本文总结了几个本人在使用 Carbondata 的时候遇到的几个问题及其解决办法。这里使用的环境是：Spark 2.1.0、Carbondata 1.2.0。

必须指定 HDFS nameservices

在初始化 CarbonSession 的时候，如果不指定 HDFS nameservices，在数据导入是没啥问题的；但是数据查询会出现相关数据找不到问题：

scala> val carbon = SparkSession.builder().tempnfig(sc.getConf).getOrCreateCarbonSession("hdfs:///user/iteblog/carb")

scala> carbon.sql("""CREATE TABLE temp.iteblog(id bigint) STORED BY 'carbondata'""")
17/11/09 16:20:58 AUDIT command.CreateTable: [www.iteblog.com][iteblog][Thread-1]Creating Table with Database name [temp] and Table name [iteblog]
17/11/09 16:20:58 WARN hive.HiveExternalCatalog: Couldn't find corresponding Hive SerDe for data source provider org.apache.spark.sql.CarbonSource. Persisting data source table `temp`.`iteblog` into Hive metastore in Spark SQL specific format, which is NOT tempmpatible with Hive.
17/11/09 16:20:59 AUDIT command.CreateTable: [www.iteblog.com][iteblog][Thread-1]Table created with Database name [temp] and Table name [iteblog]
res2: org.apache.spark.sql.DataFrame = []

scala> carbon.sql("insert overwrite table temp.iteblog select id from temp.mytable limit 10")
17/11/09 16:21:46 AUDIT rdd.CarbonDataRDDFactory: [www.iteblog.com][iteblog][Thread-1]Data load request has been received for table temp.iteblog
17/11/09 16:21:46 WARN util.CarbonDataProcessorUtil: main sort scope is set to LOCAL_SORT
17/11/09 16:23:03 AUDIT rdd.CarbonDataRDDFactory: [www.iteblog.com][iteblog][Thread-1]Data load is successful for temp.iteblog
res3: org.apache.spark.sql.DataFrame = []

scala> carbon.sql("select * from temp.iteblog limit 10").show(10,100)
17/11/09 16:23:15 WARN scheduler.TaskSetManager: Lost task 0.0 in stage 3.0 (TID 1011, static.iteblog.com, executor 2): java.lang.RuntimeException: java.io.FileNotFoundException: /user/iteblog/carb/temp/iteblog/Fact/Part0/Segment_0/part-0-0_batchno0-0-1510215706696.carbondata (No such file or directory)
  at org.apache.carbondata.tempre.indexstore.blockletindex.IndexWrapper.<init>(IndexWrapper.java:39)
  at org.apache.carbondata.tempre.scan.executor.impl.AbstractQueryExecutor.initQuery(AbstractQueryExecutor.java:141)
  at org.apache.carbondata.tempre.scan.executor.impl.AbstractQueryExecutor.getBlockExecutionInfos(AbstractQueryExecutor.java:216)
  at org.apache.carbondata.tempre.scan.executor.impl.VectorDetailQueryExecutor.execute(VectorDetailQueryExecutor.java:36)
  at org.apache.carbondata.spark.vectorreader.VectorizedCarbonRetemprdReader.initialize(VectorizedCarbonRetemprdReader.java:116)
  at org.apache.carbondata.spark.rdd.CarbonScanRDD.internalCompute(CarbonScanRDD.scala:229)
  at org.apache.carbondata.spark.rdd.CarbonRDD.tempmpute(CarbonRDD.scala:62)
  at org.apache.spark.rdd.RDD.tempmputeOrReadCheckpoint(RDD.scala:323)
  at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
  at org.apache.spark.rdd.MapPartitionsRDD.tempmpute(MapPartitionsRDD.scala:38)
  at org.apache.spark.rdd.RDD.tempmputeOrReadCheckpoint(RDD.scala:323)
  at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
  at org.apache.spark.rdd.MapPartitionsRDD.tempmpute(MapPartitionsRDD.scala:38)
  at org.apache.spark.rdd.RDD.tempmputeOrReadCheckpoint(RDD.scala:323)
  at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
  at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
  at org.apache.spark.scheduler.Task.run(Task.scala:99)
  at org.apache.spark.executor.ExecutorTaskRunner.run(Executor.scala:282)
  at java.util.tempncurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
  at java.util.tempncurrent.ThreadPoolExecutorWorker.run(ThreadPoolExecutor.java:617)
  at java.lang.Thread.run(Thread.java:745)
Caused by: java.io.FileNotFoundException: /user/iteblog/carb/temp/iteblog/Fact/Part0/Segment_0/part-0-0_batchno0-0-1510215706696.carbondata (No such file or directory)
  at java.io.FileInputStream.open(Native Method)
  at java.io.FileInputStream.<init>(FileInputStream.java:138)
  at java.io.FileInputStream.<init>(FileInputStream.java:93)
  at org.apache.carbondata.tempre.datastore.impl.FileFactory.getDataInputStream(FileFactory.java:128)
  at org.apache.carbondata.tempre.reader.ThriftReader.open(ThriftReader.java:77)
  at org.apache.carbondata.tempre.reader.CarbonHeaderReader.readHeader(CarbonHeaderReader.java:46)
  at org.apache.carbondata.tempre.util.DataFileFooterConverterV3.getSchema(DataFileFooterConverterV3.java:90)
  at org.apache.carbondata.tempre.util.CarbonUtil.readMetadatFile(CarbonUtil.java:925)
  at org.apache.carbondata.tempre.indexstore.blockletindex.IndexWrapper.<init>(IndexWrapper.java:37)
  ... 20 more

