Nginx

Agent Check

Overview

The Datadog Agent can collect many metrics from NGINX instances, including (but not limited to)::

• Total requests
• Connections (e.g. accepted, handled, active)

For users of NGINX Plus, the commercial version of NGINX, the Agent can collect the significantly more metrics that NGINX Plus provides, like:

• Errors (e.g. 4xx codes, 5xx codes)
• Upstream servers (e.g. active connections, 5xx codes, health checks, etc.)
• Caches (e.g. size, hits, misses, etc.)
• SSL (e.g. handshakes, failed handshakes, etc.)

Setup

Installation

The NGINX check pulls metrics from a local NGINX status endpoint, so your nginx binaries need to have been compiled with one of two NGINX status modules:

• stub status module - for open source NGINX
• http status module - only for NGINX Plus

NGINX Open Source

If you use open source NGINX, your instances may lack the stub status module. Verify that your nginx binary includes the module before proceeding to Configuration:

$ nginx -V 2>&1| grep -o http_stub_status_module
http_stub_status_module

If the command output does not include http_stub_status_module, you must install an NGINX package that includes the module. You can compile your own NGINX-enabling the module as you compile it-but most modern Linux distributions provide alternative NGINX packages with various combinations of extra modules built in. Check your operating system’s NGINX packages to find one that includes the stub status module.

NGINX Plus

NGINX Plus packages prior to release 13 include the http status module. For NGINX Plus release 13 and above, the status module is deprecated and you must use the new Plus API instead. See the announcement for more information.

Prepare NGINX

On each NGINX server, create a status.conf file in the directory that contains your other NGINX configuration files (e.g. /etc/nginx/conf.d/).

server {
  listen 81;
  server_name localhost;

  access_log off;
  allow 127.0.0.1;
  deny all;

  location /nginx_status {
    # Choose your status module

    # freely available with open source NGINX
    stub_status;

    # for open source NGINX < version 1.7.5
    # stub_status on;

    # available only with NGINX Plus
    # status;

    # ensures the version information can be retrieved
    server_tokens on;
  }
}

NGINX Plus

NGINX Plus users can also utilize stub_status, but since that module provides fewer metrics, Datadog recommends using status.

For NGINX Plus releases 15+, the status module is deprecated. Use the http_api_module instead. For example, enable the /api endpoint in your main NGINX configuration file (/etc/nginx/conf.d/default.conf):

server {
  listen 8080;
  location /api {
    api write=on;
  }
}

To get more detailed metrics with NGINX Plus (such as 2xx / 3xx / 4xx / 5xx response counts per second), set a status_zone on the servers you want to monitor. For example:

server {
  listen 80;
  status_zone <ZONE_NAME>;
  ...
}

Reload NGINX to enable the status or API endpoint. There’s no need for a full restart.

sudo nginx -t && sudo nginx -s reload

Configuration

http {
	#recommended log format
	log_format nginx '\$remote_addr - \$remote_user [\$time_local] '
                  '"\$request" \$status \$body_bytes_sent \$request_time '
                  '"\$http_referer" "\$http_user_agent"';

	access_log /var/log/nginx/access.log;
}

Validation

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

Data Collected

Metrics

Not all metrics shown are available to users of open source NGINX. Compare the module reference for stub status (open source NGINX) and http status (NGINX Plus) to understand which metrics are provided by each module.

A few open-source NGINX metrics are named differently in NGINX Plus; they refer to the exact same metric, though:

These metrics don’t refer exactly to the same metric, but they are somewhat related:

Finally, these metrics have no good equivalent:

Events

The NGINX check does not include any events.

Service Checks

nginx.can_connect:
Returns CRITICAL if the Agent cannot connect to NGINX to collect metrics, otherwise returns OK.

Troubleshooting

• Why do my logs not have the expected timestamp?

Need help? Contact Datadog support.

Further Reading

Additional helpful documentation, links, and articles:

• How to monitor NGINX
• How to collect NGINX metrics
• How to monitor NGINX with Datadog