Skip to main content
Scraping metrics is preferred whenever possible, but you can use the Prometheus Pushgateway for ephemeral jobs such as batch, AWS Lambda, or Google Cloud Functions that the Collector couldn’t scrape otherwise. As of v0.98.0, the Collector provides an HTTP endpoint that’s compatible with the Prometheus Pushgateway, with the following caveats:
  • The open source Prometheus Pushgateway is a cache for metrics. Because the Collector is a proxy, it forwards metric payloads to the Chronosphere backend instead of caching metrics. This distinction means:
    • HTTP DELETE requests aren’t supported.
    • HTTP PUT and POST requests are semantically the same.
    • The Query API, or GET requests, aren’t supported.
  • Neither the Prometheus Pushgateway nor the Collector support any aggregation.
  • The Collector only supports GZIP compressed payloads. Snappy-compressed payloads aren’t supported.
  • Transport Layer Security (TLS) isn’t supported. All traffic to the Collector is in plain text, but traffic to the Chronosphere backend is always encrypted.
Adhere to the following best practices when using the Prometheus Pushgateway with the Collector:
  • Ensure that at least one of the labels are unique to the instance of the calling job so that time series don’t collide. If the label IDs are sufficiently high in cardinality, you might want to aggregate the labels.
  • Push metrics periodically, ideally when a job shuts down.
  • Run Collectors as sidecars (where all traffic is over localhost), or as a formal Kubernetes service.
To enable the Prometheus Pushgateway for the Collector, modify the Collector configuration file and add the following lines: 
# This will register an HTTP endpoint on the listenerAddress for the Collector.
push:
  pushgateway:
    enabled: true

​
Examples

The following examples show how to send traffic with the Prometheus Pushgateway to the Collector, assuming that the listen address is running on port 3030.

​
Using curl

When running this curl command, replace <collector-ip> with the IP address of your Collector. 
curl --data-binary @- http://<collector-ip>:3030/pushgateway/metrics/job/some_job/job_id/my_unique_id <<EOF
# TYPE some_metric counter
some_metric{label="val1"} 42
# TYPE another_metric gauge
# HELP another_metric Just an example.
another_metric 2398.283
EOF
If you want to use localhost instead of the IP address for a Collector that’s running locally (such as for testing purposes), you need to enable port forwarding first. If your Collector is running as a Kubernetes DaemonSet or as a Kubernetes Deployment, run the following command to enable port forwarding, where <chronocollector-pod-name> is the name of the Kubernetes pod where the Collector is running: 
kubectl port-forward <chronocollector-pod-name> 3030
You can then run the previous curl command and use localhost instead of the Collector IP address.

​
Using the Prometheus Pushgateway client

import (
    "github.com/prometheus/client_golang/prometheus"
    "github.com/prometheus/client_golang/prometheus/push"
)

func main() {
    registry := prometheus.NewRegistry()

    // Push metrics on shutdown.
    defer func() {
        // The address can be replaced with a Kubernetes Service name,
        // however, the URL must end in /pushgateway.
        push.New("http://localhost:3030/pushgateway", "my_job").
            Grouping("job_id", "my_unique_id").Gatherer(registry).Push()
    }()
}