Spark Load方式导入StarRocks指南
一、背景
本文主要阐述用Spark Load方式导入数据到StarRocks。
二、环境准备
2.1 部署依赖组件
这里对于Hadoop、Hive、Spark就不做详细的阐述了，大家可以使用公司自建的系统或者自己参考官方文档部署。
注：以下为本次测试使用的环境，$HOME为家目录，可根据自己实际环境调整（建议创建用户并使用/home/starrocks），另外本次测试为自己部署依赖环境，由于资源有限，会有多个服务混合部署，可根据自己实际环境调整参数
| 服务 | 节点 | 端口 | PATH |
|---|---|---|---|
| JDK | 172.26.194.2 172.26.194.3 172.26.194.4 |
N/A | $HOME/thirdparty/jdk1.8.0_202/ |
| Hadoop | namenode（172.26.194.2) resourcemanager（172.26.194.2) |
9002 8035 |
$HOME/thirdparty/hadoop-2.10.0 |
| Spark | master（172.26.194.2） | N/A | $HOME/thirdparty/spark-2.4.4-bin-hadoop2.7 |
| Hive | metastore（172.26.194.2） | 9083 | $HOME/apache-hive-2.3.7-bin |
2.2 部署StarRocks及其依赖
2.2.1 部署StarRocks
请参考 StarRocks手动部署，并建议按如下端口等配置进行部署。以下为本次测试使用的环境，starrocks部署路径为$HOME/starrocks-1.18.2，可根据自己实际环境调整$HOME
| 服务 | 节点 | 端口 | PATH |
|---|---|---|---|
| fe | 172.26.194.2 | http_port = 8030 rpc_port = 9020 query_port = 9030 edit_log_port = 9010 |
$HOME/starrocks-1.18.2/fe |
| be | 172.26.194.2 172.26.194.3 172.26.194.4 |
be_port = 9060 webserver_port = 8040 heartbeat_service_port = 9050 brpc_port = 8060 |
$HOME/starrocks-1.18.2/be |
| broker | 172.26.194.2 172.26.194.3 172.26.194.4 |
broker_ipc_port=8000 | $HOME/starrocks-1.18.2/broker |
2.2.2 配置Spark和YARN客户端
下面配置Spark和YARN客户端操作也可以在fe.conf中spark_home_default_dir指定spark客户端路径，spark_resource_path指定spark-2x.zip路径，yarn_client_path指定YARN客户端路径，详情请参考StarRocks Spark YARN客户端配置
mkdir /home/starrocks/starrocks-1.18.2/fe/lib/yarn-client
ln -s /home/starrocks/thirdparty/spark-2.4.4-bin-hadoop2.7 /home/starrocks/starrocks-1.18.2/fe/lib/spark2x
ln -s /home/starrocks/thirdparty/hadoop-2.10.0 /home/starrocks/starrocks-1.18.2/fe/lib/yarn-client
2.2.3 打包jar包
提交spark load任务的时候会提交该依赖包到hadoop上
cd /home/starrocks/starrocks-1.18.2/fe/lib/spark2x/jars
zip -q -r spark-2x.zip *
2.2.4 配置环境变量～/.bashrc
export JAVA_HOME=/home/starrocks/thirdparty/jdk1.8.0_202/
export HADOOP_HOME=/home/starrocks/thirdparty/hadoop-2.10.0
export SPARK_HOME=/home/starrocks/thirdparty/spark-2.4.4-bin-hadoop2.7
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export HIVE_HOME= /home/starrocks/thirdparty/apache-hive-2.3.7-bin
export HIVE_CONF_DIR=$HIVE_HOME/conf
2.2.5 配置JAVA_HOME
lib/yarn-client/hadoop/libexec/hadoop-config.sh文件第一行添加JAVA_HOME声明，否则会在获取application状态的时候报如下错误(fe.warn.log中)
2021-07-27 21:34:25,341 WARN (Load etl checker|47) [SparkEtlJobHandler.killEtlJob():297] yarn application kill failed. app id: application_1620958422319_5507, load job id: 17426, msg: which: no /home/starrocks/starrocks-1.18.2/fe/lib/yarn-client/hadoop/bin/yarn in ((null))
Error: JAVA_HOME is not set and could not be found.
