Command-Line Interface

Flink provides a Command-Line Interface (CLI) to run programs that are packagedas JAR files, and control their execution. The CLI is partof any Flink setup, available in local single node setups and indistributed setups. It is located under <flink-home>/bin/flinkand connects by default to the running Flink master (JobManager) that wasstarted from the same installation directory.

The command line can be used to

  • submit jobs for execution,
  • cancel a running job,
  • provide information about a job,
  • list running and waiting jobs,
  • trigger and dispose savepoints, and

A prerequisite to using the command line interface is that the Flinkmaster (JobManager) has been started (via<flink-home>/bin/start-cluster.sh) or that another deployment target such as YARN or Kubernetes isavailable.

Deployment targets

Flink has the concept of executors for defining available deployment targets. You can see theavailable executors in the output of bin/flink —help, for example:

  1. Options for executor mode:
  2.    -D <property=value>   Generic configuration options for
  3.                          execution/deployment and for the configured executor.
  4.                          The available options can be found at
  5.                          https://ci.apache.org/projects/flink/flink-docs-stabl
  6.                          e/ops/config.html
  7.    -e,--executor <arg>   The name of the executor to be used for executing the
  8.                          given job, which is equivalent to the
  9.                          "execution.target" config option. The currently
  10.                          available executors are: "remote", "local",
  11.                          "kubernetes-session", "yarn-per-job", "yarn-session".

When running one of the bin/flink actions, the executor is specified using the —executoroption.

Examples

Job Submission Examples

These examples about how to submit a job in CLI.

  • Run example program with no arguments:
  1. ./bin/flink run ./examples/batch/WordCount.jar
  • Run example program with arguments for input and result files:
  1. ./bin/flink run ./examples/batch/WordCount.jar \
  2.                      --input file:///home/user/hamlet.txt --output file:///home/user/wordcount_out
  • Run example program with parallelism 16 and arguments for input and result files:
  1. ./bin/flink run -p 16 ./examples/batch/WordCount.jar \
  2.                      --input file:///home/user/hamlet.txt --output file:///home/user/wordcount_out
  • Run example program with flink log output disabled:
  1.     ./bin/flink run -q ./examples/batch/WordCount.jar
  • Run example program in detached mode:
  1.     ./bin/flink run -d ./examples/batch/WordCount.jar
  • Run example program on a specific JobManager:
  1. ./bin/flink run -m myJMHost:8081 \
  2.                        ./examples/batch/WordCount.jar \
  3.                        --input file:///home/user/hamlet.txt --output file:///home/user/wordcount_out
  • Run example program with a specific class as an entry point:
  1. ./bin/flink run -c org.apache.flink.examples.java.wordcount.WordCount \
  2.                        ./examples/batch/WordCount.jar \
  3.                        --input file:///home/user/hamlet.txt --output file:///home/user/wordcount_out
  1. ./bin/flink run -m yarn-cluster \
  2.                        ./examples/batch/WordCount.jar \
  3.                        --input hdfs:///user/hamlet.txt --output hdfs:///user/wordcount_out

Note When submitting Python job via flink run, Flink will run the command “python”. Please run the following command to confirm that the command “python” in current environment points to Python 3.5+:

  1. $ python --version
  2. # the version printed here must be 3.5+
  • Run Python Table program:
  1. ./bin/flink run -py examples/python/table/batch/word_count.py
  • Run Python Table program with pyFiles:
  1. ./bin/flink run -py examples/python/table/batch/word_count.py \
  2.                         -pyfs file:///user.txt,hdfs:///$namenode_address/username.txt
  • Run Python Table program with a JAR file:
  1. ./bin/flink run -py examples/python/table/batch/word_count.py -j <jarFile>
  • Run Python Table program with pyFiles and pyModule:
  1. ./bin/flink run -pym batch.word_count -pyfs examples/python/table/batch
  • Run Python Table program with parallelism 16:
  1. ./bin/flink run -p 16 -py examples/python/table/batch/word_count.py
  • Run Python Table program with flink log output disabled:
  1. ./bin/flink run -q -py examples/python/table/batch/word_count.py
  • Run Python Table program in detached mode:
  1. ./bin/flink run -d -py examples/python/table/batch/word_count.py
  • Run Python Table program on a specific JobManager:
  1. ./bin/flink run -m myJMHost:8081 \
  2.                        -py examples/python/table/batch/word_count.py
  1. ./bin/flink run -m yarn-cluster \
  2.                        -py examples/python/table/batch/word_count.py

Job Management Examples

These examples about how to manage a job in CLI.

