CRI-O is not following Kubernetes conventions for annotations. See https://kubernetes.io/docs/reference/labels-annotations-taints/ for what actual Kubernetes annotations look like. This means we could either migrate to the new style or solve the issue for future annotations by enforcing conventions (using docs?).
Refers to: #7779, kubernetes/website#45121 (comment)
Migrating the annotations will require a conversion as well as deprecation phase. We did the same some time ago for our metrics, which are still in deprecation:
|
metricOperations *prometheus.CounterVec // Deprecated: in favour of metricOperationsTotal |
|
metricOperationsLatency *prometheus.GaugeVec // Deprecated: in favour of metricOperationsLatencySeconds |
|
metricOperationsLatencyTotal *prometheus.SummaryVec // Deprecated: in favour of metricOperationsLatencySecondsTotal |
|
metricOperationsErrors *prometheus.CounterVec // Deprecated: in favour of metricOperationsErrorsTotal |
|
metricImagePullsByDigest *prometheus.CounterVec // Deprecated: in favour of metricImagePullsBytesTotal |
|
metricImagePullsByName *prometheus.CounterVec // Deprecated: in favour of metricImagePullsBytesTotal |
|
metricImagePullsByNameSkipped *prometheus.CounterVec // Deprecated: in favour of metricImagePullsSkippedBytesTotal |
|
metricImagePullsFailures *prometheus.CounterVec // Deprecated: in favour of metricImagePullsFailureTotal |
|
metricImagePullsSuccesses *prometheus.CounterVec // Deprecated: in favour of metricImagePullsSuccessTotal |
|
metricImagePullsLayerSize prometheus.Histogram |
|
metricImageLayerReuse *prometheus.CounterVec // Deprecated: in favour of metricImageLayerReuseTotal |
|
metricContainersEventsDropped prometheus.Counter |
|
metricContainersOOMTotal prometheus.Counter |
|
metricContainersOOM *prometheus.CounterVec // Deprecated: in favour of metricContainersOOMCountTotal |