periscopev1.0.7

docsusageevents

Events

The Events page is the cluster-wide event feed — every K8s event from every namespace, newest first, live-streaming via SSE so new events appear at the top without a refresh. Distinct from the per-resource events tab…

The Events page is the cluster-wide event feed — every K8s event from every namespace, newest first, live-streaming via SSE so new events appear at the top without a refresh. Distinct from the per-resource events tab (covered in workloads.md) which scopes to a single object's involvedObject feed.

What you see

Events page on peri-server with seeded broken workloads
Events page on peri-server with seeded broken workloads

The page header shows the total event count, an active-stream indicator (live pill, green when the SSE connection is healthy), the namespace picker, and the + apply yaml quick-action.

The filter strip below carries:

  • Type pills — all (default) / warning / normal. The warning count is shown explicitly (5 warnings in the screencap).
  • Search — substring match across event name, reason, message.

Each row carries:

  • OBJECT — kind / name of the event's involvedObject. Kind badges (POD, DEPLOYMENT, HORIZONTALPODAUTOSCALER, REPLICASET) render in monospace as a small uppercase chip.
  • NAMESPACE — where the involved object lives. Empty for cluster-scoped resources.
  • REASON — kubectl-style reason in monospace. Warnings render yellow (BackOff, Failed, FailedScheduling); Normal events render in muted text.
  • MESSAGE — kubelet's human-readable detail (truncated; full text on hover).
  • COUNT — kubelet collapses repeated identical events into one row with a count. Yellow at >100 occurrences, red at >999.
  • SOURCE — kubelet, replicaset-controller, horizontal-pod-autoscaler, etc.
  • AGE — relative timestamp.

Warning rows render with a left border tint so an operator scanning the feed catches them at a glance.

The screencap above is from peri-server with the seeded broken workloads — web-7476d7bdf8-* Pods showing ImagePullBackOff from the nginx:1.27.99 typo, checkout-* Pods showing BackOff from the deliberate-fail container, plus normal ScalingReplicaSet and SuccessfulCreate events from the healthy parts of the fleet. Real operational mix, not synthetic.

Live updates

The page connects to the cluster's watch streams for events specifically. New events arrive at the top in real time; the SSE connection auto-reconnects on transient drops with a small "polling fallback" indicator if the watch stream can't be reestablished (rare, usually indicates apiserver-side issues).

RBAC

The Events page needs events:list cluster-scoped (or namespace-scoped if you want a single-namespace view). The standard view ClusterRole carries this. When the role denies the list, the page renders a forbidden state with the same "your role can't see these" wording the rest of the SPA uses.

Related docs

← previousWorkloadsnext →NetworkPolicies
edit this page on github →