How to visualize the results of Flink processing or aggregation?

[Xiong Qiang]

Hi All, 

I am new to Flink, so forgive me if it is a naive question.

The context is:

We have a data streaming coming in, and we will use Flink applications to do the processing or aggregations. After the processing or aggregation, we need some approaches to visualize the results, to either build a dashboard or setup alerts, for example, using Prometheus and Grafana.

However, after reading the documents (https://flink.apache.org/flink-architecture.html and more links) and examples (https://ci.apache.org/projects/flink/flink-docs-release-1.12/try-flink/datastream_api.html) (https://github.com/ververica/flink-training/blob/master/long-ride-alerts/src/solution/java/org/apache/flink/training/solutions/longrides/LongRidesSolution.java) , I am still not able to close the gap between Flink and a monitoring/dashboard tool, e.g. Prometheus/Grafana. 

The question is:

How are processing results connected/sinked from Flink to Prometheus/Grafana? for example, in the fraud detection example, how is the account id = 3, send to Prometheus and Grafana, so that I have a dashboard showing there is one suspected account? In the taxi long rides example, how do I send the count of long rides from Flink to Prometheus/Grafana?

I understand there are sinks (https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/connectors/). However, I didn't see sinks for Prometheus.

Hope I made my question clear.

Thanks

[Fuyao Li-2]

Hello Xiong,

 

You can expose monitors through Metric system of Flink.

https://ci.apache.org/projects/flink/flink-docs-stable/ops/metrics.html

Metrics can be exposed by metric reporter:

https://ci.apache.org/projects/flink/flink-docs-release-1.12/deployment/metric_reporters.html

That includes Prometheus.

 

For the DataStream API, you can build customized metrics.

For the Flink SQL/Table API, you can only use the listed predefined metrics in Flink. There could be work arounds ways, but no direct way to supply Flink SQL customized metrics.

 

Best regards,

Fuyao

 

From: Xiong Qiang <[hidden email]>
Date: Friday, March 26, 2021 at 09:15
To: [hidden email] <[hidden email]>
Subject: [External] : How to visualize the results of Flink processing or aggregation?

Hi All, 

 

I am new to Flink, so forgive me if it is a naive question.

 

The context is:

We have a data streaming coming in, and we will use Flink applications to do the processing or aggregations. After the processing or aggregation, we need some approaches to visualize the results, to either build a dashboard or setup alerts, for example, using Prometheus and Grafana.

However, after reading the documents (https://flink.apache.org/flink-architecture.html and more links) and examples (https://ci.apache.org/projects/flink/flink-docs-release-1.12/try-flink/datastream_api.html) (https://github.com/ververica/flink-training/blob/master/long-ride-alerts/src/solution/java/org/apache/flink/training/solutions/longrides/LongRidesSolution.java) , I am still not able to close the gap between Flink and a monitoring/dashboard tool, e.g. Prometheus/Grafana. 

 

The question is:

How are processing results connected/sinked from Flink to Prometheus/Grafana? for example, in the fraud detection example, how is the account id = 3, send to Prometheus and Grafana, so that I have a dashboard showing there is one suspected account? In the taxi long rides example, how do I send the count of long rides from Flink to Prometheus/Grafana?

 

I understand there are sinks (https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/connectors/). However, I didn't see sinks for Prometheus.

 

Hope I made my question clear.

 

Thanks

[David Anderson-4]

Prometheus is a metrics system; you can use Flink's Prometheus metrics reporter to send metrics to Prometheus. 

Grafana can also be connected to influxdb, and to databases like mysql and postgresql, for which sinks are available.

And the Elasticsearch sink can be used to create visualizations with Kibana.

I'm sure there are other solutions as well, but these are some of the popular ones.

Regards,

David

On Fri, Mar 26, 2021 at 5:15 PM Xiong Qiang <[hidden email]> wrote:

Hi All, 

I am new to Flink, so forgive me if it is a naive question.

The context is:

We have a data streaming coming in, and we will use Flink applications to do the processing or aggregations. After the processing or aggregation, we need some approaches to visualize the results, to either build a dashboard or setup alerts, for example, using Prometheus and Grafana.

However, after reading the documents (https://flink.apache.org/flink-architecture.html and more links) and examples (https://ci.apache.org/projects/flink/flink-docs-release-1.12/try-flink/datastream_api.html) (https://github.com/ververica/flink-training/blob/master/long-ride-alerts/src/solution/java/org/apache/flink/training/solutions/longrides/LongRidesSolution.java) , I am still not able to close the gap between Flink and a monitoring/dashboard tool, e.g. Prometheus/Grafana. 

The question is:

How are processing results connected/sinked from Flink to Prometheus/Grafana? for example, in the fraud detection example, how is the account id = 3, send to Prometheus and Grafana, so that I have a dashboard showing there is one suspected account? In the taxi long rides example, how do I send the count of long rides from Flink to Prometheus/Grafana?

I understand there are sinks (https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/connectors/). However, I didn't see sinks for Prometheus.

Hope I made my question clear.

Thanks

[Xiong Qiang]

Thank you, @David Anderson and @Fuyao Li. This answered my question and cleared my confusions.

On Fri, Mar 26, 2021 at 11:08 AM David Anderson <[hidden email]> wrote:

Prometheus is a metrics system; you can use Flink's Prometheus metrics reporter to send metrics to Prometheus. 

Grafana can also be connected to influxdb, and to databases like mysql and postgresql, for which sinks are available.

And the Elasticsearch sink can be used to create visualizations with Kibana.

I'm sure there are other solutions as well, but these are some of the popular ones.

Regards,

David

On Fri, Mar 26, 2021 at 5:15 PM Xiong Qiang <[hidden email]> wrote:

Hi All, 

I am new to Flink, so forgive me if it is a naive question.

The context is:

We have a data streaming coming in, and we will use Flink applications to do the processing or aggregations. After the processing or aggregation, we need some approaches to visualize the results, to either build a dashboard or setup alerts, for example, using Prometheus and Grafana.

However, after reading the documents (https://flink.apache.org/flink-architecture.html and more links) and examples (https://ci.apache.org/projects/flink/flink-docs-release-1.12/try-flink/datastream_api.html) (https://github.com/ververica/flink-training/blob/master/long-ride-alerts/src/solution/java/org/apache/flink/training/solutions/longrides/LongRidesSolution.java) , I am still not able to close the gap between Flink and a monitoring/dashboard tool, e.g. Prometheus/Grafana. 

The question is:

How are processing results connected/sinked from Flink to Prometheus/Grafana? for example, in the fraud detection example, how is the account id = 3, send to Prometheus and Grafana, so that I have a dashboard showing there is one suspected account? In the taxi long rides example, how do I send the count of long rides from Flink to Prometheus/Grafana?

I understand there are sinks (https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/connectors/). However, I didn't see sinks for Prometheus.

Hope I made my question clear.

Thanks