|
|
||
|---|---|---|
| .. | ||
| prometheus/manifests | ||
| loki.yaml | ||
| opencost.yaml | ||
| prometheus.yaml | ||
| promtail.yaml | ||
| README.md | ||
Observability Stack
This directory contains an observability implementation based on Grafana tooling
Caveats
- reliance on ref-implementation for SSO
- This is possible to work around by removing the
auth.generic_oauthsection fromprometheus.yamland removing thegrafana-config.yamlandgrafana-external-secret.yamlfiles
- This is possible to work around by removing the
- using
tls_skip_verify_insecurefor oauth- This is due to using the ingress certificate. Once this is addressed, we can remove this
- Bigger memory requirement required for kind cluster
- Due to using a more robust loki deployment, the memory limits have been increased. 16 GB seems to work while leaving ample room in the cluster.
Components
The observability stack is built upon:
- Prometheus - metrics
- Loki - logging
- Promtail - log delivery
- Opencost - cost accounting
- Grafana - visualization
- Alertmanager - alerting
Installation
Note: The stack is configured to use Keycloak for SSO; therefore, the ref-implementation is required for this to work.
idpbuilder create --use-path-routing --package-dir ./ref-implementation --package-dir ./observability
A grafana-config job will be deployed into the keycloak namespace to create/patch some of the keycloak components. If deployed at the same time as the ref-implementation, this job will fail until the config job succeeds. This is normal