可以看出，如果创建 CarbonSession 的时候，如果不指定 HDFS nameservices，在数据导入是没啥问题的；查找的时候就会出现文件找不到。这个最直接的解决版本就是创建 CarbonSession 的时候指定 HDFS nameservices。针对这个问题一个改进措施是让 Carbondata 能够根据提供的 hadoop 配置信息自动补充 HDFS nameservices 信息。

不支持 tinyint 数据类型

scala> carbon.sql("""CREATE TABLE temp.iteblog(status tinyint) STORED BY 'carbondata'""")
org.apache.carbondata.spark.exception.MalformedCarbonCommandException: Unsupported data type: StructField(status,ByteType,true).getType
  at org.apache.spark.sql.parser.CarbonSpark2SqlParser$anonfungetFields1.apply(CarbonSpark2SqlParser.scala:427)
  at org.apache.spark.sql.parser.CarbonSpark2SqlParseranonfungetFields1.apply(CarbonSpark2SqlParser.scala:417)
  at scala.collection.TraversableLikeanonfunmap1.apply(TraversableLike.scala:234)
  at scala.collection.TraversableLikeanonfunmap1.apply(TraversableLike.scala:234)
  at scala.collection.immutable.List.foreach(List.scala:381)
  at scala.collection.TraversableLikeclass.map(TraversableLike.scala:234)
  at scala.collection.immutable.List.map(List.scala:285)
  at org.apache.spark.sql.parser.CarbonSpark2SqlParser.getFields(CarbonSpark2SqlParser.scala:417)
  at org.apache.spark.sql.parser.CarbonSqlAstBuilder.visitCreateTable(CarbonSparkSqlParser.scala:135)
  at org.apache.spark.sql.parser.CarbonSqlAstBuilder.visitCreateTable(CarbonSparkSqlParser.scala:72)
  at org.apache.spark.sql.catalyst.parser.SqlBaseParserCreateTableContext.accept(SqlBaseParser.java:578)
  at org.antlr.v4.runtime.tree.AbstractParseTreeVisitor.visit(AbstractParseTreeVisitor.java:42)
  at org.apache.spark.sql.catalyst.parser.AstBuilderanonfunvisitSingleStatement1.apply(AstBuilder.scala:66)
  at org.apache.spark.sql.catalyst.parser.AstBuilderanonfunvisitSingleStatement1.apply(AstBuilder.scala:66)
  at org.apache.spark.sql.catalyst.parser.ParserUtils.withOrigin(ParserUtils.scala:93)
  at org.apache.spark.sql.catalyst.parser.AstBuilder.visitSingleStatement(AstBuilder.scala:65)
  at org.apache.spark.sql.catalyst.parser.AbstractSqlParser$anonfunparsePlan1.apply(ParseDriver.scala:54)
  at org.apache.spark.sql.catalyst.parser.AbstractSqlParseranonfunparsePlan$1.apply(ParseDriver.scala:53)
  at org.apache.spark.sql.catalyst.parser.AbstractSqlParser.parse(ParseDriver.scala:82)
  at org.apache.spark.sql.parser.CarbonSparkSqlParser.parse(CarbonSparkSqlParser.scala:68)
  at org.apache.spark.sql.catalyst.parser.AbstractSqlParser.parsePlan(ParseDriver.scala:53)
  at org.apache.spark.sql.parser.CarbonSparkSqlParser.parsePlan(CarbonSparkSqlParser.scala:49)
  at org.apache.spark.sql.SparkSession.sql(SparkSession.scala:592)
  ... 50 elided

