Flink部署、执行模式

Flink的部署模式

本地模式、Standalone模式和FlinkonYARN模式是Flink的三种常见部署模式。

1.Local本地模式：

在本地模式下，Flink以单机模式运行，无需启动分布式资源管理器。这种模式适用于本地开发和测试，用于验证Flink代码的正确性和性能。

2.Standalone模式：

在Standalone模式下，Flink作为一个独立的集群运行。需要启动Flink的JobManager和TaskManager，JobManager负责接收和调度任务，而TaskManager负责执行任务。

3.Flink on YARN模式：

在FlinkonYARN模式下，Flink在YARN（Hadoop的资源调度和集群管理系统）之上运行。Flink作为一个YARN应用程序，利用YARN来管理资源分配和任务调度。使用这种模式，可以充分利用Hadoop集群的资源，实现Flink的分布式计算。

Flink的执行模式

Flink可以通过以下三种方式之一执行应用程序：

1.Session Mode：会话模式

会话模式需要先启动一个集群，保持一个会话，在这个会话中通过客户端提交作业。集群启动时所有资源就都已经确定，所有提交的作业会竞争集群中的资源。适合任务规模小，执行时间短的大量作业。

Flink的作业执行环境会一直保留在集群上，直到会话被显式终止。这样，可以提交多个作业，它们可以共享相同的集群资源和状态，从而实现更高的效率和资源利用。

2.Per-Job Mode：单作业模式

每个Flink应用程序作为一个独立的作业被提交和执行。

每次提交的Flink应用程序都会创建一个独立的作业执行环境，该作业执行环境仅用于执行该特定的作业。

当作业完成后，作业执行环境会被释放，集群关闭，资源释放

3.Application Mode：应用模式

应用模式算是前2种模式的升级，前2种模式中，Flink程序代码是在客户端执行，然后客户端提交给JobManager，客户端需要占用大量网络带宽。

应用模式需要为每一个提交的应用单独启动一个JobManager（应用程序在JobManager执行），也就是创建一个集群。这个JobManager只为执行这一个应用而存在，执行结束之后JobManager关闭。

4.三种模式的区别：

集群生命周期和资源隔离保证应用程序的main()方法是在客户端还是在集群上执行

Local本地模式

Local模式是Flink提供的最简单部署模式，可以在单台服务器上运行，适用于日常的开发和调试。

注意：Flink的运行依赖JAVA环境，需要预先安装好JDK

下载安装

Flink下载地址： https://archive.apache.org/dist/flink/

下载Flink

wget https://repo.huaweicloud.com/apache/flink/flink-1.17.0/flink-1.17.0-bin-scala_2.12.tgz

解压、重命名

tar  -zxvf flink-1.17.0-bin-scala_2.12.tgz mv flink-1.17.0 flink

启动、停止Flink

不需要进行任何配置，直接使用Flink默认配置，直接运行脚本启动

bin/start-cluster.sh

停止Flink

bin/stop-cluster.sh

直接访问：http://IP:8081，可以看到Flink的后台管理界面

每个taskmanager有3个solt

提交测试任务

提交一个测试任务：

./bin/flink run examples/batch/WordCount.jar

在控制台直接看到输出

[root@node01 flink]# ./bin/flink run examples/batch/WordCount.jar
SLF4J: Class path contains multiple SLF4J bindings.
SLF4J: Found binding in [jar:file:/usr/local/program/flink/lib/log4j-slf4j-impl-2.17.1.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/usr/local/program/hadoop/share/hadoop/common/lib/slf4j-log4j12-1.7.25.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: See http://www.slf4j.org/codes.html#multiple_bindings for an explanation.
SLF4J: Actual binding is of type [org.apache.logging.slf4j.Log4jLoggerFactory]
Executing WordCount example with default input data set.
Use --input to specify file input.
Printing result to stdout. Use --output to specify output path.
Job has been submitted with JobID a946d0abf84ac6848a823cec43f7056f
Program execution finished
Job with JobID a946d0abf84ac6848a823cec43f7056f has finished.
Job Runtime: 584 ms
Accumulator Results: 
- 1a50b4c9582d4d35a854872c62391768 (java.util.ArrayList) [170 elements](a,5)
(action,1)
(after,1)
(against,1)
(all,2)
(and,12)
(arms,1)
(arrows,1)
(awry,1)