  • Display the optimized execution plan for the WordCount example program as JSON:
  1. ./bin/flink info ./examples/batch/WordCount.jar \
  2.                         --input file:///home/user/hamlet.txt --output file:///home/user/wordcount_out
  • List scheduled and running jobs (including their JobIDs):
  1. ./bin/flink list
  • List scheduled jobs (including their JobIDs):
  1. ./bin/flink list -s
  • List running jobs (including their JobIDs):
  1. ./bin/flink list -r
  • List all existing jobs (including their JobIDs):
  1. ./bin/flink list -a
  • List running Flink jobs inside Flink YARN session:
  1. ./bin/flink list -m yarn-cluster -yid <yarnApplicationID> -r
  • Cancel a job:
  1. ./bin/flink cancel <jobID>
  • Cancel a job with a savepoint (deprecated; use “stop” instead):
  1. ./bin/flink cancel -s [targetDirectory] <jobID>
  • Gracefully stop a job with a savepoint (streaming jobs only):
  1. ./bin/flink stop [-p targetDirectory] [-d] <jobID>

Savepoints

Savepoints are controlled via the command line client:

Trigger a Savepoint

  1. ./bin/flink savepoint <jobId> [savepointDirectory]

This will trigger a savepoint for the job with ID jobId, and returns the path of the created savepoint. You need this path to restore and dispose savepoints.

Furthermore, you can optionally specify a target file system directory to store the savepoint in. The directory needs to be accessible by the JobManager.

If you don’t specify a target directory, you need to have configured a default directory. Otherwise, triggering the savepoint will fail.

Trigger a Savepoint with YARN

  1. ./bin/flink savepoint <jobId> [savepointDirectory] -yid <yarnAppId>

This will trigger a savepoint for the job with ID jobId and YARN application ID yarnAppId, and returns the path of the created savepoint.

Everything else is the same as described in the above Trigger a Savepoint section.

Stop

Use the stop to gracefully stop a running streaming job with a savepoint.

  1. ./bin/flink stop [-p targetDirectory] [-d] <jobID>

A “stop” call is a more graceful way of stopping a running streaming job, as the “stop” signal flows fromsource to sink. When the user requests to stop a job, all sources will be requested to send the last checkpoint barrierthat will trigger a savepoint, and after the successful completion of that savepoint, they will finish by calling theircancel() method. If the -d flag is specified, then a MAX_WATERMARK will be emitted before the last checkpointbarrier. This will result all registered event-time timers to fire, thus flushing out any state that is waiting fora specific watermark, e.g. windows. The job will keep running until all sources properly shut down. This allows the job to finish processing all in-flight data.

Cancel with a savepoint (deprecated)

You can atomically trigger a savepoint and cancel a job.

  1. ./bin/flink cancel -s [savepointDirectory] <jobID>

If no savepoint directory is configured, you need to configure a default savepoint directory for the Flink installation (see Savepoints).

The job will only be cancelled if the savepoint succeeds.

Note: Cancelling a job with savepoint is deprecated. Use "stop" instead.

Restore a savepoint

  1. ./bin/flink run -s <savepointPath> ...

The run command has a savepoint flag to submit a job, which restores its state from a savepoint. The savepoint path is returned by the savepoint trigger command.

By default, we try to match all savepoint state to the job being submitted. If you want to allow to skip savepoint state that cannot be restored with the new job you can set the allowNonRestoredState flag. You need to allow this if you removed an operator from your program that was part of the program when the savepoint was triggered and you still want to use the savepoint.

  1. ./bin/flink run -s <savepointPath> -n ...

This is useful if your program dropped an operator that was part of the savepoint.

Dispose a savepoint

  1. ./bin/flink savepoint -d <savepointPath>

Disposes the savepoint at the given path. The savepoint path is returned by the savepoint trigger command.