这是因为 Carbondata 目前不支持 tinyint 类型，Carbondata 目前支持的数据类型可以参见：http://carbondata.apache.org/supported-data-types-in-carbondata.html。但是奇怪的是 CARBONDATA-18 这里面已经解决了这个问题，不知道为啥到当前版本却不支持了。

添加分区出现NoSuchTableException

如果你使用 ALTER TABLE temp.iteblog ADD PARTITION('2017') 语句来添加分区，你会遇到下面的异常：

scala> carbon.sql("ALTER TABLE temp.iteblog ADD PARTITION('2012')")
org.apache.spark.sql.catalyst.analysis.NoSuchTableException: Table or view 'iteblog' not found in database 'default';
  at org.apache.spark.sql.hive.client.HiveClient$anonfungetTable1.apply(HiveClient.scala:76)
  at org.apache.spark.sql.hive.client.HiveClientanonfungetTable1.apply(HiveClient.scala:76)
  at scala.Option.getOrElse(Option.scala:121)
  at org.apache.spark.sql.hive.client.HiveClientclass.getTable(HiveClient.scala:76)
  at org.apache.spark.sql.hive.client.HiveClientImpl.getTable(HiveClientImpl.scala:78)
  at org.apache.spark.sql.hive.HiveExternalCatalog$anonfunorgapachesparksqlhiveHiveExternalCataloggetRawTable1.apply(HiveExternalCatalog.scala:110)
  at org.apache.spark.sql.hive.HiveExternalCatalog$anonfunorgapachesparksqlhiveHiveExternalCataloggetRawTable1.apply(HiveExternalCatalog.scala:110)
  at org.apache.spark.sql.hive.HiveExternalCatalog.withClient(HiveExternalCatalog.scala:95)
  at org.apache.spark.sql.hive.HiveExternalCatalog.orgapachesparksqlhive$HiveExternalCataloggetRawTable(HiveExternalCatalog.scala:109)
  at org.apache.spark.sql.hive.HiveExternalCataloganonfungetTable1.apply(HiveExternalCatalog.scala:601)
  at org.apache.spark.sql.hive.HiveExternalCatalog$anonfungetTable1.apply(HiveExternalCatalog.scala:601)
  at org.apache.spark.sql.hive.HiveExternalCatalog.withClient(HiveExternalCatalog.scala:95)
  at org.apache.spark.sql.hive.HiveExternalCatalog.getTable(HiveExternalCatalog.scala:600)
  at org.apache.spark.sql.hive.HiveMetastoreCatalog.lookupRelation(HiveMetastoreCatalog.scala:106)
  at org.apache.spark.sql.hive.HiveSessionCatalog.lookupRelation(HiveSessionCatalog.scala:69)
  at org.apache.spark.sql.hive.CarbonSessionCatalog.lookupRelation(CarbonSessionState.scala:83)
  at org.apache.spark.sql.internal.CatalogImpl.refreshTable(CatalogImpl.scala:461)
  at org.apache.spark.sql.execution.command.AlterTableSplitPartitionCommand.processSchema(carbonTableSchema.scala:283)
  at org.apache.spark.sql.execution.command.AlterTableSplitPartitionCommand.run(carbonTableSchema.scala:229)
  at org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResultlzycompute(commands.scala:58)
  at org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResult(commands.scala:56)
  at org.apache.spark.sql.execution.command.ExecutedCommandExec.doExecute(commands.scala:74)
  at org.apache.spark.sql.execution.SparkPlan$anonfunexecute1.apply(SparkPlan.scala:114)
  at org.apache.spark.sql.execution.SparkPlananonfunexecute1.apply(SparkPlan.scala:114)
  at org.apache.spark.sql.execution.SparkPlananonfunexecuteQuery1.apply(SparkPlan.scala:135)
  at org.apache.spark.rdd.RDDOperationScope.withScope(RDDOperationScope.scala:151)
  at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:132)
  at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:113)
  at org.apache.spark.sql.execution.QueryExecution.toRddlzycompute(QueryExecution.scala:87)
  at org.apache.spark.sql.execution.QueryExecution.toRdd(QueryExecution.scala:87)
  at org.apache.spark.sql.Dataset.<init>(Dataset.scala:185)
  at org.apache.spark.sql.Dataset.ofRows(Dataset.scala:64)
  at org.apache.spark.sql.SparkSession.sql(SparkSession.scala:592)
  ... 50 elided

