Metrics Reference¶
The cloud-vinyl operator exposes Prometheus metrics on port 8080 at /metrics.
Metrics¶
Metric |
Type |
Labels |
Description |
|---|---|---|---|
|
Counter |
|
Total reconcile loop invocations. |
|
Histogram |
|
Reconcile loop duration in seconds. |
|
Counter |
|
Total VCL push attempts. |
|
Histogram |
|
VCL push duration in seconds. |
|
Gauge |
|
Backend health status. |
|
Gauge |
|
Cache hit ratio (0.0–1.0) aggregated across pods. |
|
Counter |
|
Total invalidation requests. |
|
Counter |
|
Total broadcast requests from the proxy. |
|
Histogram |
|
Broadcast fanout duration in seconds. |
|
Counter |
|
VCL push operations where some but not all pods failed. |
|
Gauge |
|
Number of VCL versions currently loaded in Varnish (drift indicator). |
Alerts¶
When monitoring.prometheusRules.enabled: true, the following alerts are deployed:
Alert |
Severity |
Condition |
|---|---|---|
|
warning |
VCL push error rate > 0 for 5 m |
|
warning |
Partial failure counter > 0 for 2 m |
|
critical |
Push errors and reconcile errors both > 0 for 1 m |
|
warning |
Backend health == 0 for 5 m |
|
warning |
Hit ratio < 50% for 15 m |
|
warning |
>100 invalidations/s for 5 m |
|
warning |
Reconcile error rate > 0 for 5 m |
|
critical |
No reconcile metrics present for 5 m |
|
warning |
>2 VCL versions loaded for 10 m |
|
warning |
>10% broadcast error rate for 5 m |