graphite-nginx-module

An nginx module for collecting location stats into Graphite.

This module is not distributed with the Nginx source. See the installation instructions.

Features

• Aggregation of location, server or http metrics
• Calculation of percentiles
• Sending data to Graphite over UDP or TCP in non-blocking way
• Sending custom metrics from lua

Version

This document describes graphite-nginx-module v2.3.0 released on 21 August 2018.

Synopsis

http {
    graphite_config prefix=playground server=127.0.0.1;
    server {
        location /foo/ {
            graphite_data nginx.foo;
        }
    }
}

Description

This module use shared memory segment to collect aggregated stats from all workers and send calculated values for last minute to Graphite every 60s (default) over UDP or TCP in non-blocking way. Stats aggegation made on the fly in fixed size buffer allocated on server start and does't affect server performance.

This module is in active use on Mail.Ru Sites (one of largest web-services in Russia) for about a year and considered stable and well-tested.

To collect metrics from nginx core modules (ssl, gzip, upstream) little patch must be applied on nginx source tree. See the installation instructions. You can build this module as a dynamic one, but then you won't be able to collect metrics from nginx core modules (ssl, gzip, upstream) and lua functions.

Directives

graphite_config

syntax: graphite_config key1=<value1> key2=<value2> ... keyN=<valueN>

context: http

Specify global settings for a whole server instance.

Param	Required	Default	Description
prefix			path prefix for all graphs
host		gethostname()	host name for all graphs
server	Yes		carbon-cache server IP address
protocol		udp	carbon-cache server protocol (udp or tcp)
port		2003	carbon-cache server port
frequency		60	how often send values to Graphite (seconds)
intervals		1m	aggregation intervals, time interval list, vertical bar separator (m - minutes)
params		*	limit metrics list to track, vertical bar separator
shared		2m	shared memory size, increase in case of too small shared memory error
buffer		64k	network buffer size, increase in case of too small buffer size error
package		1400	maximum UDP packet size
template			template for graph name (default is $prefix.$host.$split.$param_$interval)
error_log			path suffix for error logs graphs (*)

(*): works only when nginx_error_log_limiting*.patch is applied to the nginx source code

Example (standard):

http {
    graphite_config prefix=playground server=127.0.0.1;
}

Example (custom):

http {
    graphite_config prefix=playground server=127.0.0.1 intervals=1m|5m|15m params=rps|request_time|upstream_time template=$prefix.$host.$split.$param_$interval;
}

Example (error_log):

http {
    graphite_config prefix=playground server=127.0.0.1 error_log=log;
}

graphite_default_data

syntax: graphite_default_data <path prefix> [params=<params>] [if=<condition>]

context: http, server

Create measurement point in all nested locations. You can use "$location" or "$server" variables which represent the name of the current location and the name of current server with all non-alphanumeric characters replaced with "_." Leading and trailing "_" are deleted.

Example:

   graphite_default_data nginx.$location;

   location /foo/ {
   }

   location /bar/ {
   }

Data for /foo/ will be sent to nginx.foo, data for /bar/ - to nginx.bar. The <params> parameter (1.3.0) specifies list of params to be collected for all nested locations. To add all default params, use *. The <if> parameter (1.1.0) enables conditional logging. A request will not be logged if the condition evaluates to "0" or an empty string.

Example(with $server):

    graphite_default_data nginx.$server.$location

    server {
        server_name foo_host;

        location /foo/ {
        }
    }

    server {
        server_name bar_host;

        location /bar/ {
        }
    }

Data for /foo/ will be sent to nginx.foo_host.foo, data for /bar/ - to nginx.bar_host.bar.

graphite_data

syntax: graphite_data <path prefix> [params=<params>] [if=<condition>]

context: http, server, location, if

Create measurement point in specific location.

Example:

    location /foo/ {
        graphite_data nginx.foo;
    }

The <params> parameter (1.3.0) specifies list of params to be collected for this location. To add all default params, use *. The <if> parameter (1.1.0) enables conditional logging. A request will not be logged if the condition evaluates to "0" or an empty string.

Example:

    map $scheme $is_http { http 1; }
    map $scheme $is_https { https 1; }

    ...

    location /bar/ {
        graphite_data nginx.all.bar;
        graphite_data nginx.http.bar if=$is_http;
        graphite_data nginx.https.bar if=$is_https;
        graphite_data nginx.arg params=rps|request_time;
        graphite_data nginx.ext params=*|rps|request_time;
    }

graphite_param

syntax: graphite_param name=<path> interval=<time value> aggregate=<func>

context: location

Param	Required	Description
name	Yes	path prefix for all graphs
interval	Yes*	aggregation interval, time intrval value format (m - minutes)
aggregate	Yes*	aggregation function on values
percentile	Yes*	percentile level

aggregate functions

func	Description
sum	sum of values per interval
persec	sum of values per second (sum divided on seconds in interval)
avg	average value on interval
gauge	gauge value

Example: see below.

Nginx API for Lua

syntax: ngx.graphite(<name>,<value>[,<config>])

Write stat value into aggregator function. Floating point numbers accepted in value.

