Auto Scaling
Machines that follow the load, and replace themselves when they fail.
At a glance
- Group size
- 0 to 100 machines
- Zones
- Spread over the zones you choose
- Policies
- Target tracking, step and scheduled
What it does
- Launch templates with versions
- A group launches the default, the latest or a fixed version of a template, and a rolling instance refresh moves running machines to a new one.
- Scaling policies
- Track a CPU or network target, scale in steps on any monitoring alarm, or change the size on a cron schedule.
- Health and load balancers
- Unhealthy machines are replaced, by machine status or by load-balancer health, and new machines are registered with the group’s load balancers.
- When a type is sold out
- Give an ordered list of instance types and zones to fall back to, so a sold-out type does not stop a scale-out.
Command line
Create a group and keep its CPU at 60 %.
$ dcloud autoscaling groups create --name web --region-id de1 --launch-template-id <template> \
--min-size 2 --max-size 10 --availability-zone-ids de1a --availability-zone-ids de1b
$ dcloud autoscaling policies put-target-tracking <group> --name cpu60 \
--metric cpu_percent --target-value 60 Price
Groups, templates and policies have no charge. The machines a group launches are billed as machines, at the prices on the price list.
Documentation
Related: Compute, Load balancing, Monitoring.