SignalFx Python Library

1.1.16 verified Mon Apr 27 auth: no python maintenance

SignalFx Python Library provides a client for sending metrics and events to SignalFx, and for interacting with SignalFlow computations. Current version 1.1.16, released 2022-07. Maintenance mode; no active development.

pip install signalfx

Common errors

error ValueError: Unknown type for value: ... ↓

cause Sending invalid metric value type (e.g., string, list). Values must be int or float.

fix

Ensure metric value is numeric: {'metric': 'cpu', 'value': float(cpu_percent)}

error signalfx.exceptions.SignalFxException: 401 Unauthorized ↓

cause Invalid or missing access token.

fix

Set SIGNALFX_ACCESS_TOKEN environment variable or pass token to SignalFx().

error ImportError: cannot import name 'SignalFx' from 'signalfx' (unknown location) ↓

cause Old Python version missing the module or corrupted install.

fix

pip install --upgrade signalfx. On Python 2.7, use signalfx <=1.0.19.

Warnings

breaking Versions 1.1.7 to 1.1.11 contain serious bugs (HTTP reconnection issues, data corruption). Upgrade directly to 1.1.12 or later. ↓

fix pip install 'signalfx>=1.1.12'

deprecated The 'pyformance' integration (SignalFxRegistry) is deprecated and removed in later versions. Use OpenTelemetry or custom metrics instead. ↓

fix Migrate to OpenTelemetry SDK or direct ingest via signalfx.

gotcha The SignalFlow API (get_computation, etc.) uses a different token (stream token) than ingest token. Must use SignalFx(stream_endpoint=...) or set SIGNALFX_ACCESS_TOKEN for ingest, but for SignalFlow you need to use the SignalFx(api_endpoint=...) with the stream token. ↓

fix Use separate SignalFx instances for ingest and SignalFlow with appropriate tokens.

Imports

SignalFx
wrong
```
import signalfx
```
correct
```
from signalfx import SignalFx
```
SignalFx class is under the signalfx package, not a top-level import.

SignalFxClient

wrong

from signalfx.client import SignalFxClient

correct

from signalfx import SignalFx

Use SignalFx class; client classes are internal.

Quickstart

Send a datapoint and event using the v1 SDK with a context manager.

import os
from signalfx import SignalFx

# Set token via environment variable
sfx = SignalFx(api_endpoint='https://ingest.signalfx.com', ingest_timeout_seconds=5)
# If using a realm: sfx = SignalFx(api_endpoint='https://ingest.<REALM>.signalfx.com')
with sfx.ingest(os.environ.get('SIGNALFX_ACCESS_TOKEN', '')) as ingest:
    ingest.send_datapoints([
        {
            'metric': 'cpu.utilization',
            'value': 42.0,
            'dimensions': {'host': 'server1'}
        }
    ])
    ingest.send_events([
        {
            'eventType': 'deployment',
            'dimensions': {'host': 'server1'}
        }
    ])