If you use custom state instances (for example custom reducing state or RocksDB state), you have to specify the path to the program JAR with which the savepoint was triggered in order to dispose the savepoint with the user code class loader:

  1. ./bin/flink savepoint -d <savepointPath> -j <jarFile>

Otherwise, you will run into a ClassNotFoundException.

Usage

The command line syntax is as follows:

  1. ./flink <ACTION> [OPTIONS] [ARGUMENTS]
  3. The following actions are available:
  5. Action "run" compiles and runs a program.
  7.   Syntax: run [OPTIONS] <jar-file> <arguments>
  8.   "run" action options:
  9.      -c,--class <classname>               Class with the program entry point
  10.                                           ("main()" method). Only needed if the
  11.                                           JAR file does not specify the class in
  12.                                           its manifest.
  13.      -C,--classpath <url>                 Adds a URL to each user code
  14.                                           classloader  on all nodes in the
  15.                                           cluster. The paths must specify a
  16.                                           protocol (e.g. file://) and be
  17.                                           accessible on all nodes (e.g. by means
  18.                                           of a NFS share). You can use this
  19.                                           option multiple times for specifying
  20.                                           more than one URL. The protocol must
  21.                                           be supported by the {@link
  22.                                           java.net.URLClassLoader}.
  23.      -d,--detached                        If present, runs the job in detached
  24.                                           mode
  25.      -n,--allowNonRestoredState           Allow to skip savepoint state that
  26.                                           cannot be restored. You need to allow
  27.                                           this if you removed an operator from
  28.                                           your program that was part of the
  29.                                           program when the savepoint was
  30.                                           triggered.
  31.      -p,--parallelism <parallelism>       The parallelism with which to run the
  32.                                           program. Optional flag to override the
  33.                                           default value specified in the
  34.                                           configuration.
  35.      -py,--python <pythonFile>            Python script with the program entry
  36.                                           point. The dependent resources can be
  37.                                           configured with the `--pyFiles`
  38.                                           option.
  39.      -pyarch,--pyArchives <arg>           Add python archive files for job. The
  40.                                           archive files will be extracted to the
  41.                                           working directory of python UDF
  42.                                           worker. Currently only zip-format is
  43.                                           supported. For each archive file, a
  44.                                           target directory be specified. If the
  45.                                           target directory name is specified,
  46.                                           the archive file will be extracted to
  47.                                           a name can directory with the
  48.                                           specified name. Otherwise, the archive
  49.                                           file will be extracted to a directory
  50.                                           with the same name of the archive
  51.                                           file. The files uploaded via this
  52.                                           option are accessible via relative
  53.                                           path. '#' could be used as the
  54.                                           separator of the archive file path and
  55.                                           the target directory name. Comma (',')
  56.                                           could be used as the separator to
  57.                                           specify multiple archive files. This
  58.                                           option can be used to upload the
  59.                                           virtual environment, the data files
  60.                                           used in Python UDF (e.g.: --pyArchives
  61.                                           file:///tmp/py37.zip,file:///tmp/data.
  62.                                           zip#data --pyExecutable
  63.                                           py37.zip/py37/bin/python). The data
  64.                                           files could be accessed in Python UDF,
  65.                                           e.g.: f = open('data/data.txt', 'r').
  66.      -pyexec,--pyExecutable <arg>         Specify the path of the python
  67.                                           interpreter used to execute the python
  68.                                           UDF worker (e.g.: --pyExecutable
  69.                                           /usr/local/bin/python3). The python
  70.                                           UDF worker depends on Python 3.5+,
  71.                                           Apache Beam (version == 2.15.0), Pip
  72.                                           (version >= 7.1.0) and SetupTools
  73.                                           (version >= 37.0.0). Please ensure
