What the portal shows you

The Kubernetes section of the portal is not just where you download the kubeconfig. It carries several things you cannot get from kubectl at all, and two of them save real money.

Views

Tab	What it gives you
Overview	Quota used against your limits, and a count of what is running
Workloads	Every deployment and stateful set, with health and right-sizing suggestions
Pods	Status, restarts, age, and per-container state
Events	Recent cluster events for your namespace
Storage	Each volume's capacity, how full it is, and its autoresize settings
Usage	Month-to-date and projected cost, broken down, with a spending cap
Sleep	Which workloads are asleep and what that is saving

Right-sizing suggestions

The Workloads view compares what each workload requests against what it actually uses, taking the 95th percentile over the last seven days, and suggests a request that fits, with the monthly saving next to it.

This is the fastest way to find money. Because billing follows requests rather than usage, an over-provisioned workload costs full price while looking perfectly healthy, and nothing else will tell you. See Billing and quotas.

The suggestion is a starting point, not an instruction. A workload with a spiky profile needs headroom that a percentile does not capture, so treat it as evidence rather than a value to paste in blindly.

Cost and the spending cap

The Usage view shows month-to-date consumption, a projection for the month, and the split between CPU, memory, storage, egress and the base fee.

Two numbers there are worth understanding:

• The projection assumes the rest of the month looks like the part so far. Early in the month it is marked low-confidence, because it is.
• The floor is what the namespace costs if everything sleeps from now on: the base fee plus storage. It is the number to compare against when deciding whether scale-to-zero is worth setting up.

You can also set a spending cap on the namespace, which is the thing to do before experimenting with anything that scales.

The same things from kubectl

Some of this has a command-line equivalent, and some does not.

Current CPU and memory per pod:

kubectl top pods

Logs, including following them live, which the portal's log view does not do:

kubectl logs -f deployment/myapp
kubectl logs deployment/myapp --previous

Quota:

kubectl get resourcequota -o yaml

There is no kubectl equivalent for cost, the projection, the right-sizing suggestions or the spending cap. Those exist only in the portal.

What the portal does not do

• The log view is a snapshot, not a stream. It shows a fixed number of recent lines and refreshes on an interval. For live output use kubectl logs -f.
• Logs are not archived. Both the portal and kubectl show what the container currently holds, so a restart loses the previous output unless you ask for --previous. If you need log history, ship it somewhere yourself.
• The event list is capped at the most recent entries.

What's next

• Billing and quotas for what the numbers mean
• Scale-to-zero for cutting the idle part of the bill