Apache Flink 1.12.0 on Yarn(3.1.1) 所遇到的問題
新搭建的FLINK集群出现的问题汇总
1.新搭建的Flink集群和Hadoop集群无法正常启动Flink任务
查看这个提交任务的日志无法发现有用的错误信息。
进一步查看yarn日志：
发现只有JobManager的错误日志出现了如下的错误：/bin/bash: /bin/java: No such file or directory。
正常情况下执行配置完成java之后,执行/bin/java的会出现如下的结果：
根据查到的提示，出现这个情况（/bin/bash: /bin/java: No such file or directory）的原因是软连接的问题。因此需要在每个节点都创建软连接：ls -s /usr/java/jdk1.8.0_221 /bin/java
每个节点创建完软连接之后，再次执行：/bin/java结果如下：
每个节点的软连接已生效，再次提交任务成功。
概要
根據官方文檔配置在 $FLINK_HOME/lib 加入 flink-shaded-hadoop-3-uber-3.1.1.7.1.1.0-565-9.0.jar ，經過驗證，其實這個可以不加，只加上下面的 hadoop classpath 就行。
或者在環境變量配置文件中 加入 hadoop classpath.
## 注意：lib 後面一定要加 *
export Hadoop_CLASSPATH=$Hadoop_CLASSPATH:$HADOOP_HOME/lib/*
export HADOOP_CLASSPATH=`hadoop classpath`
问题1
启动 yarn-session.sh 出現 Exit code: 127 Stack trace: ExitCodeException exitCode=127，具體的錯誤日誌如下：
2023-11-01 14:26:44,408 ERROR org.apache.flink.yarn.cli.FlinkYarnSessionCli [] - Error while running the Flink session.
org.apache.flink.client.deployment.ClusterDeploymentException: Couldn't deploy Yarn session cluster
at org.apache.flink.yarn.YarnClusterDescriptor.deploySessionCluster(YarnClusterDescriptor.java:411) ~[flink-dist_2.11-1.12.0.jar:1.12.0]
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.run(FlinkYarnSessionCli.java:498) ~[flink-dist_2.11-1.12.0.jar:1.12.0]
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.lambda$main$4(FlinkYarnSessionCli.java:730) ~[flink-dist_2.11-1.12.0.jar:1.12.0]
at java.security.AccessController.doPrivileged(Native Method) ~[?:1.8.0_221]
at javax.security.auth.Subject.doAs(Subject.java:422) ~[?:1.8.0_221]
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1729) ~[hadoop-common-3.1.1.jar:?]
at org.apache.flink.runtime.security.contexts.HadoopSecurityContext.runSecured(HadoopSecurityContext.java:41) ~[flink-dist_2.11-1.12.0.jar:1.12.0]
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.main(FlinkYarnSessionCli.java:730) [flink-dist_2.11-1.12.0.jar:1.12.0]
Caused by: org.apache.flink.yarn.YarnClusterDescriptor$YarnDeploymentException: The YARN application unexpectedly switched to state FAILED during deployment.
Diagnostics from YARN: Application application_1617189748122_0017 failed 1 times (global limit =2; local limit is =1) due to AM Container for appattempt_1617189748122_0017_000001 exited with exitCode: 127
Failing this attempt.Diagnostics: [2023-11-01 14:26:44.107]Exception from container-launch.
Container id: container_1617189748122_0017_01_000001
Exit code: 127
[2023-11-01 14:26:44.108]Container exited with a non-zero exit code 127. Error file: prelaunch.err.
Last 4096 bytes of prelaunch.err :
[2023-11-01 14:26:44.109]Container exited with a non-zero exit code 127. Error file: prelaunch.err.
Last 4096 bytes of prelaunch.err :
For more detailed output, check the application tracking page: http://hadoop001:8088/cluster/app/application_1617189748122_0017 Then click on links to logs of each attempt.
. Failing the application.
If log aggregation is enabled on your cluster, use this command to further investigate the issue:
yarn logs -applicationId application_1617189748122_0017
at org.apache.flink.yarn.YarnClusterDescriptor.startAppMaster(YarnClusterDescriptor.java:1078) ~[flink-dist_2.11-1.12.0.jar:1.12.0]
at org.apache.flink.yarn.YarnClusterDescriptor.deployInternal(YarnClusterDescriptor.java:558) ~[flink-dist_2.11-1.12.0.jar:1.12.0]
at org.apache.flink.yarn.YarnClusterDescriptor.deploySessionCluster(YarnClusterDescriptor.java:404) ~[flink-dist_2.11-1.12.0.jar:1.12.0]
... 7 more
------------------------------------------------------------
The program finished with the following exception:
org.apache.flink.client.deployment.ClusterDeploymentException: Couldn't deploy Yarn session cluster
at org.apache.flink.yarn.YarnClusterDescriptor.deploySessionCluster(YarnClusterDescriptor.java:411)
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.run(FlinkYarnSessionCli.java:498)
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.lambda$main$4(FlinkYarnSessionCli.java:730)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1729)
at org.apache.flink.runtime.security.contexts.HadoopSecurityContext.runSecured(HadoopSecurityContext.java:41)
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.main(FlinkYarnSessionCli.java:730)
Caused by: org.apache.flink.yarn.YarnClusterDescriptor$YarnDeploymentException: The YARN application unexpectedly switched to state FAILED during deployment.
Diagnostics from YARN: Application application_1617189748122_0017 failed 1 times (global limit =2; local limit is =1) due to AM Container for appattempt_1617189748122_0017_000001 exited with exitCode: 127
Failing this attempt.Diagnostics: [2023-11-01 14:26:44.107]Exception from container-launch.
Container id: container_1617189748122_0017_01_000001
Exit code: 127
[2023-11-01 14:26:44.108]Container exited with a non-zero exit code 127. Error file: prelaunch.err.
Last 4096 bytes of prelaunch.err :
[2023-11-01 14:26:44.109]Container exited with a non-zero exit code 127. Error file: prelaunch.err.
Last 4096 bytes of prelaunch.err :
For more detailed output, check the application tracking page: http://hadoop001:8088/cluster/app/application_1617189748122_0017 Then click on links to logs of each attempt.
. Failing the application.
If log aggregation is enabled on your cluster, use this command to further investigate the issue:
yarn logs -applicationId application_1617189748122_0017
at org.apache.flink.yarn.YarnClusterDescriptor.startAppMaster(YarnClusterDescriptor.java:1078)
at org.apache.flink.yarn.YarnClusterDescriptor.deployInternal(YarnClusterDescriptor.java:558)
at org.apache.flink.yarn.YarnClusterDescriptor.deploySessionCluster(YarnClusterDescriptor.java:404)
... 7 more
2023-11-01 14:26:44,415 INFO org.apache.flink.yarn.YarnClusterDescriptor [] - Cancelling deployment from Deployment Failure Hook
2023-11-01 14:26:44,416 INFO org.apache.hadoop.yarn.client.RMProxy [] - Connecting to ResourceManager at hadoop001/192.168.100.100:8032
2023-11-01 14:26:44,418 INFO org.apache.flink.yarn.YarnClusterDescriptor [] - Killing YARN application
2023-11-01 14:26:44,429 INFO org.apache.hadoop.yarn.client.api.impl.YarnClientImpl [] - Killed application application_1617189748122_0017
2023-11-01 14:26:44,532 INFO org.apache.flink.yarn.YarnClusterDescriptor [] - Deleting files in hdfs://hadoop001:8020/user/hadoop/.flink/application_1617189748122_0017.
然後下載具體的 container 日誌:
yarn logs -applicationId application_1617189748122_0017 -containerId container_1617189748122_0017_01_000001 -out /tmp/
查看 container 日誌
LogAggregationType: AGGREGATED
====================================================================
LogType:jobmanager.err
LogLastModifiedTime:Thu Apr 01 14:26:45 +0800 2021
LogLength:48
LogContents:
/bin/bash: /bin/java: No such file or directory
End of LogType:jobmanager.err
*******************************************************************************
End of LogType:jobmanager.out
*******************************************************************************
Container: container_1617189748122_0017_01_000001 on hadoop001_53613
LogAggregationType: AGGREGATED
====================================================================
注意日誌中的，找不到 /bin/bash: /bin/java: No such file or directory
[hadoop@hadoop001 bin]$ echo $JAVA_HOME
/usr/java/jdk1.8.0_221
然後做一個軟連接
ls -s /usr/java/jdk1.8.0_221 /bin/java
问题2
啟動 yarn-session.sh 出現 Container exited with a non-zero exit code 126，具體的錯誤日誌如下：
org.apache.flink.client.deployment.ClusterDeploymentException: Couldn't deploy Yarn session cluster
at org.apache.flink.yarn.YarnClusterDescriptor.deploySessionCluster(YarnClusterDescriptor.java:411)
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.run(FlinkYarnSessionCli.java:498)
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.lambda$main$4(FlinkYarnSessionCli.java:730)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1729)
at org.apache.flink.runtime.security.contexts.HadoopSecurityContext.runSecured(HadoopSecurityContext.java:41)
at org.apache.flink.yarn.cli.FlinkYarnSessionCli.main(FlinkYarnSessionCli.java:730)
Caused by: org.apache.flink.yarn.YarnClusterDescriptor$YarnDeploymentException: The YARN application unexpectedly switched to state FAILED during deployment.
Diagnostics from YARN: Application application_1617189748122_0019 failed 1 times (global limit =2; local limit is =1) due to AM Container for appattempt_1617189748122_0019_000001 exited with exitCode: 126
Failing this attempt.Diagnostics: [2023-11-01 14:43:23.068]Exception from container-launch.
