Prometheus instrumentation library for PHP applications.

v1.2.0 2023-11-21 10:01 UTC


A prometheus client library written in PHP


This library uses Redis or APCu to do the client side aggregation. If using Redis, we recommend to run a local Redis instance next to your PHP workers.

How does it work?

Usually PHP worker processes don't share any state. You can pick from three adapters. Redis, APC or an in memory adapter. While the first needs a separate binary running, the second just needs the APC extension to be installed. If you don't need persistent metrics between requests (e.g. a long running cron job or script) the in memory adapter might be suitable to use.


Add as Composer dependency:

composer require endclothing/prometheus_client_php


A simple counter:

    ->getOrRegisterCounter('', 'some_quick_counter', 'just a quick measurement')

Write some enhanced metrics:

$registry = \Prometheus\CollectorRegistry::getDefault();

$counter = $registry->getOrRegisterCounter('test', 'some_counter', 'it increases', ['type']);
$counter->incBy(3, ['blue']);

$gauge = $registry->getOrRegisterGauge('test', 'some_gauge', 'it sets', ['type']);
$gauge->set(2.5, ['blue']);

$histogram = $registry->getOrRegisterHistogram('test', 'some_histogram', 'it observes', ['type'], [0.1, 1, 2, 3.5, 4, 5, 6, 7, 8, 9]);
$histogram->observe(3.5, ['blue']);

Manually register and retrieve metrics (these steps are combined in the getOrRegister... methods):

$registry = \Prometheus\CollectorRegistry::getDefault();

$counterA = $registry->registerCounter('test', 'some_counter', 'it increases', ['type']);
$counterA->incBy(3, ['blue']);

// once a metric is registered, it can be retrieved using e.g. getCounter:
$counterB = $registry->getCounter('test', 'some_counter')
$counterB->incBy(2, ['red']);

Expose the metrics:

$registry = \Prometheus\CollectorRegistry::getDefault();

$renderer = new RenderTextFormat();
$result = $renderer->render($registry->getMetricFamilySamples());

header('Content-type: ' . RenderTextFormat::MIME_TYPE);
echo $result;

Change the Redis options (the example shows the defaults):

        'host' => '',
        'port' => 6379,
        'password' => null,
        'timeout' => 0.1, // in seconds
        'read_timeout' => '10', // in seconds
        'persistent_connections' => false

Using the InMemory storage:

$registry = new CollectorRegistry(new InMemory());

$counter = $registry->registerCounter('test', 'some_counter', 'it increases', ['type']);
$counter->incBy(3, ['blue']);

$renderer = new RenderTextFormat();
$result = $renderer->render($registry->getMetricFamilySamples());

Advanced Usage

Advanced Histogram Usage

On passing an empty array for the bucket parameter on instantiation, a set of default buckets will be used instead. Whilst this is a good base for a typical web application, there is named constructor to assist in the generation of exponential / geometric buckets.


Histogram::exponentialBuckets(0.05, 1.5, 10);

This will start your buckets with a value of 1.5, grow them by a factor of 1.5 per bucket across a set of 10 buckets.

Also look at the examples.



  • PHP ^7.3
  • PHP Redis extension
  • PHP APCu extension
  • Composer
  • Redis

Start a Redis instance:

docker-compose up Redis

Run the tests:

composer install

# when Redis is not listening on localhost:
# export REDIS_HOST=

Black box testing

Just start the nginx, fpm & Redis setup with docker-compose:

docker-compose up

Pick the adapter you want to test.

docker-compose run phpunit env ADAPTER=apc vendor/bin/phpunit tests/Test/
docker-compose run phpunit env ADAPTER=redis vendor/bin/phpunit tests/Test/