAME); if (fs.exists(funcPath)) { numFunctions = getSubDirs(fs, funcPath).length; } diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/DatabaseEventsIterator.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/DatabaseEventsIterator.java index 72baee6881..04b880df13 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/DatabaseEventsIterator.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/DatabaseEventsIterator.java @@ -57,7 +57,7 @@ this.hiveConf = hiveConf; FileSystem fileSystem = dbLevelPath.getFileSystem(hiveConf); // this is only there for the use case where we are doing table only replication and not database level - if (!fileSystem.exists(new Path(dbLevelPath + Path.SEPARATOR + EximUtil.METADATA_NAME))) { + if (!fileSystem.exists(new Path(new Path(dbLevelPath, EximUtil.METADATA_PATH_NAME), EximUtil.METADATA_NAME))) { databaseEventProcessed = true; } @@ -80,7 +80,7 @@ public LocatedFileStatus next() throws IOException { } }; } else { - remoteIterator = fileSystem.listFiles(dbLevelPath, true); + remoteIterator = fileSystem.listFiles(new Path(dbLevelPath, EximUtil.METADATA_PATH_NAME), true); } } @@ -115,7 +115,7 @@ Path dbLevelPath() { public boolean hasNext() { try { if (!databaseEventProcessed) { - next = dbLevelPath; + next = new Path(dbLevelPath, EximUtil.METADATA_PATH_NAME); return true; } @@ -129,7 +129,8 @@ public boolean hasNext() { continue; } if (next.getPath().toString().endsWith(EximUtil.METADATA_NAME)) { - String replacedString = next.getPath().toString().replace(dbLevelPath.toString(), ""); + String replacedString = next.getPath().toString() + .replace(new Path(dbLevelPath, EximUtil.METADATA_PATH_NAME).toString(), ""); List filteredNames = Arrays.stream(replacedString.split(Path.SEPARATOR)) .filter(StringUtils::isNotBlank) .collect(Collectors.toList()); @@ -174,7 +175,8 @@ public BootstrapEvent next() { LOG.debug("functions directory: {}", next.toString()); return postProcessing(new FSFunctionEvent(next)); } - return postProcessing(new FSTableEvent(hiveConf, next.toString())); + return postProcessing(new FSTableEvent(hiveConf, next.toString(), + new Path(new Path(dbLevelPath, EximUtil.DATA_PATH_NAME), next.getName()).toString())); } private BootstrapEvent postProcessing(BootstrapEvent bootstrapEvent) { @@ -187,11 +189,14 @@ private BootstrapEvent postProcessing(BootstrapEvent bootstrapEvent) { private BootstrapEvent eventForReplicationState() { if (replicationState.partitionState != null) { BootstrapEvent - bootstrapEvent = new FSPartitionEvent(hiveConf, previous.toString(), replicationState); + bootstrapEvent = new FSPartitionEvent(hiveConf, previous.toString(), + new Path(new Path(dbLevelPath, EximUtil.DATA_PATH_NAME), previous.getName()).toString(), + replicationState); replicationState = null; return bootstrapEvent; } else if (replicationState.lastTableReplicated != null) { - FSTableEvent event = new FSTableEvent(hiveConf, previous.toString()); + FSTableEvent event = new FSTableEvent(hiveConf, previous.toString(), + new Path(new Path(dbLevelPath, EximUtil.DATA_PATH_NAME), previous.getName()).toString()); replicationState = null; return event; } diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSPartitionEvent.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSPartitionEvent.java index a79f5b7123..2d8240848a 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSPartitionEvent.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSPartitionEvent.java @@ -34,9 +34,9 @@ private final ReplicationState replicationState; private final TableEvent tableEvent; - FSPartitionEvent(HiveConf hiveConf, String metadataDir, + FSPartitionEvent(HiveConf hiveConf, String metadataDir, String dataDir, ReplicationState replicationState) { - tableEvent = new FSTableEvent(hiveConf, metadataDir); + tableEvent = new FSTableEvent(hiveConf, metadataDir, dataDir); this.replicationState = replicationState; } @@ -87,4 +87,9 @@ public boolean shouldNotReplicate() { public Path metadataPath() { return tableEvent.metadataPath(); } + + @Override + public Path dataPath() { + return tableEvent.dataPath(); + } } diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSTableEvent.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSTableEvent.java index 6d38c0309f..cd3d619332 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSTableEvent.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/events/filesystem/FSTableEvent.java @@ -52,16 +52,19 @@ import static org.apache.hadoop.hive.ql.util.HiveStrictManagedMigration.getHiveUpdater; public class FSTableEvent implements TableEvent { - private final Path fromPath; + private final Path fromPathMetadata; + private final Path fromPathData; private final MetaData metadata; private final HiveConf hiveConf; - FSTableEvent(HiveConf hiveConf, String metadataDir) { + FSTableEvent(HiveConf hiveConf, String metadataDir, String dataDir) { try { URI fromURI = EximUtil.getValidatedURI(hiveConf, PlanUtils.stripQuotes(metadataDir)); - fromPath = new Path(fromURI.getScheme(), fromURI.getAuthority(), fromURI.getPath()); + fromPathMetadata = new Path(fromURI.getScheme(), fromURI.getAuthority(), fromURI.getPath()); + URI fromURIData = EximUtil.getValidatedURI(hiveConf, PlanUtils.stripQuotes(dataDir)); + fromPathData = new Path(fromURIData.getScheme(), fromURIData.getAuthority(), fromURIData.getPath()); FileSystem fs = FileSystem.get(fromURI, hiveConf); - metadata = EximUtil.readMetaData(fs, new Path(fromPath, EximUtil.METADATA_NAME)); + metadata = EximUtil.readMetaData(fs, new Path(fromPathMetadata, EximUtil.METADATA_NAME)); this.hiveConf = hiveConf; } catch (Exception e) { throw new RuntimeException(e); @@ -82,7 +85,12 @@ public boolean shouldNotReplicate() { @Override public Path metadataPath() { - return fromPath; + return fromPathMetadata; + } + + @Override + public Path dataPath() { + return fromPathData; } /** @@ -150,7 +158,7 @@ public ImportTableDesc tableDesc(String dbName) throws SemanticException { //TODO: if partitions are loaded lazily via the iterator then we will have to avoid conversion of everything here as it defeats the purpose. for (Partition partition : metadata.getPartitions()) { // TODO: this should ideally not create AddPartitionDesc per partition - AlterTableAddPartitionDesc partsDesc = addPartitionDesc(fromPath, tblDesc, partition); + AlterTableAddPartitionDesc partsDesc = addPartitionDesc(fromPathMetadata, tblDesc, partition); descs.add(partsDesc); } return descs; diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadPartitions.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadPartitions.java index 05a590a189..c2bed2ee12 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadPartitions.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadPartitions.java @@ -243,7 +243,7 @@ private void addPartition(boolean hasMorePartitions, AlterTableAddPartitionDesc Task copyTask = ReplCopyTask.getLoadCopyTask( event.replicationSpec(), - new Path(sourceWarehousePartitionLocation, EximUtil.DATA_PATH_NAME), + new Path(event.dataPath(), sourceWarehousePartitionLocation.getName()), stagingDir, context.hiveConf, false ); diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadTable.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadTable.java index 82a30319b5..bb20687f6f 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadTable.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/bootstrap/load/table/LoadTable.java @@ -42,7 +42,6 @@ import org.apache.hadoop.hive.ql.io.AcidUtils; import org.apache.hadoop.hive.ql.metadata.HiveException; import org.apache.hadoop.hive.ql.metadata.Table; -import org.apache.hadoop.hive.ql.parse.EximUtil; import org.apache.hadoop.hive.ql.parse.ImportSemanticAnalyzer; import org.apache.hadoop.hive.ql.parse.ReplicationSpec; import org.apache.hadoop.hive.ql.parse.SemanticException; @@ -223,7 +222,7 @@ private void newTableTasks(ImportTableDesc tblDesc, Task tblRootTask, TableLo if (shouldCreateLoadTableTask) { LOG.debug("adding dependent ReplTxnTask/CopyWork/MoveWork for table"); Task loadTableTask = loadTableTask(table, replicationSpec, new Path(tblDesc.getLocation()), - event.metadataPath()); + event.dataPath()); parentTask.addDependentTask(loadTableTask); } tracker.addTask(tblRootTask); @@ -272,7 +271,7 @@ static TableLocationTuple tableLocation(ImportTableDesc tblDesc, Database parent private Task loadTableTask(Table table, ReplicationSpec replicationSpec, Path tgtPath, Path fromURI) { - Path dataPath = new Path(fromURI, EximUtil.DATA_PATH_NAME); + Path dataPath = fromURI; Path tmpPath = tgtPath; // if move optimization is enabled, copy the files directly to the target path. No need to create the staging dir. diff --git a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/util/ReplUtils.java b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/util/ReplUtils.java index 211c3f014d..939cbc3a35 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/util/ReplUtils.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/exec/repl/util/ReplUtils.java @@ -94,10 +94,6 @@ // Configuration to enable/disable dumping ACID tables. Used only for testing and shouldn't be // seen in production or in case of tests other than the ones where it's required. public static final String REPL_DUMP_INCLUDE_ACID_TABLES = "hive.repl.dump.include.acid.tables"; - //Acknowledgement for repl dump complete - public static final String DUMP_ACKNOWLEDGEMENT = "_finished_dump"; - //Acknowledgement for repl load complete - public static final String LOAD_ACKNOWLEDGEMENT = "_finished_load"; /** * Bootstrap REPL LOAD operation type on the examined object based on ckpt state. */ @@ -240,7 +236,8 @@ public static PathFilter getEventsDirectoryFilter(final FileSystem fs) { try { return fs.isDirectory(p) && !p.getName().equalsIgnoreCase(ReplUtils.INC_BOOTSTRAP_ROOT_DIR_NAME) && !p.getName().equalsIgnoreCase(ReplUtils.REPL_TABLE_LIST_DIR_NAME) - && !p.getName().equalsIgnoreCase(EximUtil.DATA_PATH_NAME); + && !p.getName().equalsIgnoreCase(EximUtil.DATA_PATH_NAME) + && !p.getName().equalsIgnoreCase(EximUtil.METADATA_PATH_NAME); } catch (IOException e) { throw new RuntimeException(e); } @@ -251,7 +248,8 @@ public static PathFilter getBootstrapDirectoryFilter(final FileSystem fs) { return p -> { try { return fs.isDirectory(p) && !p.getName().equalsIgnoreCase(ReplUtils.REPL_TABLE_LIST_DIR_NAME) - && !p.getName().equalsIgnoreCase(EximUtil.DATA_PATH_NAME); + && !p.getName().equalsIgnoreCase(EximUtil.DATA_PATH_NAME) + && !p.getName().equalsIgnoreCase(EximUtil.METADATA_PATH_NAME); } catch (IOException e) { throw new RuntimeException(e); } diff --git a/ql/src/java/org/apache/hadoop/hive/ql/parse/EximUtil.java b/ql/src/java/org/apache/hadoop/hive/ql/parse/EximUtil.java index bc90ea1db7..11148f5222 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/parse/EximUtil.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/parse/EximUtil.java @@ -51,6 +51,7 @@ import org.slf4j.LoggerFactory; import java.io.IOException; +import java.io.Serializable; import java.net.URI; import java.net.URISyntaxException; import java.nio.charset.StandardCharsets; @@ -74,6 +75,7 @@ public static final String METADATA_NAME = "_metadata"; public static final String FILES_NAME = "_files"; public static final String DATA_PATH_NAME = "data"; + public static final String METADATA_PATH_NAME = "metadata"; private static final Logger LOG = LoggerFactory.getLogger(EximUtil.class); @@ -159,7 +161,8 @@ public void setOpenTxnTask(Task openTxnTask) { /** * Wrapper class for mapping source and target path for copying managed table data. */ - public static class ManagedTableCopyPath { + public static class ManagedTableCopyPath implements Serializable { + private static final long serialVersionUID = 1L; private ReplicationSpec replicationSpec; private static boolean nullSrcPathForTest = false; private Path srcPath; diff --git a/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSemanticAnalyzer.java b/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSemanticAnalyzer.java index 8802139e84..c4ff070da6 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSemanticAnalyzer.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSemanticAnalyzer.java @@ -32,6 +32,7 @@ import org.apache.hadoop.hive.ql.QueryState; import org.apache.hadoop.hive.ql.exec.Task; import org.apache.hadoop.hive.ql.exec.TaskFactory; +import org.apache.hadoop.hive.ql.exec.repl.ReplAck; import org.apache.hadoop.hive.ql.exec.repl.ReplDumpWork; import org.apache.hadoop.hive.ql.exec.repl.ReplLoadWork; import org.apache.hadoop.hive.ql.exec.repl.util.ReplUtils; @@ -54,6 +55,7 @@ import static org.apache.hadoop.hive.conf.HiveConf.ConfVars.REPL_DUMP_METADATA_ONLY; import static org.apache.hadoop.hive.conf.HiveConf.ConfVars.REPL_ENABLE_MOVE_OPTIMIZATION; import static org.apache.hadoop.hive.conf.HiveConf.ConfVars.REPL_MOVE_OPTIMIZED_FILE_SCHEMES; +import static org.apache.hadoop.hive.ql.exec.repl.ReplAck.LOAD_ACKNOWLEDGEMENT; import static org.apache.hadoop.hive.ql.parse.HiveParser.TOK_DBNAME; import static org.apache.hadoop.hive.ql.parse.HiveParser.TOK_REPLACE; import static org.apache.hadoop.hive.ql.parse.HiveParser.TOK_REPL_CONFIG; @@ -424,8 +426,9 @@ private Path getCurrentLoadPath() throws IOException, SemanticException { } } Path hiveDumpPath = new Path(latestUpdatedStatus.getPath(), ReplUtils.REPL_HIVE_BASE_DIR); - if (loadPathBase.getFileSystem(conf).exists(new Path(hiveDumpPath, ReplUtils.DUMP_ACKNOWLEDGEMENT)) - && !loadPathBase.getFileSystem(conf).exists(new Path(hiveDumpPath, ReplUtils.LOAD_ACKNOWLEDGEMENT))) { + if (loadPathBase.getFileSystem(conf).exists(new Path(hiveDumpPath, + ReplAck.DUMP_ACKNOWLEDGEMENT.toString())) + && !loadPathBase.getFileSystem(conf).exists(new Path(hiveDumpPath, LOAD_ACKNOWLEDGEMENT.toString()))) { return hiveDumpPath; } } diff --git a/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSpec.java b/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSpec.java index 5c8d0edd77..70607846ef 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSpec.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/parse/ReplicationSpec.java @@ -24,7 +24,7 @@ import org.apache.hadoop.hive.ql.plan.PlanUtils; import javax.annotation.Nullable; -import java.text.Collator; +import java.io.Serializable; import java.util.Map; /** @@ -35,8 +35,8 @@ * Typically, this corresponds to the replicationClause definition * in the parser. */ -public class ReplicationSpec { - +public class ReplicationSpec implements Serializable { + private static final long serialVersionUID = 1L; private boolean isInReplicationScope = false; // default is that it's not in a repl scope private boolean isMetadataOnly = false; // default is full export/import, not metadata-only private String eventId = null; diff --git a/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/PartitionExport.java b/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/PartitionExport.java index c3b10815f6..73dc606d87 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/PartitionExport.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/PartitionExport.java @@ -23,7 +23,6 @@ import org.apache.hadoop.hive.ql.metadata.HiveException; import org.apache.hadoop.hive.ql.metadata.Partition; import org.apache.hadoop.hive.ql.metadata.PartitionIterable; -import org.apache.hadoop.hive.ql.parse.EximUtil; import org.apache.hadoop.hive.ql.parse.EximUtil.ManagedTableCopyPath; import org.apache.hadoop.hive.ql.parse.ReplicationSpec; import org.apache.hadoop.hive.ql.parse.repl.dump.io.FileOperations; @@ -118,12 +117,13 @@ // this the data copy List dataPathList = Utils.getDataPathList(partition.getDataLocation(), forReplicationSpec, hiveConf); - Path rootDataDumpDir = paths.partitionExportDir(partitionName); + Path rootDataDumpDir = paths.partitionMetadataExportDir(partitionName); new FileOperations(dataPathList, rootDataDumpDir, distCpDoAsUser, hiveConf, mmCtx) .export(isExportTask); + Path dataDumpDir = new Path(paths.dataExportRootDir(), partitionName); LOG.debug("Thread: {}, finish partition dump {}", threadName, partitionName); return new ManagedTableCopyPath(forReplicationSpec, partition.getDataLocation(), - new Path(rootDataDumpDir, EximUtil.DATA_PATH_NAME)); + dataDumpDir); } catch (Exception e) { throw new RuntimeException(e.getMessage(), e); } diff --git a/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/TableExport.java b/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/TableExport.java