Skip to main content
Kubernetes records an event when something happens to an object: a pod is scheduled, an image fails to pull, a probe fails, a node runs low on memory. Events lists them, newest first. Open it from the sidebar under Cluster, with G then E, or with ⌘6 in the desktop app (Ctrl6 on Windows and Linux).
The Events list: warnings like BackOff, Failed and FailedScheduling among normal events, each with its object, message, count and when it was last seen.The Events list: warnings like BackOff, Failed and FailedScheduling among normal events, each with its object, message, count and when it was last seen.

What happened in the cluster, newest first.

Events have hashed names, so the list has no Name column, and opening one shows its reason and object as its title. Events aren’t changed by hand: they have no actions, and their rows can’t be selected.

Only the warnings

The chips above the list count each type. Pick Warning to hide the routine events. The overview’s Warnings tile opens the list this way. The filter (/) matches an event’s reason, message and object name as well as its labels, so typing OOMKilled or a pod’s name finds what you’re after.

On a timeline

List | Timeline, in the list’s bar, shows the same events where they fell in time. The timeline has a row for each object, with its kind and name; click the name to open the object. A Normal event is a hollow circle and a Warning a triangle. One that repeated is a bar from its first time to its last, with how many times beside it where there’s room, and always in its card. Marks too close to tell apart are drawn as one, with how many, and so are repeated events too many to stack in a row; click one to zoom in and part them. A key under it says what each mark means.
The cluster's events on a timeline, warnings first: each a bar from when it began to when it was last seen.The cluster's events on a timeline, warnings first: each a bar from when it began to when it was last seen.

Events on a timeline: each a bar from when it began to when it was last seen.

It shows the events the page’s filter and chips leave, so pick Warning to see only the warnings in time. Above it, choose 15m, 1h or All (every event the cluster keeps), and drag across the times to zoom in; Reset zoom goes back. A stretch with no events says so, with Reset zoom. Hover a mark, or move to it with the keyboard, to see what happened: its reason, type and message, and when, with its first and last time and how many times for one that repeated. Click a mark, or press Enter on it, to open the event. With the timeline focused, the arrow keys move between rows and along a row’s marks, Home and End go to a row’s first and last, and Escape lets go. The choice is kept in the page’s address (?as=timeline), so a link opens the timeline too. On a phone, or where the page is too narrow for it (beside an open object, say), events stay a list.

Notifications of new warnings

Lumovi can tell you of new Warning events in a cluster as they happen, through your system’s notifications. It’s off until you turn it on, for each cluster:
A cluster's settings: new warnings told while Lumovi runs, in the namespaces it watches.A cluster's settings: new warnings told while Lumovi runs, in the namespaces it watches.

Notifications of new warnings, in a cluster’s settings.

  • On the desktop, in the cluster’s settings: Notify me of new warnings. While Lumovi is open, whatever page it shows, it checks every 30 seconds.
  • On a server, in the cluster switcher at the top of the sidebar: Notify me of new warnings. It’s yours alone, kept in this browser, and works while one of Lumovi’s tabs is open in it, hidden or not. Your browser asks first: Allow notifications, then Send a test to see one. A browser that blocks them says so, and the switch stays on until you allow them in its settings for the site; one that can’t show notifications at all says “This browser can’t show notifications.” With several tabs open over HTTPS, one of them checks, and you’re told once.
Namespaces to watch is All namespaces, or These namespaces, picked from the cluster’s. On the desktop, turned on, it starts with the namespace the cluster opens in. Lumovi reads their Warning events as you, so it sees only what you may read, and checks every 30 seconds. What’s new is an event that wasn’t there at the last check, or one whose count went up. What was there when it started is only remembered. A notification says the reason and the object, like “BackOff: Pod shop/cart-7d9f6c5b8-x2kqp”, the message’s first line, and the cluster’s name last. The same object and reason is told once in 10 minutes. Several new in one namespace at once come as one, like “5 new warnings in shop”, with “BackOff ×3, FailedScheduling, Unhealthy”. Click one to open Lumovi at the object, or at that namespace’s warnings. A check where any read fails tells nothing, and keeps what it saw, so an old warning is never told as new. If signing in fails, the cluster’s notifications pause: one notification says “Notifications paused”, and its settings say “Paused at 14:02: signing in failed. Open the cluster to sign in again, and notifications go on.” Lumovi doesn’t try signing in again by itself until you do. For a cluster that signs in through a credential plugin, its settings say so, like “Signing in (kubelogin) runs as it checks, as when you open the cluster.”: a plugin that signs in through a browser may open one at the first check.
A notification is shown wherever your system shows them, its lock screen and notification history included. An event’s message can carry what only you may read, so Lumovi cuts it as it does for a plugin’s pages: at the first place it could, like “Liveness probe failed: (the rest left out by Lumovi)”, and at 200 characters. The reason, the object’s name and the cluster’s name are shown as they are. To keep them off your lock screen, turn off previews there in your system’s settings.

An object’s own events

Every object’s detail panel has an Events tab with the events about it, newest first. It’s often the fastest way to see why a pod won’t start or a volume won’t bind. The overview lists the latest warnings across the namespace, too. When an object has more than 200 events, Filter events appears above the list. It finds events by their reason, their message or what reported them, and says how many match, like 12 of 3,000. Your browser’s own find reaches only the events on screen then, since a long list is drawn as you scroll. Where the panel is wide enough, the tab has List | Timeline too: its timeline has a row for each reason, like BackOff or Pulled, and works as the page’s does. Lumovi keeps your choice for every object’s tab.
Kubernetes keeps events for a limited time (an hour, unless the cluster is set up otherwise), so old ones disappear. The timeline says so under it: All is only what’s still kept. For a longer view of what a workload did, see Usage history and its logs.

Overview

The latest warnings, next to what needs attention.

Health and status

What the statuses on each object mean.