同样，在Flink的后台管理界面 Completed Jobs 一栏可以看到刚才提交执行的程序：

停止作业

可以直接在 WEB 界面上点击对应作业的 Cancel Job 按钮进行取消，也可以使用命令行进行取消。

使用命令行进行取消时，需要先获取到作业的JobId

bin/flink list

获取到JobId后，使用flink cancel JobId命令取消作业

bin/flink cancel a946d0abf84ac6848a823cec43f7056f

Standalone独立模式

Standalone模式是集群模式的一种，独立模式是独立运行的，不依赖任何外部的资源管理平台，存在资源不足，出现故障不会自动扩展或重分配资源的能力，一般用在开发测试或作业非常少的场景下。

优缺点：

部署相对简单，可以支持小规模，少量的任务运行缺少系统层面对集群中Job的管理，容易遭成资源分配不均匀资源隔离相对简单，任务之间资源竞争严重

会话模式

会话模式部署需要先启动集群，集群资源固定，通过Web页面客户端提交任务，可以多个任务。

搭建一个Flink集群，参考：搭建Flink集群、集群HA高可用以及配置历史服务器

1.启动 Flink 集群：

通过bin/start-cluster.sh脚本启动集群

2.打开Flink Web UI

在浏览器中输入http://node01:8081/地址打开Flink Web UI

3.提交Flink作业

在Flink Web UI中选择要提交的 Flink 作业 jar 包，并指定作业参数和作业名称。

bin/flink run ../examples/streaming/WordCount.jar

4.查看Flink作业

提交作业之后，在 Flink Web UI 上会看到作业的运行状态，可以查看作业日志和监控指标等信息。

5.停止Flink作业

可以在Flink Web UI中停止作业，也可以使用bin/flink cancel jobID命令停止指定的作业

单作业模式

Standalone集群并不支持单作业模式部署，单作业模式需要借助一些资源管理平台。

应用模式

应用模式下不会提前创建集群，因此不能调用start-cluster.sh脚本，但是可以使用在bin目录下的standalone-job.sh来创建一个JobManager。

1.将Flink应用程序的jar包放到Flink的安装路径下的lib目录下。

[root@node01 flink]# mv /root/demo-1.0-SNAPSHOT.jar  lib

2.启动netcat

[root@node01 ~]# nc -lk 8888

3.启动JobManager

直接指定作业入口类，脚本会到lib目录扫描所有的jar包

[root@node01 flink]# bin/standalone-job.sh start --job-classname cn.ybzy.demo.WordCountDemo  
Starting standalonejob daemon on host node01.

4.启动TaskManager

[root@node01 flink]# bin/taskmanager.sh start
Starting taskexecutor daemon on host node01.

5.查看进程

[root@node01 flink]# jps
11973 Jps
11240 TaskManagerRunner
11898 StandaloneApplicationClusterEntryPoint

6.查看Web UI

一直是如下所示状态，明显异常：

查看flink/log/flink-root-standalonejob-1-node01.log日志

1.异常提示资源不够：

Caused by: java.util.concurrent.CompletionException: org.apache.flink.runtime.jobmanager.scheduler.NoResourceAvailableException: Could not acquire the minimum required resources.at java.util.concurrent.CompletableFuture.encodeThrowable(CompletableFuture.java:292) ~[?:1.8.0_371]at java.util.concurrent.CompletableFuture.completeThrowable(CompletableFuture.java:308) ~[?:1.8.0_371]at java.util.concurrent.CompletableFuture.uniApply(CompletableFuture.java:607) ~[?:1.8.0_371]at java.util.concurrent.CompletableFuture$UniApply.tryFire(CompletableFuture.java:591) ~[?:1.8.0_371]at java.util.concurrent.CompletableFuture.postComplete(CompletableFuture.java:488) ~[?:1.8.0_371]at java.util.concurrent.CompletableFuture.completeExceptionally(CompletableFuture.java:1990) ~[?:1.8.0_371]

修改配置文件，调大资源，发现无效。

# jobmanager.memory.process.size: 1600m
jobmanager.memory.process.size: 2000m#taskmanager.memory.process.size: 1728m
taskmanager.memory.process.size: 2600m

