Load & resize lab
Send test traffic to one workload, resize it in place and tune its policy. Every change goes through the same safety checks as the console; in observe mode they are recorded and refused.
Test traffic
More traffic pushes CPU past the HPA target, which adds pods. Less traffic lets the HPA remove them again.
Pods 0 running
| Pod | Status | Node | IP | CPU used | CPU request / limit | Memory used | Memory request / limit | Restarts |
|---|---|---|---|---|---|---|---|---|
| Loading… | ||||||||
Automatic sizing
When on, the lab resizes the pod to match the current traffic tier by itself. Turn it off to set sizes by hand.
Resize by hand
Pick a starting point or type exact values, then apply. Running containers change in place without a restart.
Only the main application container is resized; sidecar containers (Istio, log shippers) are left alone. Savings are the requests freed; the bill drops when nodes are consolidated.
Tier and contract
What this workload is allowed to have done to it.
Preview before writing
These checks read only. Nothing is written to the cluster.
Who sets the replica count
Only one controller may own the count. nk8 refuses to scale while an HPA or KEDA owns it.
Recorded history
What the advisor and the forecast read, one row per hour.
Exemption
Keep this workload out of automation.
Safety checks
Checks that can stop a change.
Horizontal scaling
Vertical sizing
Rollout
How a resize moves through the pods.
Events
Controller decisions, in-place resizes, rate changes and HPA scaling, newest first.