/rabbitmq_exporter

Prometheus exporter for RabbitMQ

Primary LanguageGoMIT LicenseMIT

RabbitMQ Exporter Build Status Coverage Status

Prometheus exporter for RabbitMQ metrics. Data is scraped by prometheus.

Please note this an unofficial plugin. There is also an official plugin from RabbitMQ.com. See comparison to official exporter

Installation

Binary release

You can download the latest release on the release page. Docker images are push to docker hub

docker: rabbitmq container with network sharing

The rabbitmq_exporter is sharing the network interface with the rabbitmq container -> it is possible to use localhost and default user/password (guest).

  1. Start rabbitMQ

     docker run -d -e RABBITMQ_NODENAME=my-rabbit --name my-rabbit -p 9419:9419 rabbitmq:3-management
    
  2. Start rabbitmq_exporter in container.

     docker run -d --net=container:my-rabbit kbudde/rabbitmq-exporter
    

Now your metrics are exposed through http://host:9419/metrics. The management plugin does not need to be exposed.

Configuration

Rabbitmq_exporter can be configured using json config file or environment variables for configuration.

Config file

Rabbitmq_exporter expects config file in "conf/rabbitmq.conf". If you are running the exporter in a container (docker/kubernetes) the config must be in "/conf/rabbitmq.conf" The name of the file can be overriden with flag:

./rabbitmq_exporter -config-file config.example.json

You can find an example here. Note: If you are using a config file, you must provide all values as there is no default value.

Settings

Environment variable default description
RABBIT_URL http://127.0.0.1:15672 url to rabbitMQ management plugin (must start with http(s)://)
RABBIT_USER guest username for rabbitMQ management plugin. User needs monitoring tag!
RABBIT_PASSWORD guest password for rabbitMQ management plugin
RABBIT_USER_FILE location of file with username (useful for docker secrets)
RABBIT_PASSWORD_FILE location of file with password (useful for docker secrets)
PUBLISH_PORT 9419 Listening port for the exporter
PUBLISH_ADDR "" Listening host/IP for the exporter
OUTPUT_FORMAT TTY Log ouput format. TTY and JSON are suported
LOG_LEVEL info log level. possible values: "debug", "info", "warning", "error", "fatal", or "panic"
CAFILE ca.pem path to root certificate for access management plugin. Just needed if self signed certificate is used. Will be ignored if the file does not exist
CERTFILE client-cert.pem path to client certificate used to verify the exporter's authenticity. Will be ignored if the file does not exist
KEYFILE client-key.pem path to private key used with certificate to verify the exporter's authenticity. Will be ignored if the file does not exist
SKIPVERIFY false true/0 will ignore certificate errors of the management plugin
SKIP_VHOST ^$ regex, matching vhost names are not exported. First performs INCLUDE_VHOST, then SKIP_VHOST. Applies to queues and exchanges
INCLUDE_VHOST .* regex vhost filter. Only matching vhosts are exported. Applies to queues and exchanges
INCLUDE_QUEUES .* regex queue filter. Just matching names are exported
SKIP_QUEUES ^$ regex, matching queue names are not exported (useful for short-lived rpc queues). First performed INCLUDE, after SKIP
INCLUDE_EXCHANGES .* regex exchange filter. (Only exchanges in matching vhosts are exported)
SKIP_EXCHANGES ^$ regex, matching exchanges names are not exported. First performed INCLUDE, after SKIP
RABBIT_CAPABILITIES bert,no_sort comma-separated list of extended scraping capabilities supported by the target RabbitMQ server
RABBIT_EXPORTERS exchange,node,queue List of enabled modules. Possible modules: connections,shovel,federation,exchange,node,queue,memory
RABBIT_TIMEOUT 30 timeout in seconds for retrieving data from management plugin.
MAX_QUEUES 0 max number of queues before we drop metrics (disabled if set to 0)
EXCLUDE_METRICS Metric names to exclude from export. comma-seperated. e.g. "recv_oct, recv_cnt". See exporter_*.go for names

Example and recommended settings:

SKIP_QUEUES="RPC_.*" MAX_QUEUES=5000 ./rabbitmq_exporter

Extended RabbitMQ capabilities

Newer version of RabbitMQ can provide some features that reduce overhead imposed by scraping the data needed by this exporter. The following capabilities are currently supported in RABBIT_CAPABILITIES env var:

  • no_sort: By default RabbitMQ management plugin sorts results using the default sort order of vhost/name. This sorting overhead can be avoided by passing empty sort argument (?sort=) to RabbitMQ starting from version 3.6.8. This option can be safely enabled on earlier 3.6.X versions, but it'll not give any performance improvements. And it's incompatible with 3.4.X and 3.5.X.
  • bert: Since 3.6.9 (see rabbitmq/rabbitmq-management#367) RabbitMQ supports BERT encoding as a JSON alternative. Given that BERT encoding is implemented in C inside the Erlang VM, it's way more effective than pure-Erlang JSON encoding. So this greatly reduces monitoring overhead when we have a lot of objects in RabbitMQ.

Note for users of rabbmitmq < 3.6

no_sort and bert are enabled by default. You must overwrite the default settings with:

RABBIT_CAPABILITIES=nobert ./rabbitmq_exporter

Metrics

All metrics (except golang/prometheus metrics) are prefixed with "rabbitmq_".

Global

Always exported.

metric description
up Was the last scrape of rabbitmq successful.
module_up Was the last scrape of rabbitmq module successful. labels: module
module_scrape_duration_seconds Duration of the last scrape of rabbitmq module. labels: module
exporter_build_info A metric with a constant '1' value labeled by version, revision, branch and build date on which the rabbitmq_exporter was built.

Overview

Always exported. Labels: cluster

metric description
channels Number of channels
connections Number of connections
consumers Number of message consumers
queues Number of queues in use
exchanges Number of exchanges in use
queue_messages_global Number ready and unacknowledged messages in cluster.
queue_messages_ready_global Number of messages ready to be delivered to clients.
queue_messages_unacknowledged_global Number of messages delivered to clients but not yet acknowledged.
version_info A metric with a constant '1' value labeled by rabbitmq version, erlang version, node, cluster.

Queues

Labels: cluster, vhost, queue, durable, policy, self

Queues - Gauge

metric description
queue_messages_ready Number of messages ready to be delivered to clients.
queue_messages_unacknowledged Number of messages delivered to clients but not yet acknowledged.
queue_messages Sum of ready and unacknowledged messages (queue depth).
queue_messages_ack_total Number of messages delivered in acknowledgement mode in response to basic.get.
queue_messages_ready_ram Number of messages from messages_ready which are resident in ram.
queue_messages_unacknowledged_ram Number of messages from messages_unacknowledged which are resident in ram.
queue_messages_ram Total number of messages which are resident in ram.
queue_messages_persistent Total number of persistent messages in the queue (will always be 0 for transient queues).
queue_message_bytes Sum of the size of all message bodies in the queue. This does not include the message properties (including headers) or any overhead.
queue_message_bytes_ready Like message_bytes but counting only those messages ready to be delivered to clients.
queue_message_bytes_unacknowledged Like message_bytes but counting only those messages delivered to clients but not yet acknowledged.
queue_message_bytes_ram Like message_bytes but counting only those messages which are in RAM.
queue_message_bytes_persistent Like message_bytes but counting only those messages which are persistent.
queue_consumers Number of consumers.
queue_consumer_utilisation Fraction of the time (between 0.0 and 1.0) that the queue is able to immediately deliver messages to consumers. This can be less than 1.0 if consumers are limited by network congestion or prefetch count.
queue_memory Bytes of memory consumed by the Erlang process associated with the queue, including stack, heap and internal structures.
queue_head_message_timestamp The timestamp property of the first message in the queue, if present. Timestamps of messages only appear when they are in the paged-in state.
queue_max_length_bytes Total body size for ready messages a queue can contain before it starts to drop them from its head.
queue_max_length How many (ready) messages a queue can contain before it starts to drop them from its head.
queue_idle_since_seconds starttime where the queue switched to idle state; seconds since epoch (1970); only set if queue state is idle
queue_reductions_total Number of reductions which take place on this process.
queue_state A metric with a value of constant '1' if the queue is in a certain state. Labels: vhost, queue, state (running, idle, flow,..)
queue_slave_nodes_len Number of slave nodes attached to the queue
queue_synchronised_slave_nodes_len Number of slave nodes in sync to the queue

Queues - Counter

metric description
queue_disk_reads_total Total number of times messages have been read from disk by this queue since it started.
queue_disk_writes_total Total number of times messages have been written to disk by this queue since it started.
queue_messages_published_total Count of messages published.
queue_messages_confirmed_total Count of messages confirmed.
queue_messages_delivered_total Count of messages delivered in acknowledgement mode to consumers.
queue_messages_delivered_noack_total Count of messages delivered in no-acknowledgement mode to consumers.
queue_messages_get_total Count of messages delivered in acknowledgement mode in response to basic.get.
queue_messages_get_noack_total Count of messages delivered in no-acknowledgement mode in response to basic.get.
queue_messages_redelivered_total Count of subset of messages in deliver_get which had the redelivered flag set.
queue_messages_returned_total Count of messages returned to publisher as unroutable.

