Auto Scaling — AWS Auto Scaling

AWS docs: https://docs.aws.amazon.com/autoscaling/ec2/APIReference/Welcome.html

Auto Scaling groups really converge. A single background reconciler launches and terminates EC2 instances through the emulator’s own EC2 service until each group’s owned instance set matches its DesiredCapacity, advances the lifecycle state machine, honours lifecycle hooks, replaces unhealthy instances, and records a scaling activity for everything it does.

Summary

Group, launch-configuration, scaling-policy, lifecycle-hook and tag CRUD, plus SetDesiredCapacity, ExecutePolicy, TerminateInstanceInAutoScalingGroup, CompleteLifecycleAction, RecordLifecycleActionHeartbeat, SetInstanceHealth, SetInstanceProtection, DescribeAutoScalingInstances and DescribeScalingActivities.

Reconciliation

A group backed by a launch configuration converges within about a second of any change:

• CreateAutoScalingGroup, UpdateAutoScalingGroup, SetDesiredCapacity, ExecutePolicy and an alarm-driven policy all move DesiredCapacity; the reconciler then calls EC2 RunInstances / TerminateInstances to close the gap, clamped to the group’s MinSize and MaxSize.
• Instances move Pending → InService and Terminating → gone. With a matching lifecycle hook they park in Pending:Wait / Terminating:Wait first.
• DescribeAutoScalingGroups and DescribeAutoScalingInstances report the real instance set, with LifecycleState, HealthStatus, AvailabilityZone, InstanceType and ProtectedFromScaleIn.
• Instances are placed round-robin across AvailabilityZones and, when set, across VPCZoneIdentifier’s subnets. Tags marked PropagateAtLaunch are applied to each launched instance, alongside aws:autoscaling:groupName.
• Every launch and termination writes a DescribeScalingActivities entry with AWS’s StatusCode, Progress, Description and Cause wording. Activities are capped at the 200 most recent per group.
• A launch EC2 refuses (a missing AMI, for instance) is recorded as a Failed activity carrying EC2’s own error, never silently dropped.

Scaling policies

SimpleScaling and StepScaling execute for real. PutScalingPolicy accepts ChangeInCapacity, ExactCapacity and PercentChangeInCapacity, and honours MinAdjustmentMagnitude and the group’s cooldown.

Attaching a policy ARN to a CloudWatch alarm’s AlarmActions works end to end: when the alarm transitions, Auto Scaling executes the policy. For a step-scaling policy, the breach is measured as the alarm’s most recent datapoint minus its threshold, which is the number AWS compares against MetricIntervalLowerBound / MetricIntervalUpperBound.

Lifecycle hooks

PutLifecycleHook really pauses the transition it watches. The instance parks in Pending:Wait or Terminating:Wait, an EC2 Instance-launch Lifecycle Action (or …-terminate…) event is published to the default EventBridge bus with the LifecycleActionToken, and the instance stays there until CompleteLifecycleAction is called or HeartbeatTimeout elapses and the hook’s DefaultResult is applied. RecordLifecycleActionHeartbeat extends the window.

Not implemented — refused, not ignored

Per the fidelity-risk rule, a configuration the reconciler cannot converge is refused at the configuring operation rather than stored and quietly ignored:

Configuration	Operation	Result
LaunchTemplate	CreateAutoScalingGroup / UpdateAutoScalingGroup	501 — Overcast’s EC2 has no CreateLaunchTemplate to resolve ImageId/InstanceType from. Use a launch configuration. Tracked in #518.
MixedInstancesPolicy	CreateAutoScalingGroup / UpdateAutoScalingGroup	501 — there is no instance-type fleet or spot allocation to distribute over.
InstanceId (create from a running instance)	CreateAutoScalingGroup	501 — launch parameters cannot be derived from a running instance.
No launch source at all	CreateAutoScalingGroup	400 ValidationError, as on AWS.
PolicyType=TargetTrackingScaling	PutScalingPolicy	501 — there is no controller tracking a target metric value.
PolicyType=PredictiveScaling	PutScalingPolicy	501 — there is no forecasting model over historical metrics.
Warm pools, instance refresh, scheduled actions, notification configurations	—	Not registered; a protocol-correct 501.

