flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aljoscha Krettek <aljos...@apache.org>
Subject Re: Issue With Configuration File
Date Tue, 06 Jun 2017 17:41:16 GMT
Hi Will,

I’m afraid that will simply run your program in Flink local mode, where the configuration
settings are ignored because an in-process Flink Cluster is being started.

For running a Beam pipeline on YARN you have two options right now:

1. Start a Flink YARN session as described here https://ci.apache.org/projects/flink/flink-docs-release-1.3/setup/yarn_setup.html#flink-yarn-session
<https://ci.apache.org/projects/flink/flink-docs-release-1.3/setup/yarn_setup.html#flink-yarn-session>
and submit your Beam job as described here https://beam.apache.org/get-started/quickstart-java/#run-wordcount
<https://beam.apache.org/get-started/quickstart-java/#run-wordcount> once the YARN session
is up and has printed the JobManager details that you need to submit the Beam job.

2. Package your program as a fat jar. The quickstart from the Beam website should have the
rights settings in the POM for doing that. Then use regular Flink submission to submit that
jar to a YARN cluster as described here https://ci.apache.org/projects/flink/flink-docs-release-1.3/setup/yarn_setup.html#run-a-single-flink-job-on-yarn
<https://ci.apache.org/projects/flink/flink-docs-release-1.3/setup/yarn_setup.html#run-a-single-flink-job-on-yarn>

Best,
Aljoscha

> On 6. Jun 2017, at 19:05, Will Walters <wwalters@yahoo-inc.com> wrote:
> 
> Aljoscha,
> 
> You're correct that I'm using Beam. Here's the shell script I'm using to start the job:
> 
> hadoop jar **path to jar file** org.apache.beam.examples.complete.TfIdf \
> --runner=FlinkRunner \
> --input= **path to directory** \
> --output=tfIdf
> 
> There are two lines before that to set the Flink config path, as well as the Hadoop classpath.
> 
> Thank you,
> Will.
> 
> 
> On Tuesday, June 6, 2017 5:45 AM, Aljoscha Krettek <aljoscha@apache.org> wrote:
> 
> 
> Hi Will,
> 
> How are you starting your cluster/executing your program? From what I can gather you
are using Beam, is that right? The line about the FlinkMiniCluster seems strange because this
would hint at the fact that the Runner is trying to execute in local mode.
> 
> Best,
> Aljoscha
> 
>> On 6. Jun 2017, at 02:13, Will Walters <wwalters@yahoo-inc.com <mailto:wwalters@yahoo-inc.com>>
wrote:
>> 
>> Hello,
>> 
>> I'm having issues editing the default Flink memory settings. I'm attempting to run
a Flink task on a cluster at scale. The log shows my edited config settings having been read
into the program, but they're having no effect. Here's the trace:
>> 
>> 17/06/05 23:45:41 INFO flink.FlinkRunner: Starting execution of Flink program.
>> 17/06/05 23:45:41 INFO java.ExecutionEnvironment: The job has 0 registered types
and 0 default Kryo serializers
>> 17/06/05 23:45:41 INFO configuration.GlobalConfiguration: Loading configuration property:
taskmanager.memory.size, 100
>> 17/06/05 23:45:41 INFO configuration.GlobalConfiguration: Loading configuration property:
taskmanager.heap.mb, 256
>> 17/06/05 23:45:41 INFO configuration.GlobalConfiguration: Loading configuration property:
taskmanager.debug.memory.startLogThread, true
>> 17/06/05 23:45:41 INFO minicluster.FlinkMiniCluster: Disabled queryable state server
>> 17/06/05 23:45:41 INFO minicluster.FlinkMiniCluster: Starting FlinkMiniCluster.
>> .
>> .
>> .
>> 17/06/05 23:45:41 INFO network.NetworkEnvironment: Starting the network environment
and its components.
>> 17/06/05 23:45:41 INFO taskmanager.TaskManager: Limiting managed memory to 17592186044406
MB, memory will be allocated lazily.
>> 17/06/05 23:45:41 ERROR flink.FlinkRunner: Pipeline execution failed
>> java.lang.IllegalArgumentException: Size of total memory must be positive.
>> 
>> And here's the config file I'm using:
>> 
>> taskmanager.memory.size: 100
>> taskmanager.heap.mb: 256
>> taskmanager.debug.memory.startLogThread: true
>> 
>> In the shell script I'm using to run the task, I've edited the FLINK_CONF_DIR to
direct to the config file that I created.
>> 
>> If anyone has any advice or inputs it would be much appreciated.
>> 
>> Thanks!
>> Will Walters.
> 
> 
> 


Mime
View raw message