index 683f3c0362..1640e8787d 100644 --- a/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/TableExport.java +++ b/ql/src/java/org/apache/hadoop/hive/ql/parse/repl/dump/TableExport.java @@ -198,22 +198,30 @@ private boolean shouldExport() { public static class Paths { private final String astRepresentationForErrorMsg; private final HiveConf conf; - //variable access should not be done and use exportRootDir() instead. - private final Path _exportRootDir; + //metadataExportRootDir and dataExportRootDir variable access should not be done and use + // metadataExportRootDir() and dataExportRootDir() instead. + private final Path metadataExportRootDir; + private final Path dataExportRootDir; private final FileSystem exportFileSystem; - private boolean writeData, exportRootDirCreated = false; + private boolean writeData, metadataExportRootDirCreated = false, dataExportRootDirCreated = false; public Paths(String astRepresentationForErrorMsg, Path dbRoot, String tblName, HiveConf conf, boolean shouldWriteData) throws SemanticException { this.astRepresentationForErrorMsg = astRepresentationForErrorMsg; this.conf = conf; this.writeData = shouldWriteData; - Path tableRoot = new Path(dbRoot, tblName); - URI exportRootDir = EximUtil.getValidatedURI(conf, tableRoot.toUri().toString()); - validateTargetDir(exportRootDir); - this._exportRootDir = new Path(exportRootDir); + Path metadataRoot = new Path(dbRoot, EximUtil.METADATA_PATH_NAME); + Path dataRoot = new Path(dbRoot, EximUtil.DATA_PATH_NAME); + Path tableRootForMetadataDump = new Path(metadataRoot, tblName); + Path tableRootForDataDump = new Path(dataRoot, tblName); + URI metadataExportRootDir = EximUtil.getValidatedURI(conf, tableRootForMetadataDump.toUri().toString()); + validateTargetDir(metadataExportRootDir); + URI dataExportRootDir = EximUtil.getValidatedURI(conf, tableRootForDataDump.toUri().toString()); + validateTargetDataDir(dataExportRootDir); + this.metadataExportRootDir = new Path(metadataExportRootDir); + this.dataExportRootDir = new Path(dataExportRootDir); try { - this.exportFileSystem = this._exportRootDir.getFileSystem(conf); + this.exportFileSystem = this.metadataExportRootDir.getFileSystem(conf); } catch (IOException e) { throw new SemanticException(e); } @@ -223,37 +231,58 @@ public Paths(String astRepresentationForErrorMsg, String path, HiveConf conf, boolean shouldWriteData) throws SemanticException { this.astRepresentationForErrorMsg = astRepresentationForErrorMsg; this.conf = conf; - this._exportRootDir = new Path(EximUtil.getValidatedURI(conf, path)); + this.metadataExportRootDir = new Path(EximUtil.getValidatedURI(conf, path)); + this.dataExportRootDir = new Path(EximUtil.getValidatedURI(conf, path)); this.writeData = shouldWriteData; try { - this.exportFileSystem = _exportRootDir.getFileSystem(conf); + this.exportFileSystem = metadataExportRootDir.getFileSystem(conf); } catch (IOException e) { throw new SemanticException(e); } } - Path partitionExportDir(String partitionName) throws SemanticException { - return exportDir(new Path(exportRootDir(), partitionName)); + Path partitionMetadataExportDir(String partitionName) throws SemanticException { + return exportDir(new Path(metadataExportRootDir(), partitionName)); } /** - * Access to the {@link #_exportRootDir} should only be done via this method + * Access to the {@link #metadataExportRootDir} should only be done via this method * since the creation of the directory is delayed until we figure out if we want * to write something or not. This is specifically important to prevent empty non-native * directories being created in repl dump. */ - public Path exportRootDir() throws SemanticException { - if (!exportRootDirCreated) { + public Path metadataExportRootDir() throws SemanticException { + if (!metadataExportRootDirCreated) { try { - if (!exportFileSystem.exists(this._exportRootDir) && writeData) { - exportFileSystem.mkdirs(this._exportRootDir); + if (!exportFileSystem.exists(this.metadataExportRootDir) && writeData) { + exportFileSystem.mkdirs(this.metadataExportRootDir); } - exportRootDirCreated = true; + metadataExportRootDirCreated = true; } catch (IOException e) { throw new SemanticException(e); } } - return _exportRootDir; + return metadataExportRootDir; + } + + /** + * Access to the {@link #dataExportRootDir} should only be done via this method + * since the creation of the directory is delayed until we figure out if we want + * to write something or not. This is specifically important to prevent empty non-native + * directories being created in repl dump. + */ + public Path dataExportRootDir() throws SemanticException { + if (!dataExportRootDirCreated) { + try { + if (!exportFileSystem.exists(this.dataExportRootDir) && writeData) { + exportFileSystem.mkdirs(this.dataExportRootDir); + } + dataExportRootDirCreated = true; + } catch (IOException e) { + throw new SemanticException(e); + } + } + return dataExportRootDir; } private Path exportDir(Path exportDir) throws SemanticException { @@ -269,7 +298,7 @@ private Path exportDir(Path exportDir) throws SemanticException { } private Path metaDataExportFile() throws SemanticException { - return new Path(exportRootDir(), EximUtil.METADATA_NAME); + return new Path(metadataExportRootDir(), EximUtil.METADATA_NAME); } /** @@ -277,7 +306,7 @@ private Path metaDataExportFile() throws SemanticException { * Partition's data export directory is created within the export semantics of partition. */ private Path dataExportDir() throws SemanticException { - return exportDir(new Path(exportRootDir(), EximUtil.DATA_PATH_NAME)); + return exportDir(dataExportRootDir()); } /** @@ -310,6 +339,30 @@ private void validateTargetDir(URI rootDirExportFile) throws SemanticException { throw new SemanticException(astRepresentationForErrorMsg, e); } } + + /** + * this level of validation might not be required as the root directory in which we dump will + * be different for each run hence possibility of it having data is not there. + */ + private void validateTargetDataDir(URI rootDirExportFile) throws SemanticException { + try { + FileSystem fs = FileSystem.get(rootDirExportFile, conf); + Path toPath = new Path(rootDirExportFile.getScheme(), rootDirExportFile.getAuthority(), + rootDirExportFile.getPath()); + try { + FileStatus tgt = fs.getFileStatus(toPath); + // target exists + if (!tgt.isDirectory()) { + throw new SemanticException( + astRepresentationForErrorMsg + ": " + "Target is not a directory : " + + rootDirExportFile); + } + } catch (FileNotFoundException ignored) { + } + } catch (IOException e) { + throw new SemanticException(astRepresentationForErrorMsg, e); + } + } } public static class AuthEntities { @@ -343,7 +396,7 @@ public AuthEntities getAuthEntities() throws SemanticException { authEntities.inputs.add(new ReadEntity(tableSpec.tableHandle)); } } - authEntities.outputs.add(toWriteEntity(paths.exportRootDir(), conf)); + authEntities.outputs.add(toWriteEntity(paths.metadataExportRootDir(), conf)); } catch (Exception e) { throw new SemanticException(e); } diff --git a/ql/src/test/org/apache/hadoop/hive/ql/exec/repl/TestReplDumpTask.java b/ql/src/test/org/apache/hadoop/hive/ql/exec/repl/TestReplDumpTask.java index 81fac252a3..af3e7eae56 100644 --- a/ql/src/test/org/apache/hadoop/hive/ql/exec/repl/TestReplDumpTask.java +++ b/ql/src/test/org/apache/hadoop/hive/ql/exec/repl/TestReplDumpTask.java @@ -84,11 +84,11 @@ String getValidTxnListForReplDump(Hive hiveDb, long waitUntilTime) { } @Override - void dumpFunctionMetadata(String dbName, Path dumpRoot, Hive hiveDb) { + void dumpFunctionMetadata(String dbName, Path metadataRoot, Hive hiveDb) { } @Override - Path dumpDbMetadata(String dbName, Path dumpRoot, long lastReplId, Hive hiveDb) { + Path dumpDbMetadata(String dbName, Path metadataRoot, long lastReplId, Hive hiveDb) { return Mockito.mock(Path.class); } @@ -129,7 +129,7 @@ public void removeDBPropertyToPreventRenameWhenBootstrapDumpOfTableFails() throw @Override List dumpTable(String dbName, String tblName, String validTxnList, Path dbRoot, - Path replDataDir, long lastReplId, Hive hiveDb, + long lastReplId, Hive hiveDb, HiveWrapper.Tuple
tuple) throws Exception { tableDumpCount++;