Installing and setting up telegraf to monitor docker as a service on Linux

Telegraf is a data collection and aggregation agent that needs to be installed on target server. Let's say you would like to gather metric data for example cpu usage or memory usage for Server A, install Telegraf on server A. All these day will be sent to a central influxdb datastorem, lets call our influxdb ServerDatastore.

Grafana and influxdb can live on the same or separate server.

In this setup, we are going to setup telegraf to monitor docker application and send all the metric of interest to ServerDataStore. 

Installing telegraf on target server

curl ttps://dl.influxdata.com/telegraf/releases/telegraf_1.13.2-1_amd64.deb --output telegraf.deb

Placing telegraf executable in a specific folder.

After extracting our telegraf.deb, you can move the binary to /usr/local/bin folder and configuration file to /user/local/etc/telegraf/telegraf.conf

Next we need to specify what metrics we would like telegraf to send over.   Example below shows the key configuration for DEV environment that we are interested in.

The first is target database and server dns. Here we setup our hostname "urls = "http://ServerDataStore:8086" and database is called "dev_docker_telegraf".

# Configuration for sending metrics to InfluxDB

[[outputs.influxdb]]

  ## The full HTTP or UDP URL for your InfluxDB instance.

  ##

  ## Multiple URLs can be specified for a single cluster, only ONE of the

  ## urls will be written to each interval.

  # urls = ["unix:///var/run/influxdb.sock"]

  # urls = ["udp://127.0.0.1:8089"]

  urls = ["http://ServerDataStore:8086"]

  ## The target database for metrics; will be created as needed.

  ## For UDP url endpoint database needs to be configured on server side.

  database = "dev_docker_telegraf

Docker metrics is configured by uncommenting the following sections. It should look something below :-

# # Read metrics about docker containers

[[inputs.docker]]

#   ## Docker Endpoint

#   ##   To use TCP, set endpoint = "tcp://[ip]:[port]"

#   ##   To use environment variables (ie, docker-machine), set endpoint = "ENV"

endpoint = "unix:///var/run/docker.sock"

#

#   ## Set to true to collect Swarm metrics(desired_replicas, running_replicas)

gather_services = false

Example of setup for dev environment are shown below :-

Please make sure you have the configuration updated, saved and place in a specific folder before starting the service.

Running telegraf as an agent

We will setup telegraf to run as an agent. First, we need to create a configuration file call "telegraf.service".

Creating the configuration

Create a file called 'telegraf.service' under '/etc/systemd/system'. It should contain the following :-

[Unit]

Description=telegraf.servicee

[Service]

ExecStart=/usr/local/bin/telegraf --config /usr/local/etc/telegraf/telegraf.conf

[Install]

WantedBy=multi-user.target:

Running the service

sudo systemctl start telegraf.service

Checking on service  status

sudo systemctl status telegraf.service

Stop the service

This might come in handy if you would like to shutdown a service for testing purposes.

sudo systemctl stop telegraf.service