Exchanges - Counter

Labels: cluster, vhost, exchange

metric description
exchange_messages_published_in_total Count of messages published in to an exchange, i.e. not taking account of routing.
exchange_messages_published_out_total Count of messages published out of an exchange, i.e. taking account of routing.

Node - Counter

Labels: cluster, node, self

metric description
uptime Uptime in milliseconds
running number of running nodes
node_mem_used Memory used in bytes
node_mem_limit Point at which the memory alarm will go off
node_mem_alarm Whether the memory alarm has gone off
node_disk_free Disk free space in bytes.
node_disk_free_alarm Whether the disk alarm has gone off.
node_disk_free_limit Point at which the disk alarm will go off.
fd_used Used File descriptors
fd_available File descriptors available
sockets_used File descriptors used as sockets.
sockets_available File descriptors available for use as sockets
partitions Current Number of network partitions. 0 is ok. If the cluster is splitted the value is at least 2

Connections - Gauge

disabled by default. Depending on the environment and change rate it can create a high number of dead metrics. Otherwise it could be usefull and can be enabled.

Labels: cluster, vhost, node, peer_host, user, self

Please note: The data is aggregated by label values as it is possible that there are multiple connections for a certain combination of labels.

metric description
connection_channels number of channels in use
connection_received_bytes received bytes
connection_received_packets received packets
connection_send_bytes send bytes
connection_send_packets send packets
connection_send_pending Send queue size

Labels: vhost, node, peer_host, user, state (running, flow,..), self

metric description
connection_status Number of connections in a certain state aggregated per label combination. Metric will disappear if there are no connections in a state.

Shovel

disabled by default Labels: cluster, vhost, shovel, type, self, state

metric description
shovel_state A metric with a value of constant '1' for each shovel in a certain state

Memory

disabled by default Labels: cluster, node, self

metric description
memory_allocated_unused_bytes Memory preallocated by the runtime (VM allocators) but not yet used
memory_atom_bytes Memory used by atoms. Should be fairly constant
memory_binary_bytes Memory used by shared binary data in the runtime. Most of this memory is message bodies and metadata.)
memory_code_bytes Memory used by code (bytecode, module metadata). This section is usually fairly constant and relatively small (unless the node is entirely blank and stores no data).
memory_connection_channels_bytes Memory used by client connections - channels
memory_connection_other_bytes Memory used by client connection - other
memory_connection_readers Memory used by processes responsible for connection parser and most of connection state. Most of their memory attributes to TCP buffers
memory_connection_writers_bytes Memory used by processes responsible for serialization of outgoing protocol frames and writing to client connections socktes
memory_mgmt_db_bytes Management DB ETS tables + processes
memory_mnesia_bytes Internal database (Mnesia) tables keep an in-memory copy of all its data (even on disc nodes)
memory_msg_index_bytes Message index ETS + processes
memory_other_ets_bytes Other in-memory tables besides those belonging to the stats database and internal database tables
memory_other_proc_bytes Memory used by all other processes that RabbitMQ cannot categorise
memory_other_system_bytes Memory used by all other system that RabbitMQ cannot categorise
memory_plugins_bytes Memory used by plugins (apart from the Erlang client which is counted under Connections, and the management database which is counted separately).
memory_queue_procs_bytes Memory used by class queue masters, queue indices, queue state
memory_queue_slave_procs_bytes Memory used by class queue mirrors, queue indices, queue state
memory_reserved_unallocated_bytes Memory preallocated/reserved by the kernel but not the runtime
memory_total_allocated_bytes Node-local total memory - allocated
memory_total_rss_bytes Node-local total memory - rss
memory_total_erlang_bytes Node-local total memory - erlang

Comparison to official exporter

official exporter:

  • has runtime/erlang metrics
  • aggregated or per-object metrics
  • missing filter

This exporter:

  • works also with older versions of rabbitmq
  • has more configuration options/ filtering of objects
  • (bad) depends on data from management interface which can be slow/delayed

probalby best solution is to use both exporters: comment from shamil

common errors: Error while retrieving data from rabbitHost / code 500

msg: Error while retrieving data from rabbitHost statusCode: 500

This exporter expects capabilities from rabbitmq 3.6.8 or newer by default. If you are running older than 3.6.8 you have to set RABBIT_CAPABILITIES=compat.