scrapy-prometheus 0.3.1

Exporting scrapy stats as prometheus metrics through pushgateway service

Scrapy stats collector that exports scrapy stats as prometheus metrics through pushgateway service.

Installation

Via pip:

pip install scrapy_prometheus

Usage

To start using, add scrapy_prometheus.PrometheusStatsCollector to STATS_CLASS setting:

STATS_CLASS = 'scrapy_prometheus.PrometheusStatsCollector'

This stats collector works exactly like the vanilla one (because it subclasses it), but also creates prometheus metrics and pushes them to pushgateway service on spider close signal.

It supports two metric types: Counter and Gauge. Stat metric type is determined by operation used on the stat: stats.inc_value will create a Counter metric, while other methods, stats.set_value, stats.max_value, stats.min_value, will create Gauge.

Stat value must be either int or float.

Note, that trying to perform action on a metric, that is not supposed to be used with this action (set_value on Counter or inc_value on Gauge) will produce scrapy_prometheus.InvalidMetricType error. To suppress it, set PROMETHEUS_SUPPRESS_TYPE_CHECK to True.

If you want to create other custom metrics, there are good news for you: scrapy_prometheus.PrometheusStatsCollector is also a prometheus_client.CollectorRegistry. Multi-inheritance rules. Just specify it as a registry, when creating a metric.

Available settings

# Prometheus pushgateway host
PROMETHEUS_PUSHGATEWAY = 'localhost:9091'  # default

# Metric name prefix
PROMETHEUS_METRIC_PREFIX = 'scrapy_prometheus'  # default

# Timeout for pushing metrics to pushgateway
PROMETHEUS_PUSH_TIMEOUT = 5  # default

# Method to use when pushing metrics
# Read https://github.com/prometheus/pushgateway#put-method
PROMETHEUS_PUSH_METHOD = 'POST'  # default

# Do not raise scrapy_prometheus.InvalidMetricType when stat is accessed as different type metric.
# For example, doing stats.inc_value('foo', 1) and then stats.set_value('foo', 2) will raise an error,
# Because metric of type Counter was already created for stat foo.
PROMETHEUS_SUPPRESS_TYPE_CHECK = False

# job label value, applied to all metrics.
PROMETHEUS_JOB = 'scrapy'  # default

# grouping label dict, applied to all metrics.
# by default it uses spider name.
PROMETHEUS_GROUPING_KEY = {'spider': <spider name>}

How metrics are created

Metric name is build from PROMETHEUS_METRIC_PREFIX and stat name, where all / are replaced with _.

For example:

• stat foo: 67 whill produce metric scrapy_prometheus_foo{instance="...",job="scrapy",spider="..."} 67

• stat foo/bar: 67 whill produce metric scrapy_prometheus_foo_bar{instance="...",job="scrapy",spider="..."} 67

• stat foo/bar/baz: 67 whill produce metric scrapy_prometheus_foo_bar_baz{instance="...",job="scrapy",spider="..."} 67