能否将自定义数值数据存入Prometheus？如何推送该类数据？

Absolutely! Prometheus fully supports storing custom numerical data — this is a core capability that lets you monitor everything from application-specific metrics to custom sensor readings. Prometheus works with numerical metric types like counters (for incrementing values), gauges (for arbitrary numeric values that go up/down), histograms, and summaries—all perfect for custom numerical data.

Here are the two most common ways to push custom numerical data to Prometheus:

1. Use the Prometheus Pushgateway (for short-lived jobs/scripts) #

The Pushgateway is an official tool designed for scenarios where your workload is short-lived (like cron jobs, one-off scripts) and can't expose an endpoint for Prometheus to scrape. Here's how to use it:

Step 1: Install and start Pushgateway #

You can run it directly from the binary:

./pushgateway

Or use Docker:

docker run -p 9091:9091 prom/pushgateway

Step 2: Push your custom data #

You can use curl to send raw metric data, or use a client library (more on that later). For example, to push a gauge metric tracking device temperature:

echo "custom_device_temperature{device_id=\"sensor_001\", location=\"server_room\"} 24.7" | curl --data-binary @- http://localhost:9091/metrics/job/temperature_monitor

Breakdown:

• custom_device_temperature: Your custom metric name
• {device_id="sensor_001", ...}: Labels to add context to your data
• 24.7: The numerical value you want to store
• job/temperature_monitor: A job label to group related metrics

Step 3: Configure Prometheus to scrape Pushgateway #

Add this to your prometheus.yml config file:

scrape_configs:
  - job_name: 'pushgateway'
    static_configs:
      - targets: ['localhost:9091']

Restart Prometheus, and it will start pulling your custom data from the Pushgateway.

2. Use official client libraries (for long-running services or scripted pushes) #

For long-running applications (like web services, daemons), the best practice is to expose a metrics endpoint for Prometheus to pull data (aligning with Prometheus's pull-based model). But you can also use these libraries to push data directly to Pushgateway if needed.

Here's a quick Python example using the prometheus_client library:

Step 1: Install the library #

pip install prometheus-client

Step 2: Push custom data via code #

from prometheus_client import Gauge, push_to_gateway, CollectorRegistry

# Create a registry to hold your metrics
registry = CollectorRegistry()

# Define a gauge metric with labels
custom_temp = Gauge(
    'custom_device_temperature',
    'Current temperature of IoT device',
    ['device_id', 'location'],
    registry=registry
)

# Set your custom numerical value
custom_temp.labels(device_id='sensor_001', location='server_room').set(24.7)

# Push to Pushgateway
push_to_gateway('localhost:9091', job='temperature_monitor', registry=registry)

If you want Prometheus to pull data instead, you can expose an HTTP endpoint with start_http_server(8000) and let Prometheus scrape http://your-service:8000/metrics.

Key Notes #

• Use Pushgateway only for short-lived jobs: It holds metrics until they're scraped, but doesn't handle deduplication or expiration automatically.
• For long-running services, prefer the pull model (exposing an endpoint) — it's more aligned with Prometheus's design and easier to maintain.

内容的提问来源于stack exchange，提问作者Deepak N