Scaling
Configure manual replicas, Basic autoscaling, and Advanced scaling rules.
Autoscaling adjusts how many container replicas your app runs based on demand. Pick one of three modes.
Paid Plans Only
Autoscaling requires a paid plan. Free plans will see an upgrade prompt with all options disabled.
None
Autoscaling off. Manage replica count manually from the Containers page.
Basic
Set a minimum and maximum replica count (1 to 10). Galaxy scales automatically targeting 70% CPU utilization.
Drag the slider handles to set your range, then click Save Configuration when the unsaved changes bar appears. Click Discard to revert.
Advanced
Advanced mode replaces a single range with a list of named rules you control. Galaxy evaluates them on a schedule and adjusts your replica count to match. Reach for it when one CPU target isn't enough: different metrics, time-based scaling, or several conditions at once. Professional plan only.
App-wide settings
These apply to every rule:
Replicas: The floor and ceiling for your app. Drag the two slider handles to set your minimum and maximum, up to your plan's container limit. No rule scales you below the floor or above the ceiling.
Polling interval: How often Galaxy evaluates your metrics, in seconds. Defaults to 300. Lower it for faster reactions; raise it to smooth out noisy traffic. Accepts 10 to 3600.
Replicas per step: How many replicas Galaxy adds or removes at a time. Defaults to 1. Raise it to scale in larger jumps.
Scaling rules
A rule describes one scaling intent. Add up to 20, give each a clear name, and switch them on or off individually. Advanced mode needs at least one rule; if you don't want any, switch back to None.
Two ways to create one:
- Templates: Click Templates for a menu of ready-made rules (High CPU defense, Business hours boost, Off-hours baseline, Traffic surge, Memory pressure). Picking one fills the editor so you can adjust before saving. It doesn't save on its own.
- Add rule: Click Add rule to start from a blank editor.
Every rule is one of two types:
Scales when a metric crosses a threshold. Example: scale up when CPU goes above 70%.
Metric: Choose cpu, memory, requests, or connections. Thresholds use the average across replicas.
CPU and memory: Average usage as a percentage of each container's resource limit. For example, replicas using 90% and 30% CPU average 60%, so they don't exceed a 70% threshold.
Requests: The app's total request rate divided by the number of replicas.
Connections: Open HTTP and WebSocket connections divided by the number of active replicas.
Comparator and threshold: Set whether the metric should be above or below your number. A "below" condition only works as a gate alongside at least one "above" condition, so it can't drive scaling on its own.
Match all or any: Add more than one condition, then choose whether they all must be true (all) or just one (any).
Schedule (optional): Restrict the rule to specific days and hours. Leave it off to run around the clock. Outside its window, the rule has no effect; other enabled rules and your app-wide minimum still apply.
Holds a minimum replica count during a time window, with no metric involved. Example: keep at least 3 replicas running on weekdays from 8:00 to 18:00.
Schedule: Pick the days and hours the baseline applies. Use the presets (Weekdays, Weekends, All) or set them by hand.
Baseline replicas: The floor to hold during the window. Outside the window the rule does nothing; inside it, Galaxy keeps at least this many replicas running.
Schedules display in your local time and are stored in UTC, so they fire at the right moment for everyone on your team.
As you edit a rule, Galaxy shows a plain-language summary of what it does, like "Scale up to 8 replicas when CPU is above 70% on weekdays 8:00 to 18:00." If two rules could conflict (same metric and overlapping schedule), you'll see a warning so you can review or remove the duplicate.
When everything looks right, click Save Configuration.