后来仔细观察日志，发现一处核心异常如下异常：

 org.apache.flink.runtime.resourcemanager.slotmanager.DeclarativeSlotManager [] - Received resource requirements from job 6f4f54c45d7bb59531f537b966776793: [ResourceRequirement{resourceProfile=ResourceProfile{UNKNOWN}, numberOfRequiredSlots=3}]

关键词numberOfRequiredSlots=3尤为重要，JobManager启动默认只有1Slot，Slot请求资源不够！

编辑conf/flink-conf.yaml文件

# taskmanager.numberOfTaskSlots: 1
# 修改Slot数量为3
taskmanager.numberOfTaskSlots: 3

停止taskmanager、standalone-job，重新启动，Web UI显示明显正常


发送测试数据

[root@node01 ~]# nc -lk 8888
abc bcd cdf

7.停止集群

[root@node01 flink]# bin/taskmanager.sh stop
Stopping taskexecutor daemon (pid: 14117) on host node01.
[root@node01 flink]# bin/standalone-job.sh stop
No standalonejob daemon (pid: 14813) is running anymore on node01.

8.总结：

在Flink中，Slot是Flink作业管理的资源基本单位，一个任务不一定会占用1个Slot。

当向Flink提交一个任务时，Flink会为该任务分配所需的Slot数量。通常取决于以下几个因素：

任务的并行度（Parallelism）：如果任务的并行度很高，即需要同时执行多个子任务，则可能需要使用多个Slot。TaskManager的资源：如果TaskManager的资源非常丰富，例如拥有多个CPU或GPU核心，则可以分配更多的Slot来运行任务。反之，则可能只能分配较少的Slot。任务的资源需求：如果任务需要大量的内存或计算资源，则可能需要分配更多的Slot来满足需求。

个人在编写Flink程序时，设置了并行度，打包上传运行，由于JobManager的默认numberOfTaskSlots配置为1，Solt数量不够，故出现上述异常。

StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();env.setParallelism(3);

YARN运行模式

客户端把Flink应用提交给Yarn的ResourceManager，Yarn的ResourceManager会向Yarn的NodeManager申请容器。在这些容器上，Flink会部署JobManager和TaskManager的实例，从而启动集群。Flink会根据运行在JobManger上的作业所需要的Slot数量动态分配TaskManager资源。

1.安装Hadoop

安装Hadoop参考：搭建Hadoop3.X完全分布式集群环境

2.配置环境变量

# Hadoop
export HADOOP_HOME=/usr/local/program/hadoop
export PATH=$HADOOP_HOME/bin:$HADOOP_HOME/sbin:$PATH# Flink
export HADOOP_CONF_DIR=${HADOOP_HOME}/etc/hadoop
export HADOOP_CLASSPATH=`hadoop classpath`

3.启动Hadoop集群，包括HDFS和YARN

[root@node01 hadoop]# sbin/start-all.sh 

4.启动netcat

nc -lk 8888

会话模式

YARN的会话模式需要首先申请一个YARN会话（YARN Session）来启动Flink集群。

启动Hadoop集群

启动Hadoop集群，包括HDFS和YARN

[root@node01 hadoop]# sbin/start-all.sh 

申请一个YARN会话

查看yarn-session.sh命令帮助

[root@node01 flink]# bin/yarn-session.sh --help
Usage:Optional-at,--applicationType <arg>     Set a custom application type for the application on YARN-D <property=value>             use value for given property-d,--detached                   If present, runs the job in detached mode-h,--help                       Help for the Yarn session CLI.-id,--applicationId <arg>       Attach to running YARN session-j,--jar <arg>                  Path to Flink jar file-jm,--jobManagerMemory <arg>    Memory for JobManager Container with optional unit (default: MB)-m,--jobmanager <arg>           Set to yarn-cluster to use YARN execution mode.-nl,--nodeLabel <arg>           Specify YARN node label for the YARN application-nm,--name <arg>                Set a custom name for the application on YARN-q,--query                      Display available YARN resources (memory, cores)-qu,--queue <arg>               Specify YARN queue.-s,--slots <arg>                Number of slots per TaskManager-t,--ship <arg>                 Ship files in the specified directory (t for transfer)-tm,--taskManagerMemory <arg>   Memory per TaskManager Container with optional unit (default: MB)-yd,--yarndetached              If present, runs the job in detached mode (deprecated; use non-YARN specific option instead)-z,--zookeeperNamespace <arg>   Namespace to create the Zookeeper sub-paths for high availability mode