export JAVA_HOME=/home/starrocks/thirdparty/jdk1.8.0_202/
2.3 生成模拟数据
模拟数据生成脚本gen_data.py，例如“python gen_data.py m n” 可生成m列n行随机数字
#!/usr/bin/env python
import sys
import random
import time
def genRand(s = 10000):
return random.randint(1,s)
def getLine(cols = 10):
tpl = "%s\t"
line = ""
for x in range(int(cols) -1):
line = line + tpl % genRand(x + 10)
line = line + str(genRand(int(cols) + 10))
return line
def getTable(lines = 10, cols = 10):
tpl = "%s\n"
table = ""
for x in range(int(lines) ):
table = table + tpl % getLine(cols)
return table.strip()
def main():
lines = sys.argv[1]
cols = sys.argv[2]
data = getTable(lines, cols)
print(data)
if __name__ == '__main__':
main()
三、Case模拟
3.1 case csv导入
3.1.1 生成测试数据
python gen_data.py 10000 2 > demo.csv #生成1w行两列的数据
3.1.2 上传至hdfs
创建目录和上传文件都需要有对应的权限，可依据自己实际环境调整
/home/starrocks/thirdparty/hadoop-2.10.0/bin/hadoop fs -mkdir /user/starrocks
/home/starrocks/thirdparty/hadoop-2.10.0/bin/hadoop fs -put demo.csv /user/starrocks/
3.1.3 StarRocks DDL
需要提前创建demo_db数据库（create database demo_db;）
CREATE TABLE `demo_db`.`demo_table1` (
`k1` varchar(50) NULL COMMENT "",
`v1` String NULL COMMENT ""
) ENGINE=OLAP
DUPLICATE KEY(`k1` )
COMMENT "OLAP"
DISTRIBUTED BY HASH(`v1` ) BUCKETS 3
PROPERTIES (
"replication_num" = "1"
);
3.1.4 功能测试
创建spark资源，同一个spark资源只需要创建一次
需提前部署broker，我这里broker名字为broker0，下文working_dir需提前在hdfs上创建对应的路径，具体路径以实际环境为准。需要启动starrocks集群的用户有working_dir写权限
CREATE EXTERNAL RESOURCE "spark_resource0"
PROPERTIES
(
"type" = "spark",
"spark.master" = "yarn",
"spark.submit.deployMode" = "cluster",
"spark.executor.memory" = "1g",
"spark.hadoop.yarn.resourcemanager.address" = "172.26.194.2:8035",
"spark.hadoop.fs.defaultFS" = "hdfs://172.26.194.2:9002",
"working_dir" = "hdfs://172.26.194.2:9002/user/starrocks/sparketl",
"broker" = "broker0"
);
启动load任务
USE demo_db;
LOAD LABEL demo_db.label1
(
DATA INFILE("hdfs://172.26.194.2:9002/user/starrocks/demo.csv")
INTO TABLE demo_table1
COLUMNS TERMINATED BY "\t"
)
WITH RESOURCE 'spark_resource0'
(
"spark.executor.memory" = "500m",
"spark.shuffle.compress" = "true",
"spark.driver.memory" = "1500m",
"spark.testing.memory"="536870912" #512M
)
PROPERTIES
(
"timeout" = "3600",
"max_filter_ratio" = "0.2"
);
结果：
*************************** 1. row ***************************
JobId: 17407
Label: label1
State: FINISHED
Progress: ETL:100%; LOAD:100%
Type: SPARK
EtlInfo: unselected.rows=0; dpp.abnorm.ALL=1; dpp.norm.ALL=10000
TaskInfo: cluster:spark; timeout(s):3600; max_filter_ratio:0.2
ErrorMsg: NULL
CreateTime: 2021-07-27 19:15:43
EtlStartTime: 2021-07-27 19:15:56
EtlFinishTime: 2021-07-27 19:16:10
LoadStartTime: 2021-07-27 19:16:10
LoadFinishTime: 2021-07-27 19:16:15
URL: http://172.26.194.2:8098/proxy/application_1620958422319_5474/
JobDetails: {"Unfinished backends":{"00000000-0000-0000-0000-000000000000":[]},"ScannedRows":10001,"TaskNumber":1,"All backends":{"00000000-0000-0000-0000-000000000000":[-1]},"FileNumber":1,"FileSize":42883}
