An agent that polls supported backend systems and submits the results to the NewRelic platform. Currently supported backend systems are:
- Alternative PHP Cache
- Apache HTTP Server
- CouchDB
- Elasticsearch
- HAProxy
- Memcached
- MongoDB
- Nginx
- pgBouncer
- PHP FPM
- PostgreSQL
- RabbitMQ
- Redis
- Riak
- uWSGI
The agent requires Python 2.6 or 2.7 and pip
for installation. Individual plugin backends may require additional libraries and are detailed below.
The configuration file uses YAML as its format. Most tickets for non-working installs are due to configuration file formatting errors. Please make sure you are properly formatting your configuration file prior to submitting a ticket. YAML is a whitespace dependent markup format. More information on writing proper YAML can be found at http://yaml.org.
- Install via
pip
*:
$ pip install newrelic-plugin-agent
- See
pip
installation instructions at http://www.pip-installer.org/en/latest/installing.html
- Copy the configuration file example from
/opt/newrelic-plugin-agent/newrelic-plugin-agent.cfg
to/etc/newrelic/newrelic-plugin-agent.cfg
and edit the configuration in that file. - Make a
/var/log/newrelic
directory and make sure it is writable by the user specified in the configuration file - Make a
/var/run/newrelic
directory and make sure it is writable by the user specified in the configuration file - Run the app:
$ newrelic-plugin-agent -c PATH-TO-CONF-FILE [-f]
Where -f
is to run it in the foreground instead of as a daemon.
Sample configuration and init.d scripts are installed to /opt/newrelic-plugin-agent
in addition to a PHP script required for APC monitoring.
To use the MongoDB the mongodb
library is required. For the pgBouncer or PostgreSQL plugin you must install the psycopg2
library. To easily do
this, make sure you have the latest version of pip
installed (http://www.pip-installer.org/). This should be done after installing the agent itself:
$ pip install newrelic-plugin-agent[mongodb]
or:
$ pip install newrelic-plugin-agent[pgbouncer]
or:
$ pip install newrelic-plugin-agent[postgresql]
If this does not work for you, make sure you are running a recent copy of pip
(>= 1.3).
Each plugin can support gathering data from a single or multiple targets. To support multiple targets for a plugin, you create a list of target stanzas:
plugin_name: - name: target_name host: localhost foo: bar - name: target_name host: localhost foo: bar
While you can use the multi-target format for a plugin's configuration stanza like:
plugin_name: - name: target_name host: localhost foo: bar
You can also use a single mapping like follows:
plugin_name: name: target_name host: localhost foo: bar
The fields for plugin configurations can vary due to a plugin's configuration requirements. The name value in each stanza is only required when using multiple targets in a plugin. If it is only a single target, the name will be taken from the server's hostname.
Copy the apc-nrp.php
script to a directory that can be served by your web server or php-fpm
application. Edit the newrelic-plugin-agent
configuration to point to the appropriate URL.
Enable the HTTPd server status page in the default virtual host. The following example configuration snippet for Apache HTTPd 2.2 demonstrates how to do this:
<Location /server-status> SetHandler server-status Order deny,allow Deny from all Allow from 127.0.0.1 </Location>
For HTTPd 2.4, it should look something like:
<Location /server-status> SetHandler server-status Require ip 127.0.0.1 </Location>
The agent requires the extended information to parse metrics. If you are not seeing any metrics on your graphs for Apache verify that you have enabled ExtendedStatus
, the default is off so you must enable it. In your global Apache HTTP configuration you need to enable exetended status using:
ExtendedStatus On
If you are monitoring Apache HTTPd via a HTTPS connection you can use the verify_ssl_cert
configuration value in the httpd configuration section to disable SSL certificate verification.
The memcached plugin can communicate either over UNIX domain sockets using the path configuration variable or TCP/IP using the host and port variables. Do not include both.