Available after applying patch to lua-nginx-module. See the installation instructions.

ngx.graphite(name, value, config)

Example:

location /foo/ {
    graphite_param name=lua.foo_sum aggregate=sum interval=1m;
    graphite_param name=lua.foo_rps aggregate=persec interval=1m;
    graphite_param name=lua.foo_avg aggregate=avg interval=1m;
    graphite_param name=lua.foo_gauge aggregate=gauge;

    content_by_lua '
        ngx.graphite("lua.foo_sum", 0.01)
        ngx.graphite("lua.foo_rps", 1)
        ngx.graphite("lua.foo_avg", ngx.var.request_uri:len())
        ngx.graphite("lua.foo_gauge", 10)
        ngx.graphite("lua.foo_gauge", -2)
        ngx.graphite("lua.auto_rps", 1, "aggregate=persec interval=1m percentile=50|90|99")
        ngx.say("hello")
    ';
}

You must either specify the graphite_param command or pass the config argument. If you choose the second option, the data for this graph will not be sent until the first call to ngx.graphite.

Warning: If you do not declare graph using graphite_param command then memory for the graph will be allocated dynamically in module's shared memory. If module's shared memory is exhausted while nginx is running, no new graphs will be created and an error message will be logged.

syntax: ngx.graphite.get(<name>)

Get value of the gauge param with specified name.

syntax: ngx.graphite.set(<name>,<value>)

Set value to the gauge param with specified name.

Params

Param	Units	Func	Description
request_time	ms	avg	total time spent on serving request
bytes_sent	bytes	avg	http response length
body_bytes_sent	bytes	avg	http response body length
request_length	bytes	avg	http request length
ssl_handshake_time	ms	avg	time spent on ssl handsake
ssl_cache_usage	%	last	how much SSL cache used
content_time	ms	avg	time spent generating content inside nginx
gzip_time	ms	avg	time spent gzipping content ob-the-fly
upstream_time	ms	avg	time spent tailking with upstream
upstream_connect_time	ms	avg	time spent on upstream connect (nginx >= 1.9.1)
upstream_header_time	ms	avg	time spent on upstream header (nginx >= 1.9.1)
rps	rps	sum	total requests number per second
keepalive_rps	rps	sum	requests number sent over previously opened keepalive connection
response_2xx_rps	rps	sum	total responses number with 2xx code
response_3xx_rps	rps	sum	total responses number with 3xx code
response_4xx_rps	rps	sum	total responses number with 4xx code
response_5xx_rps	rps	sum	total responses number with 5xx code
response_[0-9]{3}_rps	rps	sum	total responses number with given code
upstream_cache_(miss|bypass|expired|stale|updating|revalidated|hit)_rps	rps	sum	totar responses with a given upstream cache status
lua_time	ms	avg	time spent on lua code

Percentiles

To calculate percentile value for any parameter, set percentile level via /. E.g. request_time/50|request_time/90|request_time/99.

Installation

Requirements

• nginx: 1.2.0 - 1.14.x
• lua-nginx-module: 0.8.6 - 0.10.13 (optional)

Build nginx with graphite module

wget 'http://nginx.org/download/nginx-1.9.2.tar.gz'
tar -xzf nginx-1.9.2.tar.gz
cd nginx-1.9.2/

# patch to collect ssl_cache_usage, ssl_handshake_time content_time, gzip_time, upstream_time, upstream_connect_time, upstream_header_time graphs (optional)
patch -p1 < /path/to/graphite-nginx-module/graphite_module_v1_7_7.patch

./configure --add-module=/path/to/graphite-nginx-module

make
make install

Build nginx with graphite dynamic module

wget 'http://nginx.org/download/nginx-1.9.2.tar.gz'
tar -xzf nginx-1.9.2.tar.gz
cd nginx-1.9.2/

./configure --add-dynamic-module=/path/to/graphite-nginx-module

make
make install

Build nginx with lua and graphite modules

wget 'https://github.com/chaoslawful/lua-nginx-module/archive/v0.9.16.tar.gz'
tar -xzf v0.9.16.tar.gz
cd lua-nginx-module-0.9.16/
# patch to add api for sending metrics from lua code (optional)
patch -p1 < /path/to/graphite-nginx-module/lua_module_v0_9_11.patch
cd ..

wget 'http://nginx.org/download/nginx-1.9.2.tar.gz'
tar -xzf nginx-1.9.2.tar.gz
cd nginx-1.9.2/

# patch to collect ssl_cache_usage, ssl_handshake_time content_time, gzip_time, upstream_time, upstream_connect_time, upstream_header_time graphs (optional)
patch -p1 < /path/to/graphite-nginx-module/graphite_module_v1_7_7.patch

./configure \
    --add-module=/path/to/ngx_devel_kit \
    --add-module=/path/to/lua-nginx-module \
    --add-module=/path/to/graphite-nginx-module

make
make install

Instructions on installing lua-nginx-module can be found in documentation on lua-nginx-module.

License

Copyright (c) 2013-2018, Mail.Ru Ltd.

This module is licensed under the terms of the BSD license.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.