mysql> select * from demo_table1 limit 5;
+------+------+
| k1 | v1 |
+------+------+
| 1 | 3 |
| 1 | 5 |
| 1 | 4 |
| 1 | 3 |
| 1 | 3 |
+------+------+
5 rows in set (0.05 sec)
3.2 case 源目不匹配导入
3.2.1 生成测试数据
利用上文3.1.1的测试数据即可
3.2.2 StarRocks DDL
建表字段新增了create_time和update_time
CREATE TABLE `demo_db`.`demo_table2` (
`k1` varchar(50) NULL COMMENT "",
`v1` String NULL COMMENT "",
`create_time` datetime NULL COMMENT "",
`update_time` datetime NULL COMMENT ""
) ENGINE=OLAP
DUPLICATE KEY(`k1` )
COMMENT "OLAP"
DISTRIBUTED BY HASH(`v1` ) BUCKETS 3
PROPERTIES (
"replication_num" = "1"
);
3.2.4 功能测试
使用 3.1.4 定义的spark资源
启动load任务
USE demo_db;
LOAD LABEL demo_db.label2
(
DATA INFILE("hdfs://172.26.194.2:9002/user/starrocks/demo.csv")
INTO TABLE demo_table2
COLUMNS TERMINATED BY "\t"
(k1,v1)
SET
(
create_time=current_timestamp(),
update_time=current_timestamp()
)
)
WITH RESOURCE 'spark_resource0'
(
"spark.executor.memory" = "500m",
"spark.shuffle.compress" = "true",
"spark.driver.memory" = "1500m",
"spark.testing.memory"="536870912" #512M
)
PROPERTIES
(
"timeout" = "3600",
"max_filter_ratio" = "0.2"
);
结果：
*************************** 1. row ***************************
JobId: 23122
Label: label2
State: FINISHED
Progress: ETL:100%; LOAD:100%
Type: SPARK
EtlInfo: unselected.rows=0; dpp.abnorm.ALL=0; dpp.norm.ALL=10000
TaskInfo: cluster:spark; timeout(s):3600; max_filter_ratio:0.2
ErrorMsg: NULL
CreateTime: 2021-07-27 20:37:29
EtlStartTime: 2021-07-27 20:37:41
EtlFinishTime: 2021-07-27 20:37:56
LoadStartTime: 2021-07-27 20:37:56
LoadFinishTime: 2021-07-27 20:37:58
URL: http://172.26.194.2:8098/proxy/application_1632749512570_0026/
JobDetails: {"Unfinished backends":{"00000000-0000-0000-0000-000000000000":[]},"ScannedRows":10000,"TaskNumber":1,"All backends":{"00000000-0000-0000-0000-000000000000":[-1]},"FileNumber":1,"FileSize":43532}
mysql> select * from demo_table2 limit 5;
+------+------+---------------------+---------------------+
| k1 | v1 | create_time | update_time |
+------+------+---------------------+---------------------+
| 1 | 9 | 2021-07-27 20:37:49 | 2021-07-27 20:37:49 |
| 1 | 9 | 2021-07-27 20:37:49 | 2021-07-27 20:37:49 |
| 1 | 10 | 2021-07-27 20:37:49 | 2021-07-27 20:37:49 |
| 1 | 9 | 2021-07-27 20:37:49 | 2021-07-27 20:37:49 |
| 1 | 9 | 2021-07-27 20:37:49 | 2021-07-27 20:37:49 |
+------+------+---------------------+---------------------+
5 rows in set (0.00 sec)
3.3 case hive orc导入
3.3.1 生成orc格式数据（如果使用已有orc数据，请跳过这一步）
/home/starrocks/thirdparty/spark-2.4.4-bin-hadoop2.7/bin/spark-shell
scala> sc.setLogLevel("ERROR")
scala> val df = spark.read.option("delimiter","\t").csv("hdfs://172.26.194.2:9002/user/starrocks/demo.csv").toDF("k1","v1")
scala> df.coalesce(1).write.orc("hdfs://172.26.194.2:9002/user/starrocks/demo.orc")
scala> spark.read.orc("hdfs://172.26.194.2:9002/user/starrocks/demo.orc").show(5)
+---+---+
| k1| v1|
+---+---+
| 4| 4|
| 8| 8|
| 10| 6|
| 6| 1|
| 2| 10|
+---+---+
only showing top 5 rows
3.3.2 Hive DDL