Known divergences

• Termination policy. Real Auto Scaling’s Default policy balances across Availability Zones and then prefers the oldest launch configuration; Overcast implements OldestInstance (skipping instances protected from scale-in), which gives the same answer for a single-AZ or uniform group. A TerminationPolicies value is stored and echoed but does not change the choice.
• HealthCheckType=ELB is accepted and echoed, but load-balancer target health is not a health source — only SetInstanceHealth marks an instance unhealthy. HealthCheckGracePeriod is stored and echoed.
• One hook per transition. AWS runs every hook watching a transition; Overcast parks the instance on the first matching hook by name.
• State is not partitioned by region. Auto Scaling groups are global to the emulator, and the reconciler launches into the configured default region.
• Uses the Query protocol (form-encoded POST) with API version 2011-01-01.

Summary

Category	✅ Supported
Operations	25

---

Endpoints

Operations

Operation	Status	Notes	AWS Docs
CreateAutoScalingGroup	✅ Supported	Reconciles: launches EC2 instances to DesiredCapacity. LaunchTemplate, MixedInstancesPolicy and InstanceId are refused with 501 — Overcast’s EC2 has no launch templates
UpdateAutoScalingGroup	✅ Supported	Min/max/desired changes re-converge the group; desired outside min/max is AWS’s ValidationError
DescribeAutoScalingGroups	✅ Supported	Reports the real converged instance set with LifecycleState, HealthStatus, AvailabilityZone and ProtectedFromScaleIn
DeleteAutoScalingGroup	✅ Supported	ResourceInUse while instances remain unless ForceDelete; ForceDelete terminates them
SetDesiredCapacity	✅ Supported	Launches or terminates instances to converge; out-of-range values return AWS’s ValidationError
TerminateInstanceInAutoScalingGroup	✅ Supported	Terminates the EC2 instance and records the scaling activity; honours ShouldDecrementDesiredCapacity
CreateLaunchConfiguration	✅ Supported	ImageId/InstanceType/SecurityGroups drive RunInstances at launch time
DescribeLaunchConfigurations	✅ Supported
DeleteLaunchConfiguration	✅ Supported	ResourceInUse while a group still references it
PutScalingPolicy	✅ Supported	SimpleScaling and StepScaling execute for real; TargetTrackingScaling and PredictiveScaling are refused with 501
DescribePolicies	✅ Supported
DeletePolicy	✅ Supported
ExecutePolicy	✅ Supported	Applies the policy’s adjustment to DesiredCapacity, honouring cooldown and MinAdjustmentMagnitude
PutLifecycleHook	✅ Supported	Really pauses launch/terminate in Pending:Wait / Terminating:Wait and emits the EventBridge lifecycle-action event
DescribeLifecycleHooks	✅ Supported
DeleteLifecycleHook	✅ Supported	Releases any instance parked on the hook
CompleteLifecycleAction	✅ Supported	CONTINUE moves the instance on; ABANDON terminates it
RecordLifecycleActionHeartbeat	✅ Supported	Extends the hook’s heartbeat window
CreateOrUpdateTags	✅ Supported	PropagateAtLaunch tags are applied to launched EC2 instances
DeleteTags	✅ Supported
DescribeTags	✅ Supported	Filters: auto-scaling-group, key, propagate-at-launch, value; an unimplemented filter name is refused, not ignored
DescribeAutoScalingInstances	✅ Supported	Reports the real instance set the reconciler owns
DescribeScalingActivities	✅ Supported	One activity per launch and termination, with AWS’s StatusCode and Cause wording
SetInstanceHealth	✅ Supported	Unhealthy instances are terminated and replaced by the reconciler
SetInstanceProtection	✅ Supported	Protected instances are excluded from scale-in