-

Skip to content
SHOW
August 21, 2017
Setting up node_exporter for helm based kubernetes installs.

Prometheus is an open source library for collecting metrics from cloud native
systems (or any systems, really). 

Especially in distributed system, it could get difficult to measure
things in one, but thanks to Prometheus, we could grab metrics from each node
and gain some insights on scalability and etc.

Node exporter is used to export the system level metrics on each node of our
kubernetes cluster. However, the helm recipes for kube dont start exporters
everywhere.

So, the easiest way to get your exporters on every node w/o needing docker in
privileged mode:

sudo apt install golang-go
echo export GOPATH=$HOME/go >> ~/.bash_profile
echo export PATH=$PATH:$GOROOT/bin:$GOPATH/bin >> ~/.bash_profile
source ~/.bash_profile
go get github.com/prometheus/node_exporter
cd $GOPATH/src/github.com/prometheus/node_exporter/
make
sudo ./node_exporter

Note here that I run node_exporter as a sudo privilege. As mentioned it’s for
grabbing the system level metrics, so it was necessary (/proc). Otherwise you’ll get a
lot of error message on not able to read/permissions etc.

KUBERNETES: DON'T FORGET TO ADD YOUR NODES TO THE CONFIG MAP

Once the node exporters are set up correctly on each of the node, and exporting
the metrics, I went back to the master node and modified the configmap for
setting the targets.

In the configmap where we store the configuration for Prometheus, I’ve added the
node name under the following section.

- job_name: 'node_exporter'
static_configs:
-targets: ['localhost:9100', 'ip-node1name:9100', 'ip-node2name:9100'.. ]

Node names could be retrieved by doing

kubectl get nodes

Once this is all set, open the prometheus service on the browser, which runs on
[master node public ip]:9090.

Go into the Status menu, click on Targets in order to double check if the
Prometheus is connecting to all the other nodes right.

Once all the states are ‘UP’ for the nodes, then you are good to go.

Start measuring things!

Advertisements

Share this:
TwitterFacebook

Post navigation
PREVIOUS: Previous post: Back to using pandas and Jupyter
Leave a Reply


Enter your comment here...
Blog at WordPress.com. Claire Keum
Follow
 :)