运行上面的SQL语句，Carbondata 表相关的分区其实已经添加好了，但是通过 Spark 刷新表的相关信息就出错了。从出错的信息可以看出，虽然我们已经传递了表所在的 DB 相关信息，但是 Spark 的 catalyst 并没有获取到，这个 bug 是因为代码里面并没有将表数据相关信息传递给 catalyst， 这个 bug 还影响分区的 split 相关操作。不过此 bug 在 CARBONDATA-1593 里面已经解决。

insert overwrite 操作超过三次将会出现 NPE

如果你在导数的时候执行 insert overwrite 大于等于三次，那么恭喜你，你肯定会遇到下面的异常，如下：

虽然出现 NPE 异常，但是数据其实已经导到 Carbondata 相关表里面了。引起这个异常的原因其实是因为每次执行完 insert overwrite 操作的时候，都需要删除之前的数据（也就是Segment目录）。但是 Segment 目录存在重复删除，导致找不到相关目录所以出现了 NPE 异常。这个问题在 CARBONDATA-1486 解决了。

不支持超过32767个字符的列

如果你有一列数据长度大于32767（Short.MaxValue），并且 enable.unsafe.sort=true ，那么你往 Carbondata 表导数据的时候会出现异常，如下：

java.lang.NegativeArraySizeException 
  at org.apache.carbondata.processing.newflow.sort.unsafe.UnsafeCarbonRowPage.getRow(UnsafeCarbonRowPage.java:182) 
  at org.apache.carbondata.processing.newflow.sort.unsafe.holder.UnsafeInmemoryHolder.readRow(UnsafeInmemoryHolder.java:63) 
  at org.apache.carbondata.processing.newflow.sort.unsafe.merger.UnsafeSingleThreadFinalSortFilesMerger.startSorting(UnsafeSingleThreadFinalSortFilesMerger.java:114) 
  at org.apache.carbondata.processing.newflow.sort.unsafe.merger.UnsafeSingleThreadFinalSortFilesMerger.startFinalMerge(UnsafeSingleThreadFinalSortFilesMerger.java:81) 
  at org.apache.carbondata.processing.newflow.sort.impl.UnsafeParallelReadMergeSorterImpl.sort(UnsafeParallelReadMergeSorterImpl.java:105) 
  at org.apache.carbondata.processing.newflow.steps.SortProcessorStepImpl.execute(SortProcessorStepImpl.java:62) 
  at org.apache.carbondata.processing.newflow.steps.DataWriterProcessorStepImpl.execute(DataWriterProcessorStepImpl.java:87) 
  at org.apache.carbondata.processing.newflow.DataLoadExecutor.execute(DataLoadExecutor.java:51) 
  at org.apache.carbondata.spark.rdd.NewDataFrameLoaderRDD$anon2.<init>(NewCarbonDataLoadRDD.scala:442) 
  at org.apache.carbondata.spark.rdd.NewDataFrameLoaderRDD.internalCompute(NewCarbonDataLoadRDD.scala:405) 
  at org.apache.carbondata.spark.rdd.CarbonRDD.compute(CarbonRDD.scala:62) 
  at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323) 
  at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)