主要参数：

-d：分离模式，让Flink YARN客户端后台运行，即YARN session可以后台运行-jm（--jobManagerMemory）：配置JobManager所需内存，默认单位MB-nm（--name）：配置在YARN UI界面上显示的任务名-qu（--queue）：指定YARN队列名-tm（--taskManager）：配置每个TaskManager所使用内存

执行脚本命令向YARN集群申请资源，开启一个YARN会话，启动Flink集群

[root@node01 flink]# bin/yarn-session.sh -nm flink-test
......
2023-06-12 22:03:01,088 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 22:03:01,428 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 22:03:01,457 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 22:03:01,476 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Cannot use kerberos delegation token manager, no valid kerberos credentials provided.
2023-06-12 22:03:01,480 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Submitting application master application_1686577483648_0001
2023-06-12 22:03:01,613 INFO  org.apache.hadoop.yarn.client.api.impl.YarnClientImpl        [] - Submitted application application_1686577483648_0001
2023-06-12 22:03:01,613 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Waiting for the cluster to be allocated
2023-06-12 22:03:01,615 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Deploying cluster, current state ACCEPTED
2023-06-12 22:03:06,406 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - YARN application has been deployed successfully.
2023-06-12 22:03:06,407 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node03:37824 of application 'application_1686577483648_0001'.
JobManager Web Interface: http://node03:37824

查看Yarn、Flink

访问http://node01:8088/cluster查看yarn

YARN Session启动之后会给出一个Web UI地址以及一个YARN application ID

2023-06-12 22:03:06,406 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - YARN application has been deployed successfully.
2023-06-12 22:03:06,407 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node03:37824 of application 'application_1686577483648_0001'.
JobManager Web Interface: http://node03:37824

访问给出的地址：http://node03:37824

提交作业

可以通过Web UI或者命令行两种方式提交作业

a.通过Web UI提交作业

b.通过命令行提交作业

1.将Flink程序打Jar包并上传至集群2.执行命令将任务提交到已经开启的Yarn-Session中运行

客户端可以自行确定JobManager的地址，也可以通过-m或者-jobmanager参数指定JobManager的地址。同时JobManager的地址在YARN Session的启动页面中可以找到。

[root@node01 ~]# /usr/local/program/flink/bin/flink run  -c cn.ybzy.demo.WordCountDemo  /root/demo-1.0-SNAPSHOT.jar2023-06-12 22:21:08,468 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 22:21:08,468 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 22:21:08,824 WARN  org.apache.flink.yarn.configuration.YarnLogConfigUtil        [] - The configuration directory ('/usr/local/program/flink/conf') already contains a LOG4J config file.If you want to use logback, then please delete or rename the log configuration file.
2023-06-12 22:21:08,860 INFO  org.apache.hadoop.yarn.client.RMProxy                        [] - Connecting to ResourceManager at node01/192.168.1.100:8032
2023-06-12 22:21:08,986 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - No path for the flink jar passed. Using the location of class org.apache.flink.yarn.YarnClusterDescriptor to locate the jar
2023-06-12 22:21:09,049 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node03:37824 of application 'application_1686577483648_0001'.
Job has been submitted with JobID cdf1ff7b48472b3d7bc413a1ee9700e8

