flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chesnay Schepler <ches...@apache.org>
Subject Re: CPU Cores of JobManager
Date Wed, 06 Dec 2017 09:36:31 GMT
This can only have an effect if "*taskmanager.net.detailed-metrics*" is 
enabled (by default it is disabled).
Otherwise an increase in sub-channels will neither increase the number 
of metrics nor the number of polls by the UI.

On 06.12.2017 10:22, Stephan Ewen wrote:
> A 'rebalance()' should no impose load on the JM itself. There are more 
> sub-channels in that case, but the JobManager is not coordinating 
> anything for them unless there is a failure/recovery.
>
> Each subchannel generates additional metrics, though - might that be 
> an issue?
>
> On Wed, Dec 6, 2017 at 9:50 AM, Yuta Morisawa 
> <yu-morisawa@kddi-research.jp <mailto:yu-morisawa@kddi-research.jp>> 
> wrote:
>
>     > Which version of Fllink are you using?
>     I'm using Flink 1.3.0.
>
>     > Is the high CPU usage caused by the browser or the JVM?
>     The CPU usage of browser is not high, the JVM is not also.
>
>     I found that if I remove the rebalance function in my code, the
>     WEBUI works well.
>     I don't know why but I think rebalance function imposes a load on
>     JobManager.
>
>
>     Regards,
>     Yuta
>
>
>     On 2017/12/06 17:08, Chesnay Schepler wrote:
>
>         Which version of Fllink are you using?
>
>         Is the high CPU usage caused by the browser or the JVM?
>
>         One thing we should investigate is how many requests the UI
>         sends to the JobManager,
>         but I don't know of a good way to do that without modifying
>         the source.
>
>         On 06.12.2017 01 <tel:06.12.2017%2001>:20, Yuta Morisawa wrote:
>
>             Hi Timo
>
>             I execute streaming job without checkpointing and I don't
>             configure any state backend, so it may be
>             "MemoryStateBackend".
>
>             Actually, my streaming app just reads data from kafka and
>             writes it to an external DB. Its not so complicated.
>
>             Regards,
>             Yuta
>
>             On 2017/12/05 19:55, Timo Walther wrote:
>
>                 I had some profiling tool like jvisualvm in mind. Are
>                 you executing streaming or batch jobs? If streaming,
>                 is checkpointing enabled and which type of statebackend?
>
>                 @Chesnay do you have experience with slow behavior of
>                 the Web UI?
>
>                 Regards,
>                 Timo
>
>
>                 Am 12/5/17 um 10:37 AM schrieb Yuta Morisawa:
>
>                     Hi Timo
>
>                     Thank you for your early reply.
>
>                     These are commands which I run my apps.
>                     ./bin/yarn-session.sh -n 20 -jm 6000 -tm 24000 -s 10
>                     ./bin/flink run -p 100 <jar1> <jar1 parameters>
>                     ./bin/flink run -p 100 <jar2> <jar2 parameters>
>
>                     So, JobManager Heap Memory = 6000 MB and it
>                     manages 2 jobs.
>
>                     > Maybe you can use a profiler and find out which
>                     component consumes so
>                     > much CPU resources?
>                     You mean Java Flight Recorder or JITWatch?
>                     Or, Flink has original profiler?
>                     https://ci.apache.org/projects/flink/flink-docs-release-1.3/monitoring/application_profiling.html
>                     <https://ci.apache.org/projects/flink/flink-docs-release-1.3/monitoring/application_profiling.html>
>
>
>                     Regards,
>                     Yuta
>
>                     On 2017/12/05 18:02, Timo Walther wrote:
>
>                         Hi Yuta,
>
>                         as far as I know you cannot assign more cores
>                         to a JobManager.
>
>                         Can you tell us a bit more about your
>                         environment? How many jobs does the JobManager
>                         has to manage? How much heap memory is
>                         assigned to the JobManager?
>
>                         Maybe you can use a profiler and find out
>                         which component consumes so much CPU resources?
>
>                         Regards,
>                         Timo
>
>
>                         Am 12/5/17 um 5:13 AM schrieb Yuta Morisawa:
>
>                             Hi
>
>                             Now I am looking for the way to increase
>                             the number of allocated CPU cores because
>                             my JobManagaer WEBUI is very heavy and
>                             sometimes freeze.
>
>                             I think this is caused by the resource
>                             shortage of JobManager.
>                             How can I increase the number of CPU for
>                             JobManager in YARN mode?
>
>
>
>                             Thanks
>                             Yuta
>
>
>
>
>
>
>
>
>
>
>
>


Mime
View raw message