Clever/docker-collectd-plugin

Name: docker-collectd-plugin

Owner: Clever

Description: CollectD plugin for Docker container statistics

Forked from: signalfx/docker-collectd-plugin

Created: 2017-05-04 22:53:00.0

Updated: 2017-05-04 22:53:02.0

Pushed: 2017-05-05 00:44:32.0

Homepage:

Size: 94

Language: Python

GitHub Committers

User	Most Recent Commit	# Commits

Other Committers

User	Email	Most Recent Commit	# Commits

README

docker-collectd-plugin

A Docker plugin for collectd using docker-py and collectd's Python plugin.

This uses the new stats API introduced by #9984 in Docker 1.5.

The following container stats are gathered for each container, with examples of some of the metric names reported:

• Network bandwidth:
• network.usage.rx_bytes
• network.usage.tx_bytes
• …
• Memory usage:
• memory.usage.total
• memory.usage.limit
• …
• CPU usage:
• cpu.usage.total
• cpu.usage.system
• …
• Block IO:
• blockio.io_service_bytes_recursive-<major>-<minor>.read
• blockio.io_service_bytes_recursive-<major>-<minor>.write
• …

All metrics are reported with a plugin:docker dimension, and the name of the container is used for the plugin_instance dimension.

Install

1. Checkout this repository somewhere on your system accessible by collectd; for example as /usr/share/collectd/docker-collectd-plugin.
2. Install the Python requirements with sudo pip install -r requirements.txt.
3. Configure the plugin (see below).
4. Restart collectd.

Configuration

Add the following to your collectd config:

sDB "/usr/share/collectd/docker-collectd-plugin/dockerplugin.db"
Plugin python

gin python>
dulePath "/usr/share/collectd/docker-collectd-plugin"
port "dockerplugin"

odule dockerplugin>
BaseURL "unix://var/run/docker.sock"
Timeout 3
Module>
ugin>

Extracting additional dimensions

The plugin has the ability to extract additional information about or from the monitored container and report them as extra dimensions on your metrics. The general syntax is via the Dimension directive in the Module section:

ule dockerplugin>
.
mension "<name>" "<spec>"
dule>

Where <name> is the name of the dimension you want to add, and <spec> describes how the value for this dimension should be extracted by the plugin. You can extract:

• manually specified dimension values, with raw:foobar;
• container environment variable values, with env:ENV_VAR_NAME;
• any container detail value from Docker, with inspect:Path.In.Json.

Here are some examples:

nsion "mesos_task_id" "env:MESOS_TASK_ID"
nsion "image" "inspect:Config.Image"
nsion "foo" "raw:bar"

For inspect, the parameter is expected to be a JSONPath matcher giving the path to the value you're interested in within the JSON output of docker inspect (or /containers/<container>/json via the remote API). See https://github.com/kennknowles/python-jsonpath-rw for more details about the JSONPath syntax.

Important note about additional dimensions

The additional dimensions extracted by this docker-collectd-plugin are packed into the plugin_instance dimension using the following syntax:

in_instance:value[dim1=value1,dim2=value2,...]

Because CollectD limits the size of the reported fields to 64 total characters, trying to pack too many additional dimensions will result in the whole constructed string from being truncated, making parsing those additional dimensions correctly on the target metrics system impossible. You should make sure that the total size of your plugin_instance value, including all additional dimensions names and values, does not exceed 64 characters.

How to send only basic metrics

A lot of the metrics reported by this plugin are advanced statistics you may not necessarily need. CollectD makes it easy to filter metrics you don't want to record or send. The first step is to define a PostCache chain with a rule to pass the Docker CollectD plugin's metrics through a custom filtering chain.

Plugin match_regex

in "PostCache">
ule>
<Match "regex">
  Plugin "^docker$"
</Match>
<Target "jump">
  Chain "FilterOutDetailedDockerStats"
</Target>
Rule>

rget "write"
ain>

Note: make sure you have exactly one <Chain "PostCache"> section; CollectD will consider the first one it sees, and ignore any other. If you already have a section, simply add the <Rule> block that jumps to the FilterOutDetailedDockerStats sub-chain to your existing PostCache chain.

You can then use the following filtering chain to only allow the “basic” CPU, memory, network and block I/O metrics to go through, and drop everything else from the plugin:

in "FilterOutDetailedDockerStats">
ule "CpuUsage">
<Match "regex">
  Type "^cpu.usage$"
</Match>
Target "return"
Rule>
ule "CpuPercent">
<Match "regex">
  Type "^cpu.percent$"
</Match>
Target "return"
Rule>
ule "MemoryUsage">
<Match "regex">
  Type "^memory.usage$"
</Match>
Target "return"
Rule>
ule "MemoryPercent">
<Match "regex">
  Type "^memory.percent$"
</Match>
Target "return"
Rule>
ule "NetworkUsage">
<Match "regex">
  Type "^network.usage$"
</Match>
Target "return"
Rule>
ule "BlockIO">
<Match "regex">
  Type "^blkio$"
  TypeInstance "^io_service_bytes_recursive*"
</Match>
Target "return"
Rule>

rget "stop"
ain>

For convenience, you'll find those configuration blocks in the example configuration file 10-docker.conf.

Requirements

• Docker 1.5+
• docker-py
• jsonpath_rw
• python-dateutil