查看、测试作业

通过Flink的Web UI页面查看提交任务的运行情况，Flink会根据运行在JobManger上的作业所需要的Slot数量动态分配TaskManager资源。

发送数据测试

[root@node01 program]# nc -lk 8888
abc bcd cdf

单作业模式

在YARN环境中，由于有了外部平台做资源调度，因此也可以直接向YARN提交一个单独的作业，从而启动一个Flink集群。

提交作业

执行命令提交作业

[root@node01 flink]# bin/flink run -t yarn-per-job -c cn.ybzy.demo.WordCountDemo  /root/demo-1.0-SNAPSHOT.jar
.....
2023-06-12 22:46:26,984 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 22:46:27,009 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 22:46:27,029 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Cannot use kerberos delegation token manager, no valid kerberos credentials provided.
2023-06-12 22:46:27,034 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Submitting application master application_1686577483648_0004
2023-06-12 22:46:27,061 INFO  org.apache.hadoop.yarn.client.api.impl.YarnClientImpl        [] - Submitted application application_1686577483648_0004
2023-06-12 22:46:27,061 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Waiting for the cluster to be allocated
2023-06-12 22:46:27,063 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Deploying cluster, current state ACCEPTED
2023-06-12 22:46:31,086 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - YARN application has been deployed successfully.
2023-06-12 22:46:31,087 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node02:42192 of application 'application_1686577483648_0004'.
Job has been submitted with JobID dfcb72ebf4a5f33d8e7967d6beaaf96d

注意：在使用-d参数启动时，启动过程中可能会出现如下异常：

Exception in thread "Thread-5" java.lang.IllegalStateException: Trying to access closed classloader. Please check if you store classloaders directly or indirectly in static fields. If the stacktrace suggests that the leak occurs in a third party library and cannot be fixed immediately, you can disable this check with the configuration 'classloader.check-leaked-classloader'.at org.apache.flink.util.FlinkUserCodeClassLoaders$SafetyNetWrapperClassLoader.ensureInner(FlinkUserCodeClassLoaders.java:184)at org.apache.flink.util.FlinkUserCodeClassLoaders$SafetyNetWrapperClassLoader.getResource(FlinkUserCodeClassLoaders.java:208)at org.apache.hadoop.conf.Configuration.getResource(Configuration.java:2780)at org.apache.hadoop.conf.Configuration.getStreamReader(Configuration.java:3036)at org.apache.hadoop.conf.Configuration.loadResource(Configuration.java:2995)at org.apache.hadoop.conf.Configuration.loadResources(Configuration.java:2968)at org.apache.hadoop.conf.Configuration.getProps(Configuration.java:2848)at org.apache.hadoop.conf.Configuration.get(Configuration.java:1200)at org.apache.hadoop.conf.Configuration.getTimeDuration(Configuration.java:1812)at org.apache.hadoop.conf.Configuration.getTimeDuration(Configuration.java:1789)at org.apache.hadoop.util.ShutdownHookManager.getShutdownTimeout(ShutdownHookManager.java:183)at org.apache.hadoop.util.ShutdownHookManager.shutdownExecutor(ShutdownHookManager.java:145)at org.apache.hadoop.util.ShutdownHookManager.access$300(ShutdownHookManager.java:65)at org.apache.hadoop.util.ShutdownHookManager$1.run(ShutdownHookManager.java:102)

解决方案是在flink的/conf/flink-conf.yaml配置文件中设置

classloader.check-leaked-classloader: false

查看Yarn、Flink

访问http://node01:8088/cluster查看

打开Flink Web UI页面进行监控

a.访问启动日志中的JobManager地址，如：node02:42192

b.也可以在http://node01:8088/cluster页面中跳转到Flink的Web UI界面

查看、取消作业

[root@node01 flink]# bin/flink list -t yarn-per-job -Dyarn.application.id=application_1686577483648_00042023-06-12 22:55:43,755 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 22:55:43,755 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 22:55:43,864 WARN  org.apache.flink.yarn.configuration.YarnLogConfigUtil        [] - The configuration directory ('/usr/local/program/flink/conf') already contains a LOG4J config file.If you want to use logback, then please delete or rename the log configuration file.
2023-06-12 22:55:43,927 INFO  org.apache.hadoop.yarn.client.RMProxy                        [] - Connecting to ResourceManager at node01/192.168.1.100:8032
2023-06-12 22:55:44,087 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - No path for the flink jar passed. Using the location of class org.apache.flink.yarn.YarnClusterDescriptor to locate the jar
2023-06-12 22:55:44,159 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node02:42192 of application 'application_1686577483648_0004'.
Waiting for response...
------------------ Running/Restarting Jobs -------------------
12.06.2023 22:46:30 : dfcb72ebf4a5f33d8e7967d6beaaf96d : Flink Streaming Job (RUNNING)
--------------------------------------------------------------
No scheduled jobs.

