ECS task scale

Last updated on Jun 5, 2026

ECS task scale is an AWS chaos fault that calls UpdateService on one or more ECS services to set their desired count to a chaos value (REPLICAS) for the duration of the fault, then restores the original desired count when the fault ends. The fault works with both EC2-backed and Fargate services and triggers normal ECS deployment behaviour (tasks are added or stopped to match the new desired count).

Use this fault to test how a service behaves when it suddenly scales up (cold-start storms, throttled downstream dependencies, license consumption) or scales down (request loss, latency spikes, downstream connection pool churn), and to validate that autoscaling alarms and dashboards respond as expected.

Run your first experiment

If you have not configured the chaos infrastructure yet, go to Quickstart to install the chaos infrastructure and run an experiment end to end.

Use cases

Run this fault when you want to answer concrete questions like:

Scale-down impact: When the service is scaled below its normal desired count, do remaining tasks absorb the traffic and stay within latency SLOs?
Scale-up cold start: When the service suddenly doubles in size, do new tasks start, register with the load balancer, and warm up within the deployment health check window?
Downstream connection pools: Does a scale-up cause connection-pool exhaustion in a downstream database or cache, and does the system shed load gracefully?
Autoscaling alarms: Do CloudWatch alarms (CPU, memory, request count) fire and unmuted within the expected SLA when the service deliberately undersize?
Capacity provider behaviour: For ASG-backed capacity providers, does the cluster scale out to meet the new desired count, and is the time to ready reasonable?

Prerequisites

Kubernetes version: 1.21 or later for the chaos infrastructure cluster. Go to What's supported to confirm distribution support.
Target ECS services: Every service named in SERVICE_NAMES exists in CLUSTER_NAME within REGION and has a non-zero desired count at the start of the experiment.
AWS credentials available: Either an AWS credentials file uploaded as a File Secret in Harness Secret Manager (see Authentication below) or an IAM role for service accounts (IRSA) bound to the chaos infrastructure service account.
IAM permissions granted: The credentials or role include the permissions listed below.
Capacity to scale up (when REPLICAS is larger than current): The cluster has free CPU and memory to host the additional tasks, or the capacity provider is allowed to scale.

Supported environments

Platform	Support status
Amazon ECS on EC2 launch type	Supported
Amazon ECS on Fargate launch type	Supported
Services with capacity-provider strategies	Supported
Services with deployment circuit breaker	Supported (a failed deployment rolls back automatically)
AWS regions	Supported in every commercial region; pass the region in `REGION`

Permissions required

The IAM principal that the chaos pod uses (the credentials mounted from the Harness Secret Manager file secret, the IRSA role on the chaos service account, or the role assumed via ASSUME_ROLE_ARN) needs the following AWS actions.

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "ecs:DescribeClusters",
        "ecs:DescribeServices",
        "ecs:UpdateService",
        "ecs:ListTasks",
        "ecs:DescribeTasks"
      ],
      "Resource": "*"
    }
  ]
}

ecs:DescribeServices is used to read the current desired count before the fault.
ecs:UpdateService drives the fault.
ecs:ListTasks and ecs:DescribeTasks are used by the verification step to confirm the new desired count is reached.

Go to common policy for all AWS faults to use a single superset IAM policy across every AWS fault.

Authentication

The fault supports three credential delivery models. Pick one based on how your chaos infrastructure is deployed.

Method	When to use it	How to configure
Harness Secret Manager file secret	Chaos infrastructure runs outside EKS, or you want explicit static credentials	Upload the AWS credentials file as a File Secret in Harness Secret Manager and reference its identifier via `AWS_AUTHENTICATION_SECRET`
IAM Roles for Service Accounts (IRSA)	Chaos infrastructure runs in EKS and uses an OIDC-bound service account	No tunable changes; the chaos pod inherits the role automatically. Go to AWS IAM integration to set it up
Assume role	The fault needs to act in a different account or with elevated permissions	Set `ASSUME_ROLE_ARN` to the role ARN; the chaos pod assumes the role on top of its base credentials

When using the Harness Secret Manager method, the contents of the File Secret should be the AWS credentials file in the standard ~/.aws/credentials format:

[default]
aws_access_key_id = REPLACE_WITH_ACCESS_KEY_ID
aws_secret_access_key = REPLACE_WITH_SECRET_ACCESS_KEY

Upload this file as a File Secret in Harness Secret Manager (Project Setup → Secrets → New File Secret), and pass the secret identifier in AWS_AUTHENTICATION_SECRET when configuring the fault.

Go to AWS named profile for chaos to switch between profiles inside a single credentials file.

Fault tunables

Configure the following fault parameters when you add ECS task scale to an experiment in Chaos Studio. Defaults are shown for reference.