这是 Carbondata 设计的缺陷，目前没有办法解决这个问题，不过可以实现一个类似于 varchar(size) 的数据类型。

日期格式错误导致数据丢失

如果你将带有日期类型的数据导入到 Carbondata 表中，可能会出现数据丢失：

scala> carbon.sql("""CREATE TABLE temp.iteblog(dt DATE) STORED BY 'carbondata'""")
17/11/09 16:44:46 AUDIT temp.mand.CreateTable: [www.iteblog.com][iteblog][Thread-1]Creating Table with Database name [temp] and Table name [iteblog]
17/11/09 16:44:47 WARN hive.HiveExternalCatalog: Couldn't find temp.responding Hive SerDe for data source provider org.apache.spark.sql.CarbonSource. Persisting data source table `temp`.`iteblog` into Hive metastore in Spark SQL specific format, which is NOT temp.patible with Hive.
17/11/09 16:44:47 AUDIT temp.mand.CreateTable: [www.iteblog.com][iteblog][Thread-1]Table created with Database name [temp] and Table name [iteblog]
res1: org.apache.spark.sql.DataFrame = []

scala> carbon.sql("select dt from temp.mydate").show(10,100)
17/11/09 16:44:52 ERROR lzo.LzoCodec: Failed to load/initialize native-lzo library
+--------+                                                                      
|      dt|
+--------+
|20170509|
|20170511|
|20170507|
|20170504|
|20170502|
|20170506|
|20170501|
|20170508|
|20170510|
|20170505|
+--------+
only showing top 10 rows


scala> carbon.sql("insert into table temp.iteblog select dt from temp.mydate limit 10")
17/11/09 16:45:14 AUDIT rdd.CarbonDataRDDFactory: [www.iteblog.com][iteblog][Thread-1]Data load request has been received for table temp.iteblog
17/11/09 16:45:14 WARN util.CarbonDataProcessorUtil: main sort scope is set to LOCAL_SORT
17/11/09 16:45:16 AUDIT rdd.CarbonDataRDDFactory: [www.iteblog.com][iteblog][Thread-1]Data load is successful for temp.iteblog
res3: org.apache.spark.sql.DataFrame = []

scala> carbon.sql("select * from temp.iteblog limit 10").show(10,100)
+----+
|  dt|
+----+
|null|
|null|
|null|
|null|
|null|
|null|
|null|
|null|
|null|
|null|
+----+

这是因为 Carbondata 对数据类型(DATE)有默认的格式，由参数 carbon.date.format 控制，默认值是 yyyy-MM-dd。所以你使用 yyyy-MM-dd 格式去解析 20170505 数据肯定会出现错误，从而导致数据丢失了。同理，时间戳类型(TIMESTAMP) 也有默认的格式，由参数 carbon.timestamp.format 空值，默认值为 yyyy-MM-dd HH:mm:ss。