CREATE EXTERNAL TABLE `hive_orc`(
`k1` string,
`v1` string)
ROW FORMAT SERDE
'org.apache.hadoop.hive.ql.io.orc.OrcSerde'
STORED AS INPUTFORMAT
'org.apache.hadoop.hive.ql.io.orc.OrcInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.orc.OrcOutputFormat'
LOCATION
'hdfs://172.26.194.2:9002/user/starrocks/demo.orc'
TBLPROPERTIES ( 'orc.compression'='snappy');
3.3.3 StarRocks DDL
#新建hive resource
CREATE EXTERNAL RESOURCE "hive0"
PROPERTIES (
"type" = "hive",
"hive.metastore.uris" = "thrift://172.26.194.2:9083"
);
#新建starrocks hive外表
CREATE EXTERNAL TABLE hive_external_orc
(
k1 string,
v1 string
)
ENGINE=hive
properties (
"resource" = "hive0",
"database" = "default",
"table" = "hive_orc",
"hive.metastore.uris" = "thrift://172.26.194.2:9083"
);
#新建starrocks测试表
create table demo_table3 like demo_table1;
3.3.4 功能测试
USE demo_db;
LOAD LABEL demo_db.label3
(
DATA FROM TABLE hive_external_orc
INTO TABLE demo_table3
)
WITH RESOURCE 'spark_resource0'
(
"spark.executor.memory" = "1500m",
"spark.shuffle.compress" = "true",
"spark.driver.memory" = "1500m"
)
PROPERTIES
(
"timeout" = "3600",
"max_filter_ratio" = "0.2"
);
#结果
*************************** 1. row ***************************
JobId: 19028
Label: label3
State: FINISHED
Progress: ETL:100%; LOAD:100%
Type: SPARK
EtlInfo: unselected.rows=0; dpp.abnorm.ALL=0; dpp.norm.ALL=10000
TaskInfo: cluster:spark; timeout(s):3600; max_filter_ratio:0.2
ErrorMsg: NULL
CreateTime: 2021-07-27 21:22:52
EtlStartTime: 2021-07-27 21:23:06
EtlFinishTime: 2021-07-27 21:23:34
LoadStartTime: 2021-07-27 21:23:34
LoadFinishTime: 2021-07-27 21:23:36
URL: http://172.26.194.2:8098/proxy/application_1620958422319_6070/
JobDetails: {"Unfinished backends":{"00000000-0000-0000-0000-000000000000":[]},"ScannedRows":10000,"TaskNumber":1,"All backends":{"00000000-0000-0000-0000-000000000000":[-1]},"FileNumber":0,"FileSize":0}
1 row in set (0.00 sec)
mysql> select * from demo_table3 limit 5;
+------+------+
| k1 | v1 |
+------+------+
| 1 | 11 |
| 1 | 11 |
| 1 | 9 |
| 1 | 7 |
| 1 | 9 |
+------+------+
5 rows in set (0.00 sec)
3.4 case hive parquet导入
3.4.1 生成parquet格式数据（如果使用已有parquet数据，请跳过这一步）
/home/starrocks/thirdparty/spark-2.4.4-bin-hadoop2.7/bin/spark-shell
scala> sc.setLogLevel("ERROR")
scala> val df = spark.read.option("delimiter","\t").csv("hdfs://172.26.194.2:9002/user/starrocks/demo.csv").toDF("k1","v1")
scala> df.coalesce(1).write.parquet("hdfs://172.26.194.2:9002/user/starrocks/demo.parquet")
scala> spark.read.parquet("hdfs://172.26.194.2:9002/user/starrocks/demo.parquet").show(5)
+---+---+
| k1| v1|
+---+---+
| 4| 4|
| 8| 8|
| 10| 6|
| 6| 1|
| 2| 10|
+---+---+
only showing top 5 rows
3.4.2 Hive DDL
CREATE EXTERNAL TABLE `hive_parquet`(
`k1` string,
`v1` string)
ROW FORMAT SERDE
'org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe'
STORED AS INPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.parquet.MapredParquetOutputFormat'
LOCATION
'hdfs://172.26.194.2:9002/user/starrocks/demo.parquet'