You need to install the pymongo driver, either by running pip install pymongo
or by following the "Installing Additional Requirements" above. Each database you wish to collect metrics for must be enumerated in the configuration.
There are two configuration stanza formats for MongoDB. You must use one or the other, they can not be mixed. For non-authenticated polling, you can simply enumate the databases you would like stats from as a list:
mongodb: name: hostname host: localhost port: 27017 #admin_username: foo #admin_password: bar #ssl: False #ssl_keyfile: /path/to/keyfile #ssl_certfile: /path/to/certfile #ssl_cert_reqs: 0 # Should be 0 for ssl.CERT_NONE, 1 for ssl.CERT_OPTIONAL, 2 for ssl.CERT_REQUIRED #ssl_ca_certs: /path/to/cacerts file databases: - database_name_1 - database_name_2
If your MongoDB server requires authentication, you must provide both admin credentials and database level credentials (or use delegated authentication as explained later) and the stanza is formatted as a nested array:
mongodb: name: hostname host: localhost port: 27017 #admin_username: foo #admin_password: bar #ssl: False #ssl_keyfile: /path/to/keyfile #ssl_certfile: /path/to/certfile #ssl_cert_reqs: 0 # Should be 0 for ssl.CERT_NONE, 1 for ssl.CERT_OPTIONAL, 2 for ssl.CERT_REQUIRED #ssl_ca_certs: /path/to/cacerts file databases: database_name_1: username: foo password: bar database_name_2: username: foo password: bar
If you would like to use delegated authentication you need to add auth_db
to desired databases.
Example below shows unlikely case of three different authentication scenarios:
- database_name_1 uses credentials defined in admin database (admin_username and admin_password).
- database_name_2 uses credentials defined in foobar database
- database_name_3 uses locally defined credentials.
mongodb: name: hostname host: localhost port: 27017 #admin_username: foo #admin_password: bar #ssl: False #ssl_keyfile: /path/to/keyfile #ssl_certfile: /path/to/certfile #ssl_cert_reqs: 0 # Should be 0 for ssl.CERT_NONE, 1 for ssl.CERT_OPTIONAL, 2 for ssl.CERT_REQUIRED #ssl_ca_certs: /path/to/cacerts file databases: database_name_1: auth_db: admin database_name_2: auth_db: foobar_db username: foo_2 password: bar_2 database_name_3: username: foo_3 password: bar_3
Enable the Nginx stub_status
setting on the default site in your configuration. The following example configuration snippet for Nginx demonstates how to do this:
location /nginx_stub_status { stub_status on; allow 127.0.0.1; deny all; }
If you are monitoring Nginx via a HTTPS connection you can use the verify_ssl_cert
configuration value in the httpd configuration section to disable SSL certificate verification.
The user specified must be a stats user.
By default, the specified user must be superuser to get PostgreSQL
directory listings. To skip those checks that require superuser
permissions, use the superuser: False
setting in the configuration
file.
Several of the checks take O(N) time where N is the number of relations
in the database. If you need to use this on a database with a very large
number of relations, you can skip these, using relation_stats: False
.
E.g.:
postgresql: host: localhost port: 5432 user: newrelic dbname: postgres password: newrelic superuser: False relation_stats: False
The user specified must have access to all virtual hosts you wish to monitor and should have either the Administrator tag or the Monitoring tag.
If you are monitoring RabbitMQ via a HTTPS connection you can use the verify_ssl_cert
configuration value in the httpd configuration section to disable SSL certificate verification.
For Redis daemons that are password protected, add the password configuration value, otherwise omit it. The Redis configuration section allows for multiple redis servers. The syntax to poll multiple servers is in the example below.
The Redis plugin can communicate either over UNIX domain sockets using the path configuration variable or TCP/IP using the host and port variables. Do not include both.
If you are monitoring Riak via a HTTPS connection you can use the verify_ssl_cert
configuration value in the httpd configuration section to disable SSL certificate verification.