取消作业

# 如果取消作业，整个Flink集群会停掉
bin/flink cancel -t yarn-per-job -Dyarn.application.id=application_XXXX <jobId>

[root@node01 flink]# bin/flink cancel -t yarn-per-job -Dyarn.application.id=application_1686577483648_0004  dfcb72ebf4a5f33d8e7967d6beaaf96dSLF4J: Actual binding is of type [org.apache.logging.slf4j.Log4jLoggerFactory]
2023-06-12 22:57:06,430 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 22:57:06,430 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
Cancelling job dfcb72ebf4a5f33d8e7967d6beaaf96d.
2023-06-12 22:57:06,560 WARN  org.apache.flink.yarn.configuration.YarnLogConfigUtil        [] - The configuration directory ('/usr/local/program/flink/conf') already contains a LOG4J config file.If you want to use logback, then please delete or rename the log configuration file.
2023-06-12 22:57:06,638 INFO  org.apache.hadoop.yarn.client.RMProxy                        [] - Connecting to ResourceManager at node01/192.168.1.100:8032
2023-06-12 22:57:06,830 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - No path for the flink jar passed. Using the location of class org.apache.flink.yarn.YarnClusterDescriptor to locate the jar
2023-06-12 22:57:06,895 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node02:42192 of application 'application_1686577483648_0004'.
Cancelled job dfcb72ebf4a5f33d8e7967d6beaaf96d.

应用模式

应用模式同样非常简单，与单作业模式类似，直接执行flink run-application命令即可。

提交作业

执行命令提交作业

[root@node01 flink]# bin/flink run-application -t yarn-application -c cn.ybzy.demo.WordCountDemo  /root/demo-1.0-SNAPSHOT.jar2023-06-12 23:01:00,465 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 23:01:00,751 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 23:01:00,799 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 23:01:00,817 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Cannot use kerberos delegation token manager, no valid kerberos credentials provided.
2023-06-12 23:01:00,821 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Submitting application master application_1686577483648_0005
2023-06-12 23:01:00,847 INFO  org.apache.hadoop.yarn.client.api.impl.YarnClientImpl        [] - Submitted application application_1686577483648_0005
2023-06-12 23:01:00,848 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Waiting for the cluster to be allocated
2023-06-12 23:01:00,849 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Deploying cluster, current state ACCEPTED
2023-06-12 23:01:05,123 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - YARN application has been deployed successfully.
2023-06-12 23:01:05,124 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node03:40762 of application 'application_1686577483648_0005'.

查看、取消作业

查看作业

[root@node01 flink]# bin/flink list -t yarn-application -Dyarn.application.id=application_1686577483648_00052023-06-12 23:02:55,490 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 23:02:55,490 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 23:02:55,630 WARN  org.apache.flink.yarn.configuration.YarnLogConfigUtil        [] - The configuration directory ('/usr/local/program/flink/conf') already contains a LOG4J config file.If you want to use logback, then please delete or rename the log configuration file.
2023-06-12 23:02:55,689 INFO  org.apache.hadoop.yarn.client.RMProxy                        [] - Connecting to ResourceManager at node01/192.168.1.100:8032
2023-06-12 23:02:55,844 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - No path for the flink jar passed. Using the location of class org.apache.flink.yarn.YarnClusterDescriptor to locate the jar
2023-06-12 23:02:55,905 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node03:40762 of application 'application_1686577483648_0005'.
Waiting for response...
------------------ Running/Restarting Jobs -------------------
12.06.2023 23:01:05 : a66d8fa98d23210d36b5b005ff0a1c53 : Flink Streaming Job (RUNNING)
--------------------------------------------------------------
No scheduled jobs.

