Docs / Timeline, topology and metrics
Timeline, topology and metrics
What happened across your clusters, how traffic reaches a Pod, and how much it uses.
Timeline
Timeline shows what happened, newest first, in some namespaces (or all) across every open context, on one page. It merges:
- Kubernetes Events;
- rollouts: a new ReplicaSet for a Deployment, or a new ControllerRevision for a StatefulSet or DaemonSet, with the images;
- container restarts, each with its last termination reason and exit code;
- scaling, from Deployments and autoscalers;
- the changes Klystiq saw while it ran.
Pick the last hour, 6 hours, a day or a week, and filter by kind, warnings or text. The strip above the list shows warnings and other items over time; click a bar to jump there. Warnings have Ask AI. Most clusters keep Events for about an hour, so older stretches show revisions, restarts and changes only. The page refreshes every 15 seconds.
Topology
Topology is an interactive diagram of one or more namespaces, or all of them, in one context.
The Network view runs left to right from Gateways, routes and Ingresses through Services to Pods, then the ReplicaSets, Jobs and workloads that own them. Replicas of one owner are drawn as a single stacked card ("2/3 ready", with a tally of statuses and the worst health). Double-click it, or use Show each Pod, to see every Pod with its containers. With several namespaces, each gets a frame you can fold into a summary card; more than four start folded, except those with problems.
Chips choose what else to draw:
- Traffic and Workloads;
- Config: ConfigMaps and Secrets, by name only (Secret data is never read);
- Storage: claims and their volumes;
- Access: ServiceAccounts and the RoleBindings that grant them roles;
- Policies: NetworkPolicies linked to the Pods they apply to and the peers they allow. Each Pod is marked "isolated in", "isolated out" or both, and its panel lists who may connect to it and where it may connect, read the way Kubernetes defines it. Whether your network plugin enforces policies isn't something the API shows.
Hover a box to trace its chain; click it for details and actions (open, logs, shell, Ask AI). The Placement view groups Pods by Node. Filter by name, show only problems, pan and zoom. Topology refreshes every 5 seconds, and Show in topology in a resource's actions menu opens it focused on that resource. It's fully usable from the keyboard.
Metrics
Pods, Deployments, StatefulSets, DaemonSets, ReplicaSets, Jobs and Nodes have a Metrics tab. It charts CPU, memory (working set), network and restarts over the last hour, 6 hours, 24 hours or 7 days, with request and limit (or allocatable) lines, a crosshair tooltip and a table view. Workloads chart each Pod: the eight busiest, with the rest summed as Other.
The data comes from a Prometheus that already runs in the cluster. Klystiq finds a Prometheus, Thanos Query or VictoriaMetrics Service by itself and reaches it through the API server's service proxy with your own access, so nothing is installed or exposed. To use another one, set it in the context's settings, as namespace/service:port or a URL such as a port-forward on http://localhost:9090. Restarts need kube-state-metrics.
Health and alerts
The Health page lists everything failing right now across the open contexts. Ask AI about all of these treats them as one incident and asks for their likely common causes.
Turn on Tell me when something newly fails in Settings (it's off by default), and the bell in the tab bar collects Pods, Nodes, Deployments, StatefulSets, DaemonSets and Jobs that start failing and stay failed for 20 seconds, each with Open and Explain. Problems that already exist when it starts, or appear in its first 30 seconds, go to the Health page instead. A recovered object can alert again once it has been healthy for a minute.
With alerts on, Settings can also show a system notification for each new alert (a burst becomes one summary). Separately, it can put Klystiq in the menu bar (macOS) or system tray, with a menu that lists the open contexts, their connection state and the number of new alerts. Both are off by default.