"A collected plugin, written in python, to collect statistics from RabbitMQ."
- Free software: Apache license
- Documentation: https://collectd-rabbitmq.readthedocs.org.
- For the older single file version see https://github.com/NYTimes/collectd-rabbitmq/tree/0.1.1
- Support queue, exchange, and node stats,
This plugin supports a small amount of configuration options:
- Username: The rabbitmq user. Defaults to guest
- Password: The rabbitmq user password. Defaults to guest
- Realm: The http realm for authentication. Defaults to RabbitMQ Management
- Scheme: The protocol that the rabbitmq management API is running on. Defaults to http
- Host: The hostname that the rabbitmq server running on. Defaults to localhost
- Port: The port that the rabbitmq server is listening on. Defaults to 15672
- Ignore: The queue to ignore, matching by Regex. See example.
- See this example for further details.
For each node the following statistics are gathered:
- disk_free_limit
- fd_total
- fd_used
- mem_limit
- mem_used
- proc_total
- proc_used
- processors
- run_queue
- sockets_total
- sockets_used
For each queue in each vhost the following statistics are gathered: _NOTE_: The / vhost name is sent as default
- message_stats
- deliver_get
- deliver_get_details
- rate
- get
- get_details
- rate
- publish
- publish_details
- rate
- redeliver
- redeliver_details
- rate
- messages
- messages_details
- rate
- messages_ready
- messages_ready_details
- rate
- messages_unacknowledged
- messages_unacknowledged_details * rate
- memory
- consumers
For each exchange in each vhost the following statistics are gathered: _NOTE_: The / vhost name is sent as default
- disk_free
- disk_free_limit
- fd_total
- fd_used
- mem_limit
- mem_used
- proc_total
- proc_used
- processors
- run_queue
- sockets_total
- sockets_used
This package was created with Cookiecutter and the cookiecutter-pypackage project template.