Container id: container_1617189748122_0019_01_000001
Exit code: 126
[2023-11-01 14:43:23.070]Container exited with a non-zero exit code 126. Error file: prelaunch.err.
Last 4096 bytes of prelaunch.err :
[2023-11-01 14:43:23.072]Container exited with a non-zero exit code 126. Error file: prelaunch.err.
Last 4096 bytes of prelaunch.err :
查看 container 的日誌情況:
[hadoop@hadoop001 flink-1.12.0]$ yarn logs -applicationId application_1617189748122_0019 -show_application_log_info
2023-11-01 15:09:07,880 INFO client.RMProxy: Connecting to ResourceManager at hadoop001/192.168.100.100:8032
Application State: Completed.
Container: container_1617189748122_0019_01_000001 on hadoop001_53613
下載 container 日誌，操作和上面問題 1 一樣。
查看報錯日誌
broken symlinks(find -L . -maxdepth 5 -type l -ls):
End of LogType:directory.info
*******************************************************************************
Container: container_1617189748122_0019_01_000001 on hadoop001_53613
LogAggregationType: AGGREGATED
====================================================================
LogType:jobmanager.err
LogLastModifiedTime:Thu Apr 01 14:43:24 +0800 2021
LogLength:37
LogContents:
/bin/bash: /bin/java: Is a directory
End of LogType:jobmanager.err
*******************************************************************************
注意：/bin/bash: /bin/java: Is a directory ，這個是關鍵日誌，經過排查發現是軟連接出現了錯誤。
[root@hadoop001 bin]# ln -s /usr/java/jdk1.8.0_221/bin/java /bin/java
[root@hadoop001 bin]#
[root@hadoop001 bin]#
[root@hadoop001 bin]# ll /bin/java
lrwxrwxrwx 1 root root 31 Apr 1 16:09 /bin/java -> /usr/java/jdk1.8.0_221/bin/java
[root@hadoop001 bin]#
[root@hadoop001 bin]# /bin/java -version
java version "1.8.0_221"
Java(TM) SE Runtime Environment (build 1.8.0_221-b11)
Java HotSpot(TM) 64-Bit Server VM (build 25.221-b11, mixed mode)
验证
啟動 …/bin/yarn-session.sh
如何查看正在运行的Yarn容器的日志？？
众所周知，flink on yarn 分为jobmanager的容器和taskmanager的容器。
1.yarn application -list
2.yarn applicationattempt -list <ApplicationId>
3.yarn container -list <Application AttemptId>
到了这个第3步，就能看到每个容器的访问的url ，分别对应着jobmanager的和taskmanager的,但是具体怎么区分是哪个taskmanager的就只能依靠ip去区分。 htpp就用 curl 进行访问，https就用curl -k进行访问。
4.访问的结果包含了6种不同类型日志的访问路径：
找到我们想看的日志的访问路径，然后访问，访问路径的最后的参数：-start=-4096代表了显示多少日志出来。
如果想查看完整的日志，就应该把这个数调的很大
5.为了查看方便，应该使用 > 的方式把访问url的日志的结果输出到日志文件中。
FLINK ON YARN提交方式详解
目前自己用到的：yarn-per-job和yarn-applicaiton
他们的执行方式是不同的，执行yarn-per-job需要执行flink文件。同样的yarn-application也需要执行flik文件。
./flink run -t yarn-per-job -d \
-p 1 \
-ynm test_env_job \
-yD rest.flamegraph.enabled=true \
-yD jobmanager.memory.process.size=1G \
-yD taskmanager.memory.process.size=2G \
-yD taskmanager.numberOfTaskSlots=1 \
-yD env.java.opts="-Denv=test" \
-c com.xingye.demo.TestTimer \
/cgroups_test/test/fk.jar
./flink run-application -t yarn-application -d \
-p 5 \
-ynm test_impala_job \
-D rest.flamegraph.enabled=true \
-D jobmanager.memory.process.size=2G \
-D taskmanager.memory.process.size=8G \
-D taskmanager.numberOfTaskSlots=5 \
-c com.xingye.demo.ImpalaDemo1 \
/tmp/test_flink_impala/fk.jar
通过两种命令的对比就发现区别：
flink run -t yarn-per-job
flink run-application -t yarn-application
还有需要注意的是 -y* 这个参数是特有的使用yarn的时候就能使用的参数，也就是说yarn-per-job能用，yarn-application也能用。
-yD和-D动态参数的意思，作用就是覆盖flink-conf.yaml文件中的默认配置。唯一不同的地方就在于-yD只能在使用yarn的时候指定动态参数，不能在其他模式使用比如kubernetes无法使用-yD参数。-D可以在不同的方式下指定动态参数，-D是一种更通用的指定动态参数的方式。
总结：
yarn-per-job和yarn-application运行的都是同一个文件，相同点在于都能使用yarn模式下特有的-y*的参数，并且都能使用-D动态参数。