Monitoring single-run job statistics

[Filip Łęczycki]

Hi all,

I am running filnk aps on YARN cluster and I am trying to get some benchmarks. When I start a long-running flink cluster on my YARN cluster I have an access to web UI and rest API that provide me statistics of the deployed jobs (as desribed here: https://ci.apache.org/projects/flink/flink-docs-master/internals/monitoring_rest_api.html). I was wondering is this possible to get such information about a single run job trigerred with 'flink run -m yarn-cluster ...'? After the job is finished there is no flink client running so I cannot use rest api to get stats. 

Thanks for any help:)

[Till Rohrmann-2]

Hi Filip,

at the moment it is not possible to retrieve the job statistics after the job has finished with flink run -m yarn-cluster. The reason is that the YARN cluster is only alive as long as the job is executed. Thus, I would recommend you to execute your jobs with a long running Flink cluster on YARN.

Cheers,
Till

​

On Fri, Jan 1, 2016 at 11:29 PM, Filip Łęczycki <[hidden email]> wrote:

Hi all,

I am running filnk aps on YARN cluster and I am trying to get some benchmarks. When I start a long-running flink cluster on my YARN cluster I have an access to web UI and rest API that provide me statistics of the deployed jobs (as desribed here: https://ci.apache.org/projects/flink/flink-docs-master/internals/monitoring_rest_api.html). I was wondering is this possible to get such information about a single run job trigerred with 'flink run -m yarn-cluster ...'? After the job is finished there is no flink client running so I cannot use rest api to get stats. 

Thanks for any help:)

Best regards/Pozdrawiam,
Filip Łęczycki

[Filip Łęczycki]

Hi Till,

Thank you for you answer however I am sorry to hear that. I was reluctant to execute jobs with long running Flink cluster due to the fact that multiple jobs would cloud yarn statistics regarding cpu and memory time as well as Flink's garbage collector statistics in log, as they would be stored for the whole Flink cluster, instead of a single job. 

Do you know whether is there a way to extract mentioned stats (cpu time, mem time, gc time) for a single job ran on long running Flink cluster?

I will be very grateful for an answer:)

Best regards,

Filip

2016-01-04 10:05 GMT+01:00 Till Rohrmann <[hidden email]>:

Hi Filip,

at the moment it is not possible to retrieve the job statistics after the job has finished with flink run -m yarn-cluster. The reason is that the YARN cluster is only alive as long as the job is executed. Thus, I would recommend you to execute your jobs with a long running Flink cluster on YARN.

Cheers,
Till

​

On Fri, Jan 1, 2016 at 11:29 PM, Filip Łęczycki <[hidden email]> wrote:

Hi all,

I am running filnk aps on YARN cluster and I am trying to get some benchmarks. When I start a long-running flink cluster on my YARN cluster I have an access to web UI and rest API that provide me statistics of the deployed jobs (as desribed here: https://ci.apache.org/projects/flink/flink-docs-master/internals/monitoring_rest_api.html). I was wondering is this possible to get such information about a single run job trigerred with 'flink run -m yarn-cluster ...'? After the job is finished there is no flink client running so I cannot use rest api to get stats. 

Thanks for any help:)

Best regards/Pozdrawiam,
Filip Łęczycki

[Stephan Ewen]

Hi Filip!

There are thoughts and efforts to extend Flink to push the result statistics of Flink jobs to the YARN timeline server. That way, you can explore jobs that are completed.

Since the whole web dashboard in Flink has a pure REST design, this is a quite straightforward fix.

From the capacities I see in the community, I can not promise that to be fixed immediately. Let me know, though, if you are interested in contributing an addition there, and I can walk you through the steps that would be needed.

Greetings,

Stephan

On Mon, Jan 4, 2016 at 9:17 PM, Filip Łęczycki <[hidden email]> wrote:

Hi Till,

Thank you for you answer however I am sorry to hear that. I was reluctant to execute jobs with long running Flink cluster due to the fact that multiple jobs would cloud yarn statistics regarding cpu and memory time as well as Flink's garbage collector statistics in log, as they would be stored for the whole Flink cluster, instead of a single job. 

Do you know whether is there a way to extract mentioned stats (cpu time, mem time, gc time) for a single job ran on long running Flink cluster?

I will be very grateful for an answer:)

Best regards,

Filip

Pozdrawiam,
Filip Łęczycki

2016-01-04 10:05 GMT+01:00 Till Rohrmann <[hidden email]>:

Hi Filip,

at the moment it is not possible to retrieve the job statistics after the job has finished with flink run -m yarn-cluster. The reason is that the YARN cluster is only alive as long as the job is executed. Thus, I would recommend you to execute your jobs with a long running Flink cluster on YARN.

Cheers,
Till

​

On Fri, Jan 1, 2016 at 11:29 PM, Filip Łęczycki <[hidden email]> wrote:

Hi all,

I am running filnk aps on YARN cluster and I am trying to get some benchmarks. When I start a long-running flink cluster on my YARN cluster I have an access to web UI and rest API that provide me statistics of the deployed jobs (as desribed here: https://ci.apache.org/projects/flink/flink-docs-master/internals/monitoring_rest_api.html). I was wondering is this possible to get such information about a single run job trigerred with 'flink run -m yarn-cluster ...'? After the job is finished there is no flink client running so I cannot use rest api to get stats. 

