Scrape Astra Streaming metrics with Prometheus

Prometheus collects system metrics by scraping targets at intervals. These metrics are used to monitor deployments, generate alerts, and diagnose problems.

This doc will show you how to scrape an Astra Streaming tenant with Prometheus.

Prerequisites

Astra Streaming tenant
Docker installed locally

Get the configuration file from Astra Streaming

In the Astra Portal navigation menu, click Streaming, and then click the name of your tenant.

On the Connect tab, click Prometheus to generate a new configuration file:

prometheus.yml

global:
  scrape_interval: 60s
  evaluation_interval: 60s

scrape_configs:
  - job_name: "astra-pulsar-metrics"

    scheme: 'https'
    metrics_path: '/pulsarmetrics/TENANT_NAME'
    authorization:
      credentials: 'PULSAR_TOKEN'

    static_configs:
      - targets: [https://prometheus-PROVIDER-REGION.streaming.datastax.com/pulsarmetrics/TENANT_NAME]

The default prometheus.yml scrapes the target tenant every 60 seconds, and it uses the specified PULSAR_TOKEN for authentication. The job_name is added as a label to any timeseries scraped with this configuration.

Copy the configuration YAML or download the prometheus.yml file.

Build Prometheus

Prometheus runs with a prometheus.yml file available locally or in a Docker container. This example uses Docker.

Pull the Prometheus Docker image:
```
docker pull prom/prometheus
```
Bind-mount your modified prometheus.yml file by running the Prometheus Docker container with a modified path in the -v argument:
```
docker run \
    -p 9090:9090 \
    -v /**PATH_TO_PROMETHEUS_YAML**:/etc/prometheus/prometheus.yml \
    prom/prometheus
```

Make sure the response is similar to the following example. This indicates that Prometheus is ready to scrape your Astra Streaming tenant:

ts=2022-05-10T20:40:30.877Z caller=main.go:1199 level=info msg="Completed loading of configuration file" filename=/etc/prometheus/prometheus.yml totalDuration=2.75025ms db_storage=584ns remote_storage=708ns web_handler=167ns query_engine=416ns scrape=262.125µs scrape_sd=12.208µs notify=667ns notify_sd=792ns rules=1.042µs tracing=2.959µs
ts=2022-05-10T20:40:30.877Z caller=main.go:930 level=info msg="Server is ready to receive web requests."

If you get a mounts denied permissions error, make sure your local directory is shared with Docker. In Docker Desktop, click File Sharing, click Resources, and then make sure your local directory is listed.

Scrape with Prometheus

Navigate to your Prometheus dashboard at localhost:9090.
Click Status, select Targets, and then make sure your status_config.targets endpoint’s status is UP.
Navigate to http://localhost:9090/graph, and click the Graph tab.
In the Expression field, enter pulsar_in_messages_total, and then click Execute. Prometheus now shows the total incoming Apache Pulsar™ messages for your Astra Streaming cluster.
To test the integration, produce a few messages in your tenant with the Pulsar CLI or the Astra Streaming Websocket. Make sure your Prometheus graph displays the number of incoming Pulsar messages with each 60 second scrape:

Content encoding

Astra Streaming supports content encoding with either gzip or deflate.

With your Prometheus scrape container running, you can use curl commands to decompress your Prometheus scrape data:

Deflate
Gzip

curl -v -H 'Authorization: Bearer $BEARER_TOKEN' -H 'Accept-encoding: deflate' https://prometheus-aws-useast1.streaming.datastax.com/pulsarmetrics/tenant-1 --compressed

curl -v -H 'Authorization: Bearer $BEARER_TOKEN' -H 'Accept-encoding: gzip' https://prometheus-aws-useast1.streaming.datastax.com/pulsarmetrics/tenant-1 | zcat

Deflate or Gzip will extract your scraped metrics in a format such as the following:

# TYPE pulsar_topics_count untyped
pulsar_topics_count{app="pulsar",cluster="pulsar-aws-useast1",component="broker",controller_revision_hash="pulsar-aws-useast1-broker-7444bf6f64",instance="192.168.2.120:8080",job="broker",kubernetes_namespace="pulsar",kubernetes_pod_name="pulsar-aws-useast1-broker-1",namespace="mk-tenant/default",release="astraproduction-aws-useast1-pulsar",statefulset_kubernetes_io_pod_name="pulsar-aws-useast1-broker-1",prometheus="pulsar/astraproduction-aws-useast-prometheus",prometheus_replica="prometheus-astraproduction-aws-useast-prometheus-0"} 1 1654550685678

Metrics exposed by Astra Streaming

The following Prometheus metrics are exposed by Astra Streaming:

Cluster operational metrics are not exposed to individual cluster tenants. A tenant can only access its own metrics on the broker pod or function worker pod.