Skip to content

Prometheus instrumentation library for Python applications (with UWSGI support)

License

Notifications You must be signed in to change notification settings

Lispython/pyprometheus

Repository files navigation

Prometheus instrumentation library for Python applications

The unofficial Python 2 and 3 client for Prometheus.

https://travis-ci.org/Lispython/pyprometheus.svg?branch=master

Features

  • Four types of metric are supported: Counter, Gauge, Summary(without quantiles) and Histogram.
  • InMemoryStorage (do not use it for multiprocessing apps)
  • UWSGI storage - share metrics between processes
  • UWAGI flush storage - sync metrics with uwsgi sharedarea by flush call
  • time decorator
  • time context manager

INSTALLATION

To use pyprometheus use pip or easy_install:

pip install pyprometheus

or

easy_install pyprometheus

HOW TO INSTRUMENTING CODE

Gauge

A gauge is a metric that represents a single numerical value that can arbitrarily go up and down.:

from pyprometheus import Gauge
from pyprometheus import BaseRegistry, LocalMemoryStorage

storage = LocalMemoryStorage()
registry = CollectorRegistry(storage=storage)
gauge = Gauge("job_in_progress", "Description", registry=registry)

gauge.inc(10)
gauge.dec(5)
gauge.set(21.1)

utilities:

gauge.set_to_current_time()   # Set to current unixtime

# Increment when entered, decrement when exited.
@gauge.track_in_progress()
def f():
    pass

with gauge.track_in_progress():
    pass


with gauge.time():
    time.sleep(10)

Counter

A counter is a cumulative metric that represents a single numerical value that only ever goes up.:

from pyprometheus import Counter
from pyprometheus import BaseRegistry, LocalMemoryStorage

storage = LocalMemoryStorage()
registry = CollectorRegistry(storage=storage)
counter = Counter("requests_total", "Description", registry=registry)

counter.inc(10)

Summary

Similar to a histogram, a summary samples observations (usually things like request durations and response sizes).:

from pyprometheus import Summary
from pyprometheus import BaseRegistry, LocalMemoryStorage

storage = LocalMemoryStorage()
registry = CollectorRegistry(storage=storage)
s = Summary("requests_duration_seconds", "Description", registry=registry)

s.observe(0.100)

utilities for timing code:

@gauge.time()
def func():
   time.sleep(10)

with gauge.time():
   time.sleep(10)

Histogram

A histogram samples observations (usually things like request durations or response sizes) and counts them in configurable buckets. It also provides a sum of all observed values.:

from pyprometheus import Summary
 from pyprometheus import BaseRegistry, LocalMemoryStorage

 storage = LocalMemoryStorage()
 registry = CollectorRegistry(storage=storage)
 histogram = Histogram("requests_duration_seconds", "Description", registry=registry)

 histogram.observe(1.1)

utilities for timing code:

@histogram.time()
def func():
   time.sleep(10)

with histogram.time():
   time.sleep(10)

Labels

All metrics can have labels, allowing grouping of related time series.

Example:

from pyprometheus import Counter
c = Counter('my_requests_total', 'HTTP Failures', ['method', 'endpoint'])
c.labels('get', '/').inc()
c.labels('post', '/submit').inc()

or labels as keyword arguments:

from pyprometheus import Counter
c = Counter('my_requests_total', 'HTTP Failures', ['method', 'endpoint'])
c.labels(method='get', endpoint='/').inc()
c.labels(method='post', endpoint='/submit').inc()

STORAGES

Currently library support 2 storages: LocalMemoryStorage and UWSGIStorage

Every registry MUST have link to storage:

from pyprometheus import BaseRegistry, LocalMemoryStorage

storage = LocalMemoryStorage()
registry = CollectorRegistry(storage=storage)

Use LocalMemoryStorage

Simple storage that store samples to application memory. It can be used with threads.:

from pyprometheus import BaseRegistry, LocalMemoryStorage

storage = LocalMemoryStorag()

Use UWSGIStorage

UWSGIStorage allow to use uwsgi sharedarea to sync metrics between processes.:

from pyprometheus.contrib.uwsgi_features import UWSGICollector, UWSGIStorage

SHAREDAREA_ID = 0
storage = UWSGIStorage(SHAREDAREA_ID)

also need to configure UWSGI sharedaread pages.

EXPORTING

Library have some helpers to export metrics

To text format

You can convert registry to text format:

from pyprometheus import BaseRegistry, LocalMemoryStorage
from pyprometheus.utils.exposition import registry_to_text
from pyprometheus import Gauge

storage = LocalMemoryStorage()
registry = CollectorRegistry(storage=storage)
g = Gauge('raid_status', '1 if raid array is okay', registry=registry)
g.set(1)
print(registry_to_text(registry))

Text file export

This is useful for monitoring cronjobs, or for writing cronjobs to expose metrics about a machine system.:

from pyprometheus import BaseRegistry, LocalMemoryStorage
from pyprometheus.utils.exposition import registry_to_text, write_to_textfile
from pyprometheus import Gauge

storage = LocalMemoryStorage()
registry = CollectorRegistry(storage=storage)
g = Gauge('raid_status', '1 if raid array is okay', registry=registry)
g.set(1)
write_to_textfile(registry, "/path/to/file/metrics.prom")

You can configure text file collector to use generated file.

TODO

Some features that we plan to do:

  • [ ] Add mmap storage
  • [ ] Add features for async frameworks
  • [ ] Optimize UWSGI storage byte pad
  • [ ] Add quantiles

EXAMPLE PROJECT

We create example project to show hot to use pyprometheus in real project.

CONTRIBUTE

Fork https://github.com/Lispython/pyprometheus/ , create commit and pull request to develop.