Skip to content

Latest commit

 

History

History
247 lines (188 loc) · 6.33 KB

README.md

File metadata and controls

247 lines (188 loc) · 6.33 KB
Raw

collectd role

This role takes care of adding collectd to any given server.

It can be used to:

  1. Collect metrics about a machines, and optionally forward them to an aggregation machine.

  2. Create an aggregation machine, which will receive the metrics from a given set of machine within the same datacenter. In this case too, metrics can optionally be forwarded to another service.

Configuration

For any machines

Simply add the collector role to your playbook.

By default, we will save all metrics to RRD files under /var/lib/collectd/rrd/${hostname}.

For forwarding machines

Inventory

Normally, we put all aggregation on a single machine. However, it is possible to decouple them, and provision separate machines for any aggregators we might have. This example demonstrate how to create a dedicated collectd aggregator. collectd_forwarder defines the host that is the dedicated collectd aggregator. Default is an empty string, '', which configures the network plugin to not send to a collectd aggregator.

[collectd:children]
collectd-somedc-prod

[collectd-somedc-prod]
collectd1.somedc.prod         ansible_ssh_host=10.0.1.111

[... skipping ...]
[somedc-prod:children]
collectd-somedc-prod

[somedc-prod:vars]
collectd_forwarder = collectd1.somedc.prod

#
# Optional: if you are aggregating logs with logstash
# and are using ElasticSearch and Kibana, you might
# want to also keep track of your metrics through
# Kibana. If so, simply add the following to your
# global configuration - keep in mind that
# activating this will turn off any other
# forwarding you might have configured in
# favor of sending all entries to logstash
#
logstash_forwarder = collectd2.somedc.prod

collectd_forward_to_logstash = true

The following may also be added to your inventory.

  • collectd_interval: at what interval in seconds to take measurements (default: 60)
  • check_disk: a value which we will use to select what disk to monitor (default: xvde)
  • collectd_df_fs_type: the file system type to monitor (Collectd 5.4 default: ext4. Collectd 5.5 default: undefined)
  • collectd_df_device: the dev device to monitor.
  • collectd_df_mountpoint: the mount point to monitor.
  • collectd_df_IgnoreSelected: see wiki (default: false)
  • collectd_df_ReportByDevice: see wiki (default: false)
  • collectd_df_ReportInodes: see wiki (default: false)
  • collectd_df_ValuesAbsolute: see wiki (default: true)
  • collectd_df_ValuesPercentage: see wiki (default: false)
  • monitor_coretemp: set to true if you want to monitor coretemp (only useful on real hardware)
  • private_ip: set to the IP address of the collectd_forwarder if collectd_forwarder has multiple IPs. Example:
collectd1.somedc.prod         ansible_ssh_host=10.0.1.111   private_ip=172.16.1.111

Plugin configuration

The collectd_plugins variable defines the plugins to be loaded. Default plugins are below:

  - syslog
  - cpu
  - interface
  - load
  - memory
  - vmem
  - swap
  - disk
  - rrdtool
  - network
  - librato
  - exec
  - write_graphite

Disk

Default: xvde

check_disk: xvde

DF

All options are optional.

The following should be set as an array. Example 1: (Single value)

collectd_df_device: '["/dev/xvda1"]'
collectd_df_mountpoint: '["/"]'
collectd_df_fs_type: '["xfs"]'

Example 2: (Multiple values)

collectd_df_device: '["/dev/xvda1","/dev/xvdb1"]'
collectd_df_mountpoint: '["/","/data"]'
collectd_df_fs_type: '["xfs","ext4"]'

You can change the rest of the defaults.

collectd_df_IgnoreSelected: false
collectd_df_ReportByDevice: false
collectd_df_ReportInodes: false
collectd_df_ValuesAbsolute: true
collectd_df_ValuesPercentage: false

Install plugins

To indicate the plugins that you would like installed, list all the plugins in the packages variable in your group_vars folder.

For example, below installs only the collect-rrdtools, collectd-rrdtool and collectd-iptables plugins.

packages:
  - collectd-rrdcached
  - collectd-rrdtool
  - collectd-iptables

If you do not set packages, the following default packages that will be installed.

  - git
  - collectd
  - collectd-amqp
  - collectd-apache
  - collectd-bind
  - collectd-curl
  - collectd-curl_json
  - collectd-curl_xml
  - collectd-dbi
  - collectd-dns
  - collectd-generic-jmx
  - collectd-iptables
  - collectd-ipvs
  - collectd-java
  - collectd-lvm
  - collectd-mysql
  - collectd-netlink
  - collectd-notify_desktop
  - collectd-postgresql
  - collectd-rrdcached
  - collectd-rrdtool
  - collectd-sensors
  - collectd-snmp
  - collectd-write_riemann

The following command will list the possible packages that can be installed:

  • CentOS 6: sudo yum --enablerepo="epel,gf-plus" list available | grep collectd-
  • CentOS 7: sudo yum --enablerepo="epel" list available | grep collectd-

Roles addition

It is possible for any roles to add their own custom metric collection configuration.

In myrole/templates/collectd.conf.j2:

<Plugin "whatever">
    ...
</Plugin>

Then, in myrole/tasks/main.yml:

- name: Adding collectd monitoring for myrole
  template: >
    src=collectd.conf.j2
    dest=/etc/collectd.d/myrole.conf
  notify:
    - Restart collectd
  tags:
    - myrole
    - files
    - collectd

Execution

For any machines

Simply follow the standard provisioning method.

For the collector machine

When provisioning with this role, you can also add Librato and Graphite support.

As you should not add any credentials to your inventory, it should be done by using --extra-vars as described below.

Librato

--extra-vars="use_librato=true" \
--extra-vars="[email protected]" \
--extra-vars="librato_token=API-KEY"

Graphite

--extra-vars="use_graphite=true" \
--extra-vars="graphite_host=some.host.com" \
--extra-vars="graphite_port="

See also