TBLPROPERTIES ( 'parquet.compression'='snappy');
3.4.3 StarRocks DDL
如之前已经创建hive resource，这里就不需要再次创建
create table demo_table4 like demo_table1;
#新建hive resource
CREATE EXTERNAL RESOURCE "hive0"
PROPERTIES (
"type" = "hive",
"hive.metastore.uris" = "thrift://172.26.194.2:9083"
);
#新建Starrocks hive外表
CREATE EXTERNAL TABLE hive_external_parquet
(
k1 string,
v1 string
)
ENGINE=hive
properties (
"resource" = "hive0",
"database" = "default",
"table" = "hive_parquet",
"hive.metastore.uris" = "thrift://172.26.194.2:9083"
);
3.4.4 功能测试
USE demo_db;
LOAD LABEL demo_db.label4
(
DATA FROM TABLE hive_external_parquet
INTO TABLE demo_table4
)
WITH RESOURCE 'spark_resource0'
(
"spark.executor.memory" = "1500m",
"spark.shuffle.compress" = "true",
"spark.driver.memory" = "1500m"
)
PROPERTIES
(
"timeout" = "3600",
"max_filter_ratio" = "0.2"
);
#结果
*************************** 1. row ***************************
JobId: 17424
Label: label4
State: FINISHED
Progress: ETL:100%; LOAD:100%
Type: SPARK
EtlInfo: unselected.rows=0; dpp.abnorm.ALL=0; dpp.norm.ALL=10000
TaskInfo: cluster:spark; timeout(s):3600; max_filter_ratio:0.2
ErrorMsg: NULL
CreateTime: 2021-07-27 21:17:48
EtlStartTime: 2021-07-27 21:17:59
EtlFinishTime: 2021-07-27 21:18:14
LoadStartTime: 2021-07-27 21:18:14
LoadFinishTime: 2021-07-27 21:18:18
URL: http://172.26.194.2:8098/proxy/application_1620958422319_5489/
JobDetails: {"Unfinished backends":{"00000000-0000-0000-0000-000000000000":[]},"ScannedRows":10000,"TaskNumber":1,"All backends":{"00000000-0000-0000-0000-000000000000":[-1]},"FileNumber":0,"FileSize":0}
1 row in set (0.00 sec)
mysql> select * from demo_table4 limit 5;
+------+------+
| k1 | v1 |
+------+------+
| 1 | 11 |
| 1 | 11 |
| 1 | 9 |
| 1 | 7 |
| 1 | 9 |
+------+------+
5 rows in set (0.02 sec)
四、排查问题方法
4.1 log/spark_launcher_log #有一些spark-submit失败的原因会在这里
4.2 log/fe.log #可能存在获取状态失败问题
4.3 /home/starrocks/thirdparty/hadoop-2.10.0/logs/userlogs/application_1620958422319_xxxx #作业具体失败原因
五、常见问题
5.1 yarn application status failed
yarn application status failed. spark app id: application_1620958422319_5441, load job id: 17389, timeout: 30000, msg: which: no /home/starrocks/starrocks-1.18.2-fe/fe-dd87506f-6e0b-4b33-8ebb-ce82952433a0/lib/yarn-client/hadoop/bin/yarn in ((null))
Error: JAVA_HOME is not set and could not be found
原因：
lib/yarn-client/hadoop/libexec/hadoop-config.sh中未识别JAVA_HOME
解决方案：
在lib/yarn-client/hadoop/libexec/hadoop-config.sh中声明JAVA_HOME
export JAVA_HOME=/home/starrocks/thirdparty/jdk1.8.0_202/
5.2 内存超限制
21/07/27 16:37:07 INFO yarn.Client: Verifying our application has not requested more than the maximum memory capability of the cluster (2048 MB per container)
Exception in thread "main" java.lang.IllegalArgumentException: Required executor memory (2048), overhead (384 MB), and PySpark memory (0 MB) is above the max threshold (2048 MB) of this cluster! Please check the values of 'yarn.scheduler.maximum-allocation-mb' and/or 'yarn.nodemanager.resource.memory-mb'.
