Skip to main content

Autoscale

Purpose​

Autoscale — define per-VM policies that automatically grow or shrink a VM's vCPUs and memory within set bounds based on load, and review the history of scaling actions that were triggered.

Policies are per VM (one policy per VM). Bounds and cooldown prevent runaway growth; recent scale events show what actually fired.

When to use it​

  • To let a VM's resources flex automatically instead of manually resizing it under load
  • To cap how far a VM is allowed to scale (min/max vCPUs, min/max memory) so autoscaling can't run away
  • To check what scaling actions actually fired and when, via the recent scale events log
  • When a workload is bursty but still needs a hard ceiling for host capacity / quotas
  • After reviewing Optimizer recommendations, if you want ongoing auto adjust instead of one-shot apply

How to get there​

  • Route / id: /autoscale
  • Nav: Operations → Autoscale (sidebar, command palette, or desktop nav)

Operate from the console (UX)​

  1. Review the policy table — one row per VM: vCPU range, memory range, CPU scale-up/scale-down thresholds, and cooldown (seconds between actions).
  2. Create Policy — pick a VM (only VMs without an existing policy), set min/max vCPUs, min/max memory (MB), CPU scale-up and scale-down thresholds (%), and cooldown seconds.
  3. Delete a policy from its row (confirmation) to stop autoscaling that VM.
  4. Check Recent scale events — last 20 actions (VM, action, resource, timestamp).
  5. On a read-only account, create/delete controls are hidden and a read-only notice is shown.

Typical flow: set conservative max bounds → create policy → watch Recent scale events under load → tighten thresholds or cooldown if it flaps. Confirm host capacity on Capacity and Quotas.

Scale events only appear after real threshold crossings; idle VMs with a policy and no load will show an empty recent log.

Empty / fail: Error banner, empty table, or failed toast — confirm you are signed in, zyvor-fabricd is healthy (/readyz on http://127.0.0.1:<port> or https://<host>), and any backend this page needs (FluxVM, storage, network) is reachable. See Admin basics.

Success: Live data loads without error; creates/updates appear in the list or detail view and any confirmation toast clears cleanly.