Re: Help: how to get latency value comfortable in Flink1.5?

[Till Rohrmann]

Hi Rui,

1) Flink should always fetch records from Kafka if there some independent of the parallelism of the consumer. The only problem which could appear is that if you set the parallelism higher than the number of partitions, some of the source operators won't get a partition assigned. Due to that, they won't read data from Kafka.

2) numRecordsOutPerSecond should give you a good value for the throughput.

3) Could you check the logs whether the GraphiteReporter could be properly started? Maybe the reporter jar has not been put in the lib folder.

Cheers,

Till

On Tue, Sep 25, 2018 at 11:30 AM Wang, Rui2 <[hidden email]> wrote:

Hi Till,

 

I have integrated Kafka 2.0 with Flink1.5.2, and it works well. Now, I need to know the performance of Flink, such as throughput vs latency.

Here I just set my stream pipeline as: Kafka(source) -> flatMap -> map -> print(sink), with Kafka partitions:  32.  I confront some issues:

 

  1). If set each parallelism to 1, the job runs ok, but flatMap couldn’t receive records from Kafka.

Well, I just set the parallelism of flatMap above 1, the pipeline work fine. Does  Flink must be parallel when fetch records from Kafka?

 

  2).Can I get the throughput by "numRecordsOutPerSecond"?

Also, I think it may be right to override the open function which extends RichFlatMapFunction.

 

  3).Regarding to the latency of operator(e.g. flatMap), after setting env.getConfig().setLatencyTrackingInterval(1000),  I can get the latency only by the URL as below:

input : http://server:8081/jobs/jobid/metrics?get=latency.source_id.X1.source_subtask_index.22.operator_id.Y1.operator_subtask_index.22.latency_stddev

result: [{"id":"latency.source_id.X1.source_subtask_index.22.operator_id.Y1.operator_subtask_index.22.latency_stddev","value":"0.3854355088676483"}]

 

But, it's a poor user experience, so I want a third-party to gather the value and display it, such as Graphite. I have configured the metrics in flink-conf.yaml as list below,

 

metrics.reporters: varuy

metrics.reporters.varuy.host: tracing115

metrics.reporters.varuy.port: 2003

metrics.reporters.varuy.class: org.apache.flink.metrics.graphite.GraphiteReporter

metrics.reporters.varuy.interval: 1 SECONDS

metrics.reporters.varuy.protocol: TCP

 

when use cmd "lsof -i:2003" nothing return, I think the graphite seems not work.

Could you give some advice to continue my work when you’re free, I’m looking forward to your reply!

 

Best Regards & Thanks

Rui, Wang

[varuy322]

Hi Till,

I have copy the flink-metrics-graphite-1.5.2.jar to lib/, and restart
flink.It seems make no difference.
When I checked the jobmanager log, the metrics configured in
flink-conf.yaml, could be load since it printed as below:

"2018-09-27 09:50:23,953 INFO
org.apache.flink.configuration.GlobalConfiguration            - Loading
configuration property: metrics.reporters, varuy
2018-09-27 09:50:23,953 INFO
org.apache.flink.configuration.GlobalConfiguration            - Loading
configuration property: metrics.reporters.varuy.host, tracing115
2018-09-27 09:50:23,953 INFO
org.apache.flink.configuration.GlobalConfiguration            - Loading
configuration property: metrics.reporters.varuy.port, 2003
2018-09-27 09:50:23,953 INFO
org.apache.flink.configuration.GlobalConfiguration            - Loading
configuration property: metrics.reporters.varuy.class,
org.apache.flink.metrics.graphite.GraphiteReporter
2018-09-27 09:50:23,954 INFO
org.apache.flink.configuration.GlobalConfiguration            - Loading
configuration property: metrics.reporters.varuy.interval, 1 SECONDS
2018-09-27 09:50:23,954 INFO
org.apache.flink.configuration.GlobalConfiguration            - Loading
configuration property: metrics.reporters.varuy.protocol, TCP"

However, I found the GraphiteReporter do not run correctly for the info
list:

"2018-09-27 09:50:24,938 INFO
org.apache.flink.runtime.metrics.MetricRegistryImpl           - No metrics
reporter configured, no metrics will be exposed/reported."

Should I install Graphite Server independently?

Best Regards & Thanks

Rui,Wang
 



-----
stay hungry, stay foolish.
--
Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/

[Chesnay Schepler]

So there's 2 issues here:
1) your reporter configuration is wrong, configuration values for a
specific reporter are prefixed with "metrics.reporter", not
"metrics.reporters" (note the "s"). See below for a correct config.

metrics.reporters: varuy
metrics.reporter.varuy.host: tracing115
metrics.reporter.varuy.port: 2003
metrics.reporter.varuy.class:
org.apache.flink.metrics.graphite.GraphiteReporter
metrics.reporter.varuy.interval: 1 SECONDS
metrics.reporter.varuy.protocol: TCP

2) The reporter only exports data to an external graphite server, so yes
you have to start that one separately.

On 27.09.2018 08:13, varuy322 wrote:

> Hi Till,
>
> I have copy the flink-metrics-graphite-1.5.2.jar to lib/, and restart
> flink.It seems make no difference.
> When I checked the jobmanager log, the metrics configured in
> flink-conf.yaml, could be load since it printed as below:
>
> "2018-09-27 09:50:23,953 INFO
> org.apache.flink.configuration.GlobalConfiguration            - Loading
> configuration property: metrics.reporters, varuy
> 2018-09-27 09:50:23,953 INFO
> org.apache.flink.configuration.GlobalConfiguration            - Loading
> configuration property: metrics.reporters.varuy.host, tracing115
> 2018-09-27 09:50:23,953 INFO
> org.apache.flink.configuration.GlobalConfiguration            - Loading
> configuration property: metrics.reporters.varuy.port, 2003
> 2018-09-27 09:50:23,953 INFO
> org.apache.flink.configuration.GlobalConfiguration            - Loading
> configuration property: metrics.reporters.varuy.class,
> org.apache.flink.metrics.graphite.GraphiteReporter
> 2018-09-27 09:50:23,954 INFO
> org.apache.flink.configuration.GlobalConfiguration            - Loading
> configuration property: metrics.reporters.varuy.interval, 1 SECONDS
> 2018-09-27 09:50:23,954 INFO
> org.apache.flink.configuration.GlobalConfiguration            - Loading
> configuration property: metrics.reporters.varuy.protocol, TCP"
>
> However, I found the GraphiteReporter do not run correctly for the info
> list:
>
> "2018-09-27 09:50:24,938 INFO
> org.apache.flink.runtime.metrics.MetricRegistryImpl           - No metrics
> reporter configured, no metrics will be exposed/reported."
>
> Should I install Graphite Server independently?
>
> Best Regards & Thanks
>
> Rui,Wang
>  
>
>
>
> -----
> stay hungry, stay foolish.
> --
> Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
>