Skip to content

collectors RabbitMQCollector

Rob Smith edited this page Sep 16, 2014 · 14 revisions

RabbitMQCollector

Collects data from RabbitMQ through the admin interface

Notes

** With added support for breaking down queue metrics by vhost, we have attempted to keep results generated by existing configurations from changing. This means that the old behaviour of clobbering queue metrics when a single queue name exists in multiple vhosts still exists if the configuration is not updated. If no vhosts block is defined it will also keep the metric path as it was historically with no vhost name in it.

    old path => systems.myServer.rabbitmq.queues.myQueue.*
    new path => systems.myServer.rabbitmq.myVhost.queues.myQueue.*

** If a [vhosts] section exists but is empty, then no queues will be polled. ** To poll all vhosts and all queues, add the following. ** [vhosts] ** * = * **

Dependencies

  • pyrabbit

Options - Generic Options

Setting Default Description Type
byte_unit byte Default numeric output(s) str
cluster False If this node is part of a cluster, will collect metrics on the cluster health bool
enabled False Enable collecting these metrics bool
host localhost:55672 Hostname and port to collect from str
measure_collector_time False Collect the collector run time in ms bool
metrics_blacklist None Regex to match metrics to block. Mutually exclusive with metrics_whitelist NoneType
metrics_whitelist None Regex to match metrics to transmit. Mutually exclusive with metrics_blacklist NoneType
password guest Password str
queues Queues to publish. Leave empty to publish all.
queues_ignored , A list of queues or regexes for queue names not to report on. list
user guest Username str
vhosts A list of vhosts and queues for which we want to collect

Example Output

__EXAMPLESHERE__

This file was generated from the python source

Please edit the source to make changes

Clone this wiki locally