The UWSGI plugin can communicate either over UNIX domain sockets using the path configuration variable or TCP/IP using the host and port variables. Do not include both.
Make sure you have enabled stats server in your uwsgi config.
%YAML 1.2 --- Application: license_key: REPLACE_WITH_REAL_KEY poll_interval: 60 #newrelic_api_timeout: 10 #proxy: http://localhost:8080 apache_httpd: - name: hostname1 scheme: http host: localhost port: 80 path: /server-status #verify_ssl_cert: true - name: hostname2 scheme: http host: localhost port: 80 path: /server-status #verify_ssl_cert: true couchdb: - name: localhost host: localhost port: 5984 #verify_ssl_cert: true #username: foo #password: bar - name: localhost host: localhost port: 5984 #verify_ssl_cert: true #username: foo #password: bar elasticsearch: name: clustername host: localhost port: 9200 haproxy: name: my-haproxy-server host: localhost port: 80 path: /haproxy?stats;csv scheme: http #verify_ssl_cert: true #username: foo #password: bar mongodb: name: hostname host: localhost port: 27017 admin_username: foo admin_password: bar databases: database_name_1: username: foo password: bar database_name_2: username: foo password: bar memcached: - name: localhost host: localhost port: 11211 path: /path/to/unix/socket - name: localhost host: localhost port: 11211 path: /path/to/unix/socket nginx: - name: hostname host: localhost port: 80 path: /nginx_stub_status #verify_ssl_cert: true - name: hostname host: localhost port: 80 path: /nginx_stub_status #verify_ssl_cert: true pgbouncer: - host: localhost port: 6000 user: stats php_apc: scheme: http host: localhost port: 80 path: /apc-nrp.php #username: foo #password: bar #verify_ssl_cert: t php_fpm: - name: fpm-pool scheme: https host: localhost port: 443 path: /fpm_status query: json postgresql: - host: localhost port: 5432 user: postgres dbname: postgres superuser: True rabbitmq: - name: rabbitmq@localhost host: localhost port: 15672 username: guest password: guest #verify_ssl_cert: true api_path: /api redis: - name: localhost host: localhost port: 6379 db_count: 16 password: foobar #path: /var/run/redis/redis.sock - name: localhost host: localhost port: 6380 db_count: 16 password: foobar #path: /var/run/redis/redis.sock riak: - name: localhost host: localhost port: 8098 #verify_ssl_cert: true Daemon: user: newrelic pidfile: /var/run/newrelic/newrelic-plugin-agent.pid Logging: formatters: verbose: format: '%(levelname) -10s %(asctime)s %(process)-6d %(processName) -15s %(threadName)-10s %(name) -25s %(funcName) -25s L%(lineno)-6d: %(message)s' handlers: file: class : logging.handlers.RotatingFileHandler formatter: verbose filename: /var/log/newrelic/newrelic-plugin-agent.log maxBytes: 10485760 backupCount: 3 loggers: newrelic-plugin-agent: level: INFO propagate: True handlers: [console, file] requests: level: ERROR propagate: True handlers: [console, file]
- If the installation does not install the
newrelic-plugin-agent
application in/usr/bin
then it is likely thatsetuptools
ordistribute
is not up to date. The following commands can be run to installdistribute
andpip
for installing the application:
$ curl http://python-distribute.org/distribute_setup.py | python $ curl https://raw.github.com/pypa/pip/master/contrib/get-pip.py | python
- If the application installs but doesn't seem to be submitting status, check the logfile which at
/tmp/newrelic-plugin-agent.log
if the default example logging configuration is used. - If the agent starts but dies shortly after ensure that
/var/log/newrelic
and/var/run/newrelic
are writable by the same user specified in the daemon section of the configuration file. - If the agent has died and won't restart, remove any files found in
/var/run/newrelic/
- If using the Apache HTTP plugin and your stats are blank, ensure the ExtendedStatus directive is on.