  74.                                           that the specified environment meets
  75.                                           the above requirements.
  76.      -pyfs,--pyFiles <pythonFiles>        Attach custom python files for job.
  77.                                           These files will be added to the
  78.                                           PYTHONPATH of both the local client
  79.                                           and the remote python UDF worker. The
  80.                                           standard python resource file suffixes
  81.                                           such as .py/.egg/.zip or directory are
  82.                                           all supported. Comma (',') could be
  83.                                           used as the separator to specify
  84.                                           multiple files (e.g.: --pyFiles
  85.                                           file:///tmp/myresource.zip,hdfs:///$na
  86.                                           menode_address/myresource2.zip).
  87.      -pym,--pyModule <pythonModule>       Python module with the program entry
  88.                                           point. This option must be used in
  89.                                           conjunction with `--pyFiles`.
  90.      -pyreq,--pyRequirements <arg>        Specify a requirements.txt file which
  91.                                           defines the third-party dependencies.
  92.                                           These dependencies will be installed
  93.                                           and added to the PYTHONPATH of the
  94.                                           python UDF worker. A directory which
  95.                                           contains the installation packages of
  96.                                           these dependencies could be specified
  97.                                           optionally. Use '#' as the separator
  98.                                           if the optional parameter exists
  99.                                           (e.g.: --pyRequirements
  100.                                           file:///tmp/requirements.txt#file:///t
  101.                                           mp/cached_dir).
  102.      -s,--fromSavepoint <savepointPath>   Path to a savepoint to restore the job
  103.                                           from (for example
  104.                                           hdfs:///flink/savepoint-1537).
  105.      -sae,--shutdownOnAttachedExit        If the job is submitted in attached
  106.                                           mode, perform a best-effort cluster
  107.                                           shutdown when the CLI is terminated
  108.                                           abruptly, e.g., in response to a user
  109.                                           interrupt, such as typing Ctrl + C.
  110.   Options for yarn-cluster mode:
  111.      -d,--detached                        If present, runs the job in detached
  112.                                           mode
  113.      -m,--jobmanager <arg>                Address of the JobManager (master) to
  114.                                           which to connect. Use this flag to
  115.                                           connect to a different JobManager than
  116.                                           the one specified in the
  117.                                           configuration.
  118.      -yat,--yarnapplicationType <arg>     Set a custom application type for the
  119.                                           application on YARN
  120.      -yD <property=value>                 use value for given property
  121.      -yd,--yarndetached                   If present, runs the job in detached
  122.                                           mode (deprecated; use non-YARN
  123.                                           specific option instead)
  124.      -yh,--yarnhelp                       Help for the Yarn session CLI.
  125.      -yid,--yarnapplicationId <arg>       Attach to running YARN session
  126.      -yj,--yarnjar <arg>                  Path to Flink jar file
  127.      -yjm,--yarnjobManagerMemory <arg>    Memory for JobManager Container with
  128.                                           optional unit (default: MB)
  129.      -ynl,--yarnnodeLabel <arg>           Specify YARN node label for the YARN
  130.                                           application
  131.      -ynm,--yarnname <arg>                Set a custom name for the application
  132.                                           on YARN
  133.      -yq,--yarnquery                      Display available YARN resources
  134.                                           (memory, cores)
  135.      -yqu,--yarnqueue <arg>               Specify YARN queue.
  136.      -ys,--yarnslots <arg>                Number of slots per TaskManager
  137.      -yt,--yarnship <arg>                 Ship files in the specified directory
  138.                                           (t for transfer)
  139.      -ytm,--yarntaskManagerMemory <arg>   Memory per TaskManager Container with
  140.                                           optional unit (default: MB)
  141.      -yz,--yarnzookeeperNamespace <arg>   Namespace to create the Zookeeper
  142.                                           sub-paths for high availability mode
  143.      -z,--zookeeperNamespace <arg>        Namespace to create the Zookeeper
  144.                                           sub-paths for high availability mode
  146.   Options for executor mode:
  147.      -D <property=value>   Generic configuration options for
  148.                            execution/deployment and for the configured executor.
  149.                            The available options can be found at
