Observability
You should not need Grafana, three browser tabs, and a spreadsheet to answer how the fleet is doing. In Cloud Admin, open Monitoring & Ops → Observability for Prometheus-backed charts, inventory, and drill-downs—from a quick summary to node- and pod-level detail.
Screenshots from the live product. Each note explains what you are looking at and when you would open this screen.
Real-time Performance Analytics area chart for CPU, memory, GPU, and workloads over 24 hours. This is the same interface your team uses in production.
Click the screenshot to open full size, zoom, and pan.
Historical trends reveal whether usage is seasonal, spiking, or creeping—real-time performance analytics line chart with metric summary cards. Pair with capacity planning before users feel pain.
Click the screenshot to open full size, zoom, and pan.
In Memory Pool, the Overview view answers one operational question at a time—memory pool analytics overview with total, used, free memory and utilization. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In Memory Pool, the Usage view answers one operational question at a time—memory pool analytics usage tab with working set and cached memory. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In Memory Pool, the Breakdown view answers one operational question at a time—memory breakdown by namespace and top memory-consuming pods. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In Memory Pool, the Trends view answers one operational question at a time—24-hour memory usage trend with min, max, and average. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In Memory Pool, the Issues view answers one operational question at a time—memory health issues tab with pressure and oom checks. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In CPU Performance, the Overview view answers one operational question at a time—cpu performance analytics overview with usage, cores, and limits. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In CPU Performance, the Usage view answers one operational question at a time—cpu usage tab with used, idle cpu and overcommit ratio. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In CPU Performance, the Breakdown view answers one operational question at a time—cpu breakdown by namespace and top cpu pods. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In CPU Performance, the Trends view answers one operational question at a time—24-hour cpu usage trend chart and statistics. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
In CPU Performance, the Issues view answers one operational question at a time—cpu health checks for usage, overcommit, and throttling. Part of Monitoring workspace; use it when you need evidence before changing limits, scaling, or opening a ticket.
Click the screenshot to open full size, zoom, and pan.
Fleet KPIs, saturation charts, flow telemetry, and Prometheus-backed alerts share one investigation cadence: confirm scope, isolate the layer that moved, validate mitigation.
Slice traffic by cluster total, namespace, node NIC, or workload; correlate spikes with pod-level tables instead of tcpdump theater.
Allocation donuts, 24h trends, namespace and node breakdowns, plus ranked pods—ideal when HPA metrics disagree with user-visible latency.
Alert summaries pair severity counts with the charts that explain them—fewer circular conversations about “whether Prometheus fired.”
Nodes, namespaces, controllers, ConfigMaps, Secrets, NetworkPolicies, and Services appear beside charts—no separate CMDB hunt mid-incident.