Monitoring

How to install and configure collectd?

install and configure collectd
Install and configure collectd

How to install and configure collectd?

This post helps to install collectd monitoring agent and explains it’s configuration. collectd as its name implies collects performance metrics from different sources like servers or applications and it will transfer that data to another system for analysis.

collectd only collects and transmits metrics, and any number of tools can be arranged to receive that data. The design of collectd is agent based, so you need to run collectd on hosts that you want to monitor.

There are 95+ input plugins and 15+ write plugins are available in collectd. The input plugins will help to collect the metrics from a system and the output plugins will send that metrics to a remote system for analysis.

Examples of Input (read) plugins
  • CPU
  • Memory
  • MySQL
  • SNMP
Examples of Output (write) plugins
  • Graphite
  • RRDtool
  • MongoDB
  • HTTP

Install collectd on CentOS 7

If you are installing collectd on CentOS 7, you need to enable epel repo first.

yum install epel-release
yum install collectd

Install collectd on Ubuntu or Debian based distros

apt install collectd

Configure collectd

Once the package is installed, open /etc/collectd.conf and configure it.

vim /etc/collectd.conf

Then enable desired plugins and set the hostname. The hostname should be the local hostname. Also, if you would like to send the metrics to a graphite server, enable write_graphite plugin too:

Add the following into it:

Hostname    "SERVER01"

LoadPlugin syslog
LoadPlugin cpu
LoadPlugin load
LoadPlugin memory
LoadPlugin uptime
LoadPlugin write_graphite

<Plugin write_graphite>
  <Node "metrics">                         # Your Graphite server name
    Host "mtrics.grepitout.com"            # Your graphite server hostname
    Port "2003"                            # graphite server port
    Protocol "tcp"
    ReconnectInterval 0
    LogSendErrors true
    Prefix "collectd."
    Postfix ".os"
    StoreRates true
    AlwaysAppendDS false
    EscapeCharacter "_"
    SeparateInstances false
    PreserveSeparator false
    DropDuplicateFields false
  </Node>
</Plugin>
Sample configuration file:
[root@application01]# cat /etc/collectd.conf
Interval     60

Hostname "application01"

TypesDB "/opt/collectd/share/collectd/types.db"
TypesDB "/opt/collectd/share/collectd/types.d.custom"

LoadPlugin syslog
LoadPlugin cpu
LoadPlugin load
LoadPlugin memory
LoadPlugin uptime
LoadPlugin vmem
LoadPlugin entropy

CollectInternalStats true

Include "/etc/collectd.d/*.conf"
[root@application01]

You can add seperate configuration files for your plugins in the directory /etc/collectd.d/ as I have added an Include option in the main configuration file.

Example: /etc/collectd.d/plugin-write_graphite.conf

[root@application01]# cat /etc/collectd.d/plugin-write_graphite.conf
LoadPlugin write_graphite

<Plugin write_graphite>
  <Node "metrics">                         # Your Graphite server name
    Host "metrics.grepitout.com"           # Your graphite server hostname
    Port "2003"                            # graphite server port
    Protocol "tcp"
    ReconnectInterval 0
    LogSendErrors true
    Prefix "collectd."
    Postfix ".os"
    StoreRates true
    AlwaysAppendDS false
    EscapeCharacter "_"
    SeparateInstances false
    PreserveSeparator false
    DropDuplicateFields false
  </Node>
</Plugin>
[root@application01]#
CollectInternalStats

When we set CollectInternalStats to true, the collectd daemon statistics will be collected with “collectd” as the plugin name and send it to the tool that is arranged to receive that data.

TypesDB

The types.db file contains collectd’s metric type specifications. Each line describes one metric type, which is called “data set”.
Read more about types.db: GitHub: types.db

Also read more about Data source types (counter, gauge, derive or absolute): collectd.org: Data Source

The types.db and types.db.custom files will not be created upon collectd package installation. You can check sample types.db file in the follwoing link and place that in /opt/collectd/share/collectd/types.db

Sample types.db file: GitHub: types.db Sample

Enable and start collectd service

Once you made the desired changes enable and start the collectd service.

systemctl enable --now collectd.service
systemctl status collectd.service

Do this on all nodes that you need to monitor.

Verify

Go to the graphite host and check the metrics are receiving on the port 2003

tcpdump -i any -A -vv port 2003

That’s it!

Also Read:

If you like this post and wish to receive more articles from us, please like our FB page: Grepitout

Your suggestions and feedback will encourage us and help to improve further, please feel free to write your comments.

For more details on our services, please drop us an E-mail at info@grepitout.com

Add Comment

Click here to post a comment

Topics