  150.                            https://ci.apache.org/projects/flink/flink-docs-stabl
  151.                            e/ops/config.html
  152.      -e,--executor <arg>   The name of the executor to be used for executing the
  153.                            given job, which is equivalent to the
  154.                            "execution.target" config option. The currently
  155.                            available executors are: "remote", "local",
  156.                            "kubernetes-session", "yarn-per-job", "yarn-session".
  158.   Options for default mode:
  159.      -m,--jobmanager <arg>           Address of the JobManager (master) to which
  160.                                      to connect. Use this flag to connect to a
  161.                                      different JobManager than the one specified
  162.                                      in the configuration.
  163.      -z,--zookeeperNamespace <arg>   Namespace to create the Zookeeper sub-paths
  164.                                      for high availability mode
  168. Action "info" shows the optimized execution plan of the program (JSON).
  170.   Syntax: info [OPTIONS] <jar-file> <arguments>
  171.   "info" action options:
  172.      -c,--class <classname>           Class with the program entry point
  173.                                       ("main()" method). Only needed if the JAR
  174.                                       file does not specify the class in its
  175.                                       manifest.
  176.      -p,--parallelism <parallelism>   The parallelism with which to run the
  177.                                       program. Optional flag to override the
  178.                                       default value specified in the
  179.                                       configuration.
  182. Action "list" lists running and scheduled programs.
  184.   Syntax: list [OPTIONS]
  185.   "list" action options:
  186.      -a,--all         Show all programs and their JobIDs
  187.      -r,--running     Show only running programs and their JobIDs
  188.      -s,--scheduled   Show only scheduled programs and their JobIDs
  189.   Options for yarn-cluster mode:
  190.      -m,--jobmanager <arg>            Address of the JobManager (master) to
  191.                                       which to connect. Use this flag to connect
  192.                                       to a different JobManager than the one
  193.                                       specified in the configuration.
  194.      -yid,--yarnapplicationId <arg>   Attach to running YARN session
  195.      -z,--zookeeperNamespace <arg>    Namespace to create the Zookeeper
  196.                                       sub-paths for high availability mode
  198.   Options for executor mode:
  199.      -D <property=value>   Generic configuration options for
  200.                            execution/deployment and for the configured executor.
  201.                            The available options can be found at
  202.                            https://ci.apache.org/projects/flink/flink-docs-stabl
  203.                            e/ops/config.html
  204.      -e,--executor <arg>   The name of the executor to be used for executing the
  205.                            given job, which is equivalent to the
  206.                            "execution.target" config option. The currently
  207.                            available executors are: "remote", "local",
  208.                            "kubernetes-session", "yarn-per-job", "yarn-session".
  210.   Options for default mode:
  211.      -m,--jobmanager <arg>           Address of the JobManager (master) to which
  212.                                      to connect. Use this flag to connect to a
  213.                                      different JobManager than the one specified
  214.                                      in the configuration.
  215.      -z,--zookeeperNamespace <arg>   Namespace to create the Zookeeper sub-paths
  216.                                      for high availability mode
  220. Action "stop" stops a running program with a savepoint (streaming jobs only).
  222.   Syntax: stop [OPTIONS] <Job ID>
  223.   "stop" action options:
  224.      -d,--drain                           Send MAX_WATERMARK before taking the
  225.                                           savepoint and stopping the pipelne.
  226.      -p,--savepointPath <savepointPath>   Path to the savepoint (for example
  227.                                           hdfs:///flink/savepoint-1537). If no
  228.                                           directory is specified, the configured
  229.                                           default will be used
  230.                                           ("state.savepoints.dir").
  231.   Options for yarn-cluster mode:
  232.      -m,--jobmanager <arg>            Address of the JobManager (master) to
  233.                                       which to connect. Use this flag to connect
  234.                                       to a different JobManager than the one
  235.                                       specified in the configuration.
  236.      -yid,--yarnapplicationId <arg>   Attach to running YARN session
  237.      -z,--zookeeperNamespace <arg>    Namespace to create the Zookeeper
  238.                                       sub-paths for high availability mode
  240.   Options for executor mode:
  241.      -D <property=value>   Generic configuration options for
  242.                            execution/deployment and for the configured executor.
  243.                            The available options can be found at
  244.                            https://ci.apache.org/projects/flink/flink-docs-stabl
  245.                            e/ops/config.html
