/kind bug
What happened:
The occm Helm chart exposes Prometheus metrics and documents them in docs/metrics.md, but the default ServiceMonitor does not successfully scrape those metrics, especially when deployed with prometheus.
we observed some issues with the generated resources,
- The
ServiceMonitor selector does not match the Service created by the chart.
- The
ServiceMonitor still relies on the deprecated bearerTokenFile authentication.
- As a result, OCCM metrics are not collected well.
without working metrics, the occm reconciliation failures or repeated OpenStack API errors can go unnoticed until users begin experiencing faileds.
What you expected to happen:
Using the ServiceMonitor should allow Prometheus to scrape OCCM metrics without additional manual configuration.
OCCM operators should be able to consume the metrics described in docs/metrics.md and build alerts for controller reconciliation failures and openstack api errors.
How to reproduce it:
1. deploy `openstack-cloud-controller-manager` using the upstream Helm chart.
2. enable `ServiceMonitor`.
3. deploy prometheus operator.
4. observe that the `ServiceMonitor` does not successfully scrape the OCCM metrics,
because the generated resources do not correctly reference the `Service` and still use deprecated authentication.
Anything else we need to know?:
This issue was identified while integrating OCCM monitoring into production deployments.
The proposed fix is available in PR #3039, which is,
- updates the ServiceMonitor selector to match the generated Service,
- removes the deprecated
bearerTokenFile usage,
- replaces the generated Service and ServiceMonitor manifests with resources following the current recommendations from
docs/metrics.md.
The goal is to make OCCM metrics available without any additional configurations!
Environment:
- openstack-cloud-controller-manager(or other related binary) version: master branch
- OpenStack version:
- Others: prometheus operator, helm chart deployment
/kind bug
What happened:
The occm Helm chart exposes Prometheus metrics and documents them in
docs/metrics.md, but the defaultServiceMonitordoes not successfully scrape those metrics, especially when deployed with prometheus.we observed some issues with the generated resources,
ServiceMonitorselector does not match theServicecreated by the chart.ServiceMonitorstill relies on the deprecatedbearerTokenFileauthentication.without working metrics, the occm reconciliation failures or repeated OpenStack API errors can go unnoticed until users begin experiencing faileds.
What you expected to happen:
Using the
ServiceMonitorshould allow Prometheus to scrape OCCM metrics without additional manual configuration.OCCM operators should be able to consume the metrics described in
docs/metrics.mdand build alerts for controller reconciliation failures and openstack api errors.How to reproduce it:
Anything else we need to know?:
This issue was identified while integrating OCCM monitoring into production deployments.
The proposed fix is available in PR #3039, which is,
bearerTokenFileusage,docs/metrics.md.The goal is to make OCCM metrics available without any additional configurations!
Environment: