The easiest way to get Openshift installed is using OpenShift Container Platform
You can set AWS machines types or GCP machines types to your master and workers nodes using plaform
parameter in your install-config.yaml
file.
Example for AWS:
platform:
aws:
type: c5.4xlarge
Example for GCP:
platform:
gcp:
type: n2-standard-4
This allow you to get more ressources in order to deploy Loki for example.
You can customize your OCP installation using install-config.yaml
file.
Generate the configuration using:
openshift-install create install-config --dir=<installation_directory>
NetObserv comes with a bunch of metrics, however they are not scraped by default by OpenShift Cluster Monitoring (the OpenShift cluster Prometheus that is used for infra monitoring).
You can install your own Prometheus to scrape NetObserv’s metrics.
As an alternative, you can also tell OpenShift Cluster Monitoring to scrape all user metrics, not just infra (provided as “USE AT YOUR OWN RISK”: depending on your running workloads and their metrics, it may put pressure on Prometheus and make it unstable).
We provide some YAML to do so:
oc apply -f https://raw.githubusercontent.com/netobserv/documents/main/examples/metrics/monitoring.yaml
It will create a Service
for flowlogs-pipeline metrics, two ServiceMonitors
(for flowlogs-pipeline and the console plugin), and configure Cluster Monitoring to scrape user metrics.
The generated metrics are prefixed with netobserv_
or flow_
.