  246.      -e,--executor <arg>   The name of the executor to be used for executing the
  247.                            given job, which is equivalent to the
  248.                            "execution.target" config option. The currently
  249.                            available executors are: "remote", "local",
  250.                            "kubernetes-session", "yarn-per-job", "yarn-session".
  252.   Options for default mode:
  253.      -m,--jobmanager <arg>           Address of the JobManager (master) to which
  254.                                      to connect. Use this flag to connect to a
  255.                                      different JobManager than the one specified
  256.                                      in the configuration.
  257.      -z,--zookeeperNamespace <arg>   Namespace to create the Zookeeper sub-paths
  258.                                      for high availability mode
  262. Action "cancel" cancels a running program.
  264.   Syntax: cancel [OPTIONS] <Job ID>
  265.   "cancel" action options:
  266.      -s,--withSavepoint <targetDirectory>   **DEPRECATION WARNING**: Cancelling
  267.                                             a job with savepoint is deprecated.
  268.                                             Use "stop" instead.
  269.                                             Trigger savepoint and cancel job.
  270.                                             The target directory is optional. If
  271.                                             no directory is specified, the
  272.                                             configured default directory
  273.                                             (state.savepoints.dir) is used.
  274.   Options for yarn-cluster mode:
  275.      -m,--jobmanager <arg>            Address of the JobManager (master) to
  276.                                       which to connect. Use this flag to connect
  277.                                       to a different JobManager than the one
  278.                                       specified in the configuration.
  279.      -yid,--yarnapplicationId <arg>   Attach to running YARN session
  280.      -z,--zookeeperNamespace <arg>    Namespace to create the Zookeeper
  281.                                       sub-paths for high availability mode
  283.   Options for executor mode:
  284.      -D <property=value>   Generic configuration options for
  285.                            execution/deployment and for the configured executor.
  286.                            The available options can be found at
  287.                            https://ci.apache.org/projects/flink/flink-docs-stabl
  288.                            e/ops/config.html
  289.      -e,--executor <arg>   The name of the executor to be used for executing the
  290.                            given job, which is equivalent to the
  291.                            "execution.target" config option. The currently
  292.                            available executors are: "remote", "local",
  293.                            "kubernetes-session", "yarn-per-job", "yarn-session".
  295.   Options for default mode:
  296.      -m,--jobmanager <arg>           Address of the JobManager (master) to which
  297.                                      to connect. Use this flag to connect to a
  298.                                      different JobManager than the one specified
  299.                                      in the configuration.
  300.      -z,--zookeeperNamespace <arg>   Namespace to create the Zookeeper sub-paths
  301.                                      for high availability mode
  305. Action "savepoint" triggers savepoints for a running job or disposes existing ones.
  307.   Syntax: savepoint [OPTIONS] <Job ID> [<target directory>]
  308.   "savepoint" action options:
  309.      -d,--dispose <arg>       Path of savepoint to dispose.
  310.      -j,--jarfile <jarfile>   Flink program JAR file.
  311.   Options for yarn-cluster mode:
  312.      -m,--jobmanager <arg>            Address of the JobManager (master) to
  313.                                       which to connect. Use this flag to connect
  314.                                       to a different JobManager than the one
  315.                                       specified in the configuration.
  316.      -yid,--yarnapplicationId <arg>   Attach to running YARN session
  317.      -z,--zookeeperNamespace <arg>    Namespace to create the Zookeeper
  318.                                       sub-paths for high availability mode
  320.   Options for executor mode:
  321.      -D <property=value>   Generic configuration options for
  322.                            execution/deployment and for the configured executor.
  323.                            The available options can be found at
  324.                            https://ci.apache.org/projects/flink/flink-docs-stabl
  325.                            e/ops/config.html
  326.      -e,--executor <arg>   The name of the executor to be used for executing the
  327.                            given job, which is equivalent to the
  328.                            "execution.target" config option. The currently
  329.                            available executors are: "remote", "local",
  330.                            "kubernetes-session", "yarn-per-job", "yarn-session".
  332.   Options for default mode:
  333.      -m,--jobmanager <arg>           Address of the JobManager (master) to which
  334.                                      to connect. Use this flag to connect to a
  335.                                      different JobManager than the one specified
  336.                                      in the configuration.
  337.      -z,--zookeeperNamespace <arg>   Namespace to create the Zookeeper sub-paths
  338.                                      for high availability mode