The changes introduce a new metric generator `kube_pod_status_unscheduled_time` to record the Unix timestamp when pods transition to unscheduled status. This enhances monitoring capabilities for scheduling issues by capturing exact transition times during `PodScheduled` condition changes.
pkg/metric_generator/generator.go:32:22: 16 bytes saved: struct with 96 pointer bytes could be 80
pkg/metrics_store/metrics_store.go:31:19: 40 bytes saved: struct with 64 pointer bytes could be 24
pkg/options/options.go:42:14: 24 bytes saved: struct of size 384 could be 360
internal/store/builder.go:67:14: 8 bytes saved: struct of size 200 could be 192
internal/store/pod.go:1351:16: 8 bytes saved: struct with 16 pointer bytes could be 8
internal/store/pod.go:1477:20: 8 bytes saved: struct with 16 pointer bytes could be 8
internal/store/testutils.go:32:30: 16 bytes saved: struct with 136 pointer bytes could be 120
pkg/metricshandler/metrics_handler.go:46:21: 8 bytes saved: struct of size 104 could be 96
internal/discovery/types.go:39:19: 32 bytes saved: struct with 88 pointer bytes could be 56
pkg/customresourcestate/config.go:51:15: 16 bytes saved: struct with 112 pointer bytes could be 96
pkg/customresourcestate/config.go:134:16: 8 bytes saved: struct with 88 pointer bytes could be 80
pkg/customresourcestate/config.go:150:13: 8 bytes saved: struct with 40 pointer bytes could be 32
pkg/customresourcestate/config_metrics_types.go:29:18: 8 bytes saved: struct with 64 pointer bytes could be 56
pkg/customresourcestate/config_metrics_types.go:42:17: 8 bytes saved: struct with 40 pointer bytes could be 32
pkg/customresourcestate/registry_factory.go:125:21: 8 bytes saved: struct with 40 pointer bytes could be 32
pkg/customresourcestate/registry_factory.go:212:20: 16 bytes saved: struct with 88 pointer bytes could be 72
pkg/customresourcestate/registry_factory.go:377:23: 8 bytes saved: struct with 104 pointer bytes could be 96
pkg/customresourcestate/registry_factory.go:497:21: 8 bytes saved: struct with 64 pointer bytes could be 56
pkg/customresourcestate/registry_factory.go:549:13: 8 bytes saved: struct with 24 pointer bytes could be 16
Returning empty string instead of "<none>" value for "owner_kind", "owner_name" and "owner_is_controller" dimensions when no metadata.ownerReferences exists in Kubernetes resoures.
Recommend metrics exposed by the kube-scheduler, in case of the ones below:
* `kube_pod_container_resource_limits`
* `kube_pod_container_resource_requests`
Signed-off-by: Pranshu Srivastava <rexagod@gmail.com>
Add experimental kube_pod_ips metric to show all IPs on a pod.
This is useful for dual-stack clusters where pods will have at least one
IPv4 and one IPv6 address. Current kube_pod_info metric only shows
the single IP exposed by .status.podIP.
Signed-off-by: Braxton Schafer <bschafer@cloudflare.com>
Problem: In https://github.com/kubernetes/kube-state-metrics/pull/1723 a
potential panic in the pod metrics gathering was fixed by working around
a disconnect of `Spec.Containers` and `Status.ContainerStatuses`. The
slice storing the resulting metrics however was still defined based on
the length of the `Status.ContainerStatuses` list.
Solution: Make the slice dynamic and append metrics to it.
Signed-off-by: Jan Fajerski <jfajersk@redhat.com>
Nour
yshngg
Carlos Moreno
Rick Rackow
Kubernetes Prow Robot
leiwingqueen
Manuel Rüger
Tetiana Kravchenko
changhyuni
adinhodovic
Kubernetes Prow Robot
opeco17
Swarup Srinivasan
Pranshu Srivastava
Catherine Fang
Ryan R. Olds
Ryan Olds
Lan Liang
frezes
Paweł Kubica
Shaun Sabo
Julian van den Berkmortel
Braxton Schafer
honpey
Jan Fajerski
Michal Fojtik
Sherif Abdel-Naby
Eric Hole
Min Wang
Gabi Davar