取消作业

[root@node01 flink]# bin/flink cancel -t yarn-application -Dyarn.application.id=application_1686577483648_0005 a66d8fa98d23210d36b5b005ff0a1c532023-06-12 23:03:49,038 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
2023-06-12 23:03:49,038 INFO  org.apache.flink.yarn.cli.FlinkYarnSessionCli                [] - Found Yarn properties file under /tmp/.yarn-properties-root.
Cancelling job a66d8fa98d23210d36b5b005ff0a1c53.
2023-06-12 23:03:49,156 WARN  org.apache.flink.yarn.configuration.YarnLogConfigUtil        [] - The configuration directory ('/usr/local/program/flink/conf') already contains a LOG4J config file.If you want to use logback, then please delete or rename the log configuration file.
2023-06-12 23:03:49,204 INFO  org.apache.hadoop.yarn.client.RMProxy                        [] - Connecting to ResourceManager at node01/192.168.1.100:8032
2023-06-12 23:03:49,364 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - No path for the flink jar passed. Using the location of class org.apache.flink.yarn.YarnClusterDescriptor to locate the jar
2023-06-12 23:03:49,427 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node03:40762 of application 'application_1686577483648_0005'.
Cancelled job a66d8fa98d23210d36b5b005ff0a1c53.

从HDFS读取提交任务

通过yarn.provided.lib.dirs配置选项指定位置，将flink的依赖上传到远程

将Flink本身的依赖和用户jar预先上传到HDFS，而不需要单独发送到集群，这就使得作业提交更加轻量了

上传flink的lib和plugins到HDFS上

[root@node01 flink]#  hadoop fs -mkdir /flink-dist
[root@node01 flink]# hadoop fs -put lib/ /flink-dist
[root@node01 flink]# hadoop fs -put plugins/ /flink-dist

上传Flink开发程序jar包到HDFS

[root@node01 flink]# hadoop fs -mkdir /flink-jar
[root@node01 flink]# hadoop fs -put /root/demo-1.0-SNAPSHOT.jar /flink-jar

提交作业

[root@node01 flink]# bin/flink run-application -t yarn-application -Dyarn.provided.lib.dirs="hdfs://node01:9000/flink-dist"  -c cn.ybzy.demo.WordCountDemo hdfs://node01:9000/flink-jar/demo-1.0-SNAPSHOT.jar2023-06-12 23:19:20,128 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Cluster specification: ClusterSpecification{masterMemoryMB=2500, taskManagerMemoryMB=2200, slotsPerTaskManager=3}
2023-06-12 23:19:20,617 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 23:19:20,721 INFO  org.apache.hadoop.hdfs.protocol.datatransfer.sasl.SaslDataTransferClient [] - SASL encryption trust check: localHostTrusted = false, remoteHostTrusted = false
2023-06-12 23:19:20,783 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Cannot use kerberos delegation token manager, no valid kerberos credentials provided.
2023-06-12 23:19:20,788 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Submitting application master application_1686577483648_0009
2023-06-12 23:19:20,816 INFO  org.apache.hadoop.yarn.client.api.impl.YarnClientImpl        [] - Submitted application application_1686577483648_0009
2023-06-12 23:19:20,816 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Waiting for the cluster to be allocated
2023-06-12 23:19:20,817 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Deploying cluster, current state ACCEPTED
2023-06-12 23:19:24,086 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - YARN application has been deployed successfully.
2023-06-12 23:19:24,086 INFO  org.apache.flink.yarn.YarnClusterDescriptor                  [] - Found Web Interface node02:43653 of application 'application_1686577483648_0009'.

Yarn模式高可用

Standalone模式中, 同时启动多个Jobmanager, 一个为leader其他为standby, 当leader挂了, 其他的才会有一个成为leader

yarn的高可用是同时只启动一个Jobmanager, 当这个Jobmanager挂了之后, yarn会再次启动一个, 其实是利用的yarn的重试次数来实现的高可用

在yarn-site.xml中配置

<property><name>yarn.resourcemanager.am.max-attempts</name><value>4</value><description>The maximum number of application master execution attempts.  </description>
</property>

在flink-conf.yaml中配置

# 次数应该小于yarn-site.xml中配置重试次数
yarn.application-attempts: 3
high-availability.type: zookeeper
high-availability.storageDir: hdfs://node01:9000/flink/yarn/ha
high-availability.zookeeper.quorum: node01:2181,node02:2181,node03:2181
high-availability.zookeeper.path.root: /flink-yarn

启动yarn-session

[root@node01 flink]# bin/yarn-session.sh -nm flink-test

kill一个Jobmanager，查看复活情况

jpskill -9 pid