at org.apache.spark.deploy.yarn.Client.verifyClusterResources(Client.scala:346)
at org.apache.spark.deploy.yarn.Client.submitApplication(Client.scala:176)
at org.apache.spark.deploy.yarn.Client.run(Client.scala:1135)
at org.apache.spark.deploy.yarn.YarnClusterApplication.start(Client.scala:1527)
at org.apache.spark.deploy.SparkSubmit.org$apache$spark$deploy$SparkSubmit$$runMain(SparkSubmit.scala:845)
at org.apache.spark.deploy.SparkSubmit.doRunMain$1(SparkSubmit.scala:161)
at org.apache.spark.deploy.SparkSubmit.submit(SparkSubmit.scala:184)
at org.apache.spark.deploy.SparkSubmit.doSubmit(SparkSubmit.scala:86)
at org.apache.spark.deploy.SparkSubmit$$anon$2.doSubmit(SparkSubmit.scala:920)
at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:929)
at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)
21/07/27 16:37:07 INFO util.ShutdownHookManager: Shutdown hook called
原因：
driver内存超过yarn的单个container分配的最大内存，我们测试环境是2048M；spark启动程序的时候需要额外分配非堆内内存用于用于考虑 VM 开销、内部字符串、其他本机开销等开销，默认最小为348M，本次测试配置的jvm为1800M，1800+348=2148M>2048M，所以抛异常了，详细可参考spark.driver.memoryOverhead参数 http://spark.apache.org/docs/latest/configuration.html
5.3 spark内存配置
Exception in thread "main" java.lang.IllegalArgumentException: System memory 466092032 must be at least 471859200.Please increase heap size using the --driver-memory option or spark.driver.memory in Spark configuration.
at org.apache.spark.memory.UnifiedMemoryManager$.getMaxMemory(UnifiedMemoryManager.scala:217)
at org.apache.spark.memory.UnifiedMemoryManager$.apply(UnifiedMemoryManager.scala:199)
at org.apache.spark.SparkEnv$.create(SparkEnv.scala:330)
at org.apache.spark.SparkEnv$.createExecutorEnv(SparkEnv.scala:200)
at org.apache.spark.executor.CoarseGrainedExecutorBackend$$anonfun$run$1.apply$mcV$sp(CoarseGrain
edExecutorBackend.scala:221)
at org.apache.spark.deploy.SparkHadoopUtil$$anon$2.run(SparkHadoopUtil.scala:65)
at org.apache.spark.deploy.SparkHadoopUtil$$anon$2.run(SparkHadoopUtil.scala:64)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1698)
at org.apache.spark.deploy.SparkHadoopUtil.runAsSparkUser(SparkHadoopUtil.scala:64)
at org.apache.spark.executor.CoarseGrainedExecutorBackend$.run(CoarseGrainedExecutorBackend.scala
:188)
原因：
JVM申请的memory不够导致无法启动SparkContext，可调整spark.testing.memory避免，以下为源码可参考
解决方案：
load中添加参数
"spark.testing.memory"="536870912"
WITH RESOURCE 'spark'
(
"spark.executor.memory" = "500m",
"spark.shuffle.compress" = "true",
"spark.driver.memory" = "1500m",
"spark.testing.memory"="536870912"
)
PROPERTIES
(
"timeout" = "3600",
"max_filter_ratio" = "0.2"
);
5.4 Path does not exist
org.apache.spark.sql.AnalysisException: Path does not exist: hdfs://namenode:9002/user/starrocks/sparketl/jobs/10054/label1/17370/configs/jobconfig.json;
at org.apache.spark.sql.execution.datasources.DataSource$$anonfun$org$apache$spark$sql$execution$datasources$Da
taSource$$checkAndGlobPathIfNecessary$1.apply(DataSource.scala:558)
at org.apache.spark.sql.execution.datasources.DataSource$$anonfun$org$apache$spark$sql$execution$datasources$Da
taSource$$checkAndGlobPathIfNecessary$1.apply(DataSource.scala:545)
at scala.collection.TraversableLike$$anonfun$flatMap$1.apply(TraversableLike.scala:241)
at scala.collection.TraversableLike$$anonfun$flatMap$1.apply(TraversableLike.scala:241)
at scala.collection.immutable.List.foreach(List.scala:392)
at scala.collection.TraversableLike$class.flatMap(TraversableLike.scala:241)
at scala.collection.immutable.List.flatMap(List.scala:355)
at org.apache.spark.sql.execution.datasources.DataSource.org$apache$spark$sql$execution$datasources$DataSource$
$checkAndGlobPathIfNecessary(DataSource.scala:545)
原因：
任务失败主动清理目录导致
解决方案：
查看fe.warn.log或者fe.log具体的异常信息
5.5 Permission denied
Exception in thread "main" org.apache.hadoop.security.AccessControlException: Permission denied: user=starrocks, access=WRITE, inode="/user/starrocks":starrocks:supergroup:drwxr-xr-x
原因：
启动spark任务的时候会在指定的working_dir下上传依赖，启动spark任务的用户（即starrocks集群启动的用户，这里启动集群用户是starrocks）没有权限写working_dir目录导致的
解决方案：
./bin/hadoop fs -chmod 777 /user/starrocks
