Overview

The Datadog Agent collects one main metric about Gunicorn: the number of worker processes running. It also sends one service check: whether or not Gunicorn is running.

Gunicorn itself can provide further metrics using DogStatsD, including:

• Total request rate
• Request rate by status code (2xx, 3xx, 4xx, 5xx)
• Request duration (average, median, max, 95th percentile, etc.)
• Log message rate by log level (critical, error, warning, exception)

Setup

Installation

The Datadog Agent’s Gunicorn check is included in the Datadog Agent package, so you don’t need to install anything else on your Gunicorn servers.

The Gunicorn check requires your Gunicorn app’s Python environment to have the setproctitle package; without it, the Datadog Agent reports that it cannot find a gunicorn master process (and hence, cannot find workers, either). Install the setproctitle package in your app’s Python environment if you want to collect the gunicorn.workers metric.

Configuration

Edit the gunicorn.d/conf.yaml file, in the conf.d/ folder at the root of your Agent’s configuration directory to start collecting your Gunicorn metrics and logs. See the sample gunicorn.yaml for all available configuration options.

Metric collection

Connect Gunicorn to DogStatsD

1. As of version 19.1, Gunicorn provides an option to send its metrics to a daemon that implements the StatsD protocol, such as DogStatsD. As with many Gunicorn options, you can either pass it to gunicorn on the CLI (--statsd-host) or set it in your app’s configuration file (statsd_host). To ensure that you collect all Gunicorn metrics, configure your app to send metrics to DogStatsD at "localhost:8125", and restart the app.

2. Add this configuration block to your gunicorn.d/conf.yaml file to start gathering Gunicorn metrics:

init_config:

instances:
    ## @param proc_name - string - required
    ## The name of the gunicorn process. For the following gunicorn server:
    ##
    ## gunicorn --name <WEB_APP_NAME> <WEB_APP_CONFIG>.ini
    ##
    ## the name is `<WEB_APP_NAME>`
  - proc_name: <YOUR_APP_NAME>

3. Restart the Agent to begin sending Gunicorn metrics to Datadog.

Log collection

Available for Agent versions >6.0

1. Collecting logs is disabled by default in the Datadog Agent, enable it in your datadog.yaml file:

   logs_enabled: true
   

2. Use the following command to configure the path of the access log file: --access-logfile <MY_FILE_PATH>

3. Use the following command to configure the path of the error log file: --error-logfile FILE, --log-file <MY_FILE_PATH>

4. Add this configuration block to your gunicorn.d/conf.yaml file to start collecting your Gunicorn logs:

   logs:
     - type: file
       path: /var/log/gunicorn/access.log
       service: "<MY_SERVICE>"
       source: gunicorn
   
     - type: file
       path: /var/log/gunicorn/error.log
       service: "<MY_SERVICE>"
       source: gunicorn
       log_processing_rules:
         - type: multi_line
           name: log_start_with_date
           pattern: \[\d{4}-\d{2}-\d{2}
   

   Change the service and path parameter values and configure them for your environment. See the sample gunicorn.yaml for all available configuration options.

5. Restart the Agent.

Validation

Run the Agent’s status subcommand and look for gunicorn under the Checks section.

If the status is not OK, see the Troubleshooting section.

Use netstat to verify that Gunicorn is sending its metrics, too:

$ sudo netstat -nup | grep "127.0.0.1:8125.*ESTABLISHED"
udp        0      0 127.0.0.1:38374         127.0.0.1:8125          ESTABLISHED 15500/gunicorn: mas

Data Collected

Metrics

Events

The Gunicorn check does not include any events.

Service Checks

gunicorn.is_running
Returns CRITICAL if the Agent is unable to find the Gunicorn master process. Returns OK otherwise.
Statuses: ok, critical

Troubleshooting

Agent cannot find Gunicorn process

  Checks
  ======

    gunicorn (5.12.1)
    -----------------
      - instance #0 [ERROR]: 'Found no master process with name: gunicorn: master [my_web_app]'
      - Collected 0 metrics, 0 events & 1 service check
      - Dependencies:
          - psutil: 4.4.1

Either Gunicorn really isn’t running, or your app’s Python environment doesn’t have the setproctitle package installed.

If setproctitle is not installed, Gunicorn appears in the process table like so:

$ ps -ef | grep gunicorn
ubuntu   18013 16695  2 20:23 pts/0    00:00:00 /usr/bin/python /usr/bin/gunicorn --config test-app-config.py gunicorn-test:app
ubuntu   18018 18013  0 20:23 pts/0    00:00:00 /usr/bin/python /usr/bin/gunicorn --config test-app-config.py gunicorn-test:app
ubuntu   18019 18013  0 20:23 pts/0    00:00:00 /usr/bin/python /usr/bin/gunicorn --config test-app-config.py gunicorn-test:app

If it is installed, gunicorn processes appear in the format the Datadog Agent expects:

$ ps -ef | grep gunicorn
ubuntu   18457 16695  5 20:26 pts/0    00:00:00 gunicorn: master [my_app]
ubuntu   18462 18457  0 20:26 pts/0    00:00:00 gunicorn: worker [my_app]
ubuntu   18463 18457  0 20:26 pts/0    00:00:00 gunicorn: worker [my_app]

Further Reading

• Monitor Gunicorn performance with Datadog