Prometheus on Docker

In this article, we shall explore how to run Prometheus as a docker container, and also see how it captures and displays application metrics.

What is Prometheus

Prometheus is a time series database that can scrape time-series data through HTTP pull at given time intervals. It is good to store metrics about applications.

We use Prometheus in our company to store and analyze application metrics of our spring-boot micro-services.

Why on Docker

Prometheus is officially released as a docker container - and is the best way to use it for learning. It takes away the pain of installing its dependencies and Prometheus itself. Official docker documentation is available here.

Run

Prometheus is configured through yml files. Let us create a folder and put a configuration file in it first.

mkdir ~/docker/prometheus

And let us create a yml file

touch prometheus.yml

And now run docker image by providing newly created folder as a shared volume

docker run \

  --detach \

  --name prometheus \

  --publish 9090:9090 \

  --volume /Users/hsingh/docker/prometheus:/etc/prometheus \

  prom/prometheus

Test

Prometheus exposes its metrics in its format. So let us use it.

First, open http://localhost:9090/metrics to understand the format and also see what data is being exposed by Prometheus as metrics.

The format is simple - multiple lines of

any_label data

Edit prometheus.yml file to read this:

global:

  external_labels:

    monitor: 'codelab-monitor'

scrape_configs:

  - job_name: 'prometheus'

    scrape_interval: 5s

    static_configs:

      - targets: ['localhost:9090']

scrape_configs: --> it defines scrape jobs - usually 1 per application we wish to monitor.

rest all are self-explanatory

Note: targets is an array. So we can monitor multiple instances of the same application - useful for multi-instance micro-services.

Restart docker image after updating yml. The documentation says that Prometheus auto-reloads. It didn't happen in my case.

Now to see metrics in Prometheus, open localhost:9090.

Got to Menu:: Status > Targets

  It tells the jobs that are currently running in Prometheus - i.e. applications being monitored. You should see an entry for prometheus itself.

Go to Menu:: Graph 

  Enter any metric in the text box. e.g. 'prometheus_target_interval_length_seconds'. And press the "Execute" button. We should see a table format of data. Change tab to "Graph" and we can see the graphical representation of the same data

Conclusion

We have explored

• running Prometheus as a docker container
• tested using Prometheus's metrics

In our company, we do not use Prometheus in isolation. Our spring boot micro-services are monitored using

1. Micro-meter (middleware to expose spring boot metrics in Prometheus format)
2. Prometheus (collects metric as time series)
3. Grafana (for visualization)