Thanks for any help:)

Best regards/Pozdrawiam,
Filip Łęczycki

[Filip Łęczycki]

Hi Stephan,

Thank you for you answer. I would love to contribute but currently I have no capacity as I am buried with my thesis. 

I will reach out after graduating :)

Bestr regards.

Filip

2016-01-05 10:35 GMT+01:00 Stephan Ewen <[hidden email]>:

Hi Filip!

There are thoughts and efforts to extend Flink to push the result statistics of Flink jobs to the YARN timeline server. That way, you can explore jobs that are completed.

Since the whole web dashboard in Flink has a pure REST design, this is a quite straightforward fix.

From the capacities I see in the community, I can not promise that to be fixed immediately. Let me know, though, if you are interested in contributing an addition there, and I can walk you through the steps that would be needed.

Greetings,

Stephan

On Mon, Jan 4, 2016 at 9:17 PM, Filip Łęczycki <[hidden email]> wrote:

Hi Till,

Thank you for you answer however I am sorry to hear that. I was reluctant to execute jobs with long running Flink cluster due to the fact that multiple jobs would cloud yarn statistics regarding cpu and memory time as well as Flink's garbage collector statistics in log, as they would be stored for the whole Flink cluster, instead of a single job. 

Do you know whether is there a way to extract mentioned stats (cpu time, mem time, gc time) for a single job ran on long running Flink cluster?

I will be very grateful for an answer:)

Best regards,

Filip

Pozdrawiam,
Filip Łęczycki

2016-01-04 10:05 GMT+01:00 Till Rohrmann <[hidden email]>:

Hi Filip,

at the moment it is not possible to retrieve the job statistics after the job has finished with flink run -m yarn-cluster. The reason is that the YARN cluster is only alive as long as the job is executed. Thus, I would recommend you to execute your jobs with a long running Flink cluster on YARN.

Cheers,
Till

​

On Fri, Jan 1, 2016 at 11:29 PM, Filip Łęczycki <[hidden email]> wrote:

Hi all,

I am running filnk aps on YARN cluster and I am trying to get some benchmarks. When I start a long-running flink cluster on my YARN cluster I have an access to web UI and rest API that provide me statistics of the deployed jobs (as desribed here: https://ci.apache.org/projects/flink/flink-docs-master/internals/monitoring_rest_api.html). I was wondering is this possible to get such information about a single run job trigerred with 'flink run -m yarn-cluster ...'? After the job is finished there is no flink client running so I cannot use rest api to get stats. 

Thanks for any help:)

Best regards/Pozdrawiam,
Filip Łęczycki

[Stephan Ewen]

Good luck with the Thesis!

On Wed, Jan 6, 2016 at 6:18 PM, Filip Łęczycki <[hidden email]> wrote:

Hi Stephan,

Thank you for you answer. I would love to contribute but currently I have no capacity as I am buried with my thesis. 

I will reach out after graduating :)

Bestr regards.

Filip

Pozdrawiam,
Filip Łęczycki

2016-01-05 10:35 GMT+01:00 Stephan Ewen <[hidden email]>:

Hi Filip!

There are thoughts and efforts to extend Flink to push the result statistics of Flink jobs to the YARN timeline server. That way, you can explore jobs that are completed.

Since the whole web dashboard in Flink has a pure REST design, this is a quite straightforward fix.

From the capacities I see in the community, I can not promise that to be fixed immediately. Let me know, though, if you are interested in contributing an addition there, and I can walk you through the steps that would be needed.

Greetings,

Stephan

On Mon, Jan 4, 2016 at 9:17 PM, Filip Łęczycki <[hidden email]> wrote:

Hi Till,

Thank you for you answer however I am sorry to hear that. I was reluctant to execute jobs with long running Flink cluster due to the fact that multiple jobs would cloud yarn statistics regarding cpu and memory time as well as Flink's garbage collector statistics in log, as they would be stored for the whole Flink cluster, instead of a single job. 

Do you know whether is there a way to extract mentioned stats (cpu time, mem time, gc time) for a single job ran on long running Flink cluster?

I will be very grateful for an answer:)

Best regards,

Filip

Pozdrawiam,
Filip Łęczycki

2016-01-04 10:05 GMT+01:00 Till Rohrmann <[hidden email]>:

Hi Filip,

at the moment it is not possible to retrieve the job statistics after the job has finished with flink run -m yarn-cluster. The reason is that the YARN cluster is only alive as long as the job is executed. Thus, I would recommend you to execute your jobs with a long running Flink cluster on YARN.

Cheers,
Till

​

On Fri, Jan 1, 2016 at 11:29 PM, Filip Łęczycki <[hidden email]> wrote:

Hi all,

I am running filnk aps on YARN cluster and I am trying to get some benchmarks. When I start a long-running flink cluster on my YARN cluster I have an access to web UI and rest API that provide me statistics of the deployed jobs (as desribed here: https://ci.apache.org/projects/flink/flink-docs-master/internals/monitoring_rest_api.html). I was wondering is this possible to get such information about a single run job trigerred with 'flink run -m yarn-cluster ...'? After the job is finished there is no flink client running so I cannot use rest api to get stats. 

Thanks for any help:)

Best regards/Pozdrawiam,
Filip Łęczycki