Required parameters

Tunable	Description	Default
`CLUSTER_NAME`	Name of the target ECS cluster.	(required)
`SERVICE_NAMES`	Comma-separated list of ECS service names to scale. Every named service is set to `REPLICAS`.	(required)
`REGION`	AWS region that hosts the ECS cluster (for example `us-east-1`).	(required)
`REPLICAS`	Desired count to set for each target service during the fault.	(required)

Chaos parameters

Tunable	Description	Default
`TOTAL_CHAOS_DURATION`	Duration of the fault in seconds. The service is held at `REPLICAS` for this period, then restored to its original count.	`60`
`SEQUENCE`	Order in which multiple services are scaled: `parallel` issues all `UpdateService` calls at once; `serial` issues them one at a time.	`parallel`
`RAMP_TIME`	Wait period in seconds before and after the fault. Go to ramp time to read how it is applied.	`0`

Authentication

Tunable	Description	Default
`ASSUME_ROLE_ARN`	ARN of an IAM role to assume on top of the base credentials. Leave empty to use the base credentials directly.	`""`
`AWS_AUTHENTICATION_SECRET`	Identifier of the File Secret in Harness Secret Manager that contains the AWS credentials file. Not required when using IRSA.	`""`

Tunables that apply to every fault are documented in common tunables for all faults. AWS-specific shared tunables are documented in common AWS fault tunables.

Fault execution in brief

Reads the current desired count of every service in SERVICE_NAMES, calls UpdateService to set the desired count to REPLICAS, waits for TOTAL_CHAOS_DURATION seconds, then calls UpdateService again to restore each service's original desired count.

Expected behavior during fault execution

When REPLICAS is lower than the current desired count, the ECS service controller stops current - REPLICAS tasks (oldest first by default).
When REPLICAS is higher than the current desired count, the controller calls RunTask to launch REPLICAS - current additional tasks; if there is insufficient capacity, the deployment stalls until capacity becomes available or the deployment circuit breaker fires.
The load balancer target group is updated as tasks register and deregister.
CloudWatch metrics such as RunningTaskCount and (for ALB-backed services) HealthyHostCount reflect the new size.
Autoscaling alarms (if configured) may trigger; this fault overrides the autoscaling controller only for the duration of the experiment.

When the fault ends

The chaos pod calls UpdateService again with the original desired count. The service controller drains or launches tasks until that count is restored. Autoscaling resumes acting on the restored baseline.

Signals to watch

Attach resilience probes to assert each layer:

Application availability: Use an HTTP probe to detect downtime when scaling down, and to measure cold-start latency when scaling up.
Running task count: Use a Prometheus probe on aws_ecs_running_task_count to confirm the target count is reached.
Latency: Use a Prometheus probe on a percentile latency metric (for example aws_applicationelb_target_response_time) to detect SLO breaches during scale changes.
Downstream dependencies: Use a Prometheus probe on database connection counts or queue depths to detect scale-induced fan-out problems.

Verify the fault execution effect

While the experiment is running, confirm the service is at the chaos size and then restored:

Inspect service desired vs running count.

aws ecs describe-services \
  --cluster <cluster> \
  --services <service> \
  --region <region> \
  --query "services[].[serviceName,desiredCount,runningCount,pendingCount]"

During the fault desiredCount should equal REPLICAS. After recovery it should return to the pre-fault value.

List running tasks.

aws ecs list-tasks \
  --cluster <cluster> \
  --service-name <service> \
  --desired-status RUNNING \
  --region <region>

The number of returned ARNs should match the current desired count once the deployment stabilizes.

Watch the service event stream.
```
aws ecs describe-services \
  --cluster <cluster> \
  --services <service> \
  --region <region> \
  --query "services[].events[0:10]"
```
Look for "service ... has started ... tasks" and "service ... has stopped ... tasks" events confirming the scale change.

Recovery and cleanup

End of duration: The chaos pod calls UpdateService with the original desired count. The ECS service controller restores the count on its own schedule.
Abort the experiment: Stopping the experiment from Chaos Studio also triggers the restore call.
Manual recovery: If the fault exits before the restore call (for example because of an IAM failure on the restore), set the desired count manually with aws ecs update-service --desired-count <original>.
Autoscaling reconciliation: Application Auto Scaling may take a few minutes to converge on the steady-state target after the fault ends; this is normal.

Limitations

Single chaos size per experiment: The fault sets every service in SERVICE_NAMES to the same REPLICAS value. Use multiple experiments if different services need different chaos sizes.
Autoscaling override window: While the fault is active, the desired count is held; autoscaling alarms may fire but cannot scale the service until the fault ends.
Deployment circuit breaker can roll back: If the new desired count cannot be reached because of capacity or task-start failures, the deployment circuit breaker rolls the service back, which the fault treats as a failure.
No per-task targeting: This fault changes service capacity, not individual task identity. Use ECS task stop when you need to kill specific replicas.
Cross-region targeting: A single experiment targets one region (the value of REGION).

Troubleshooting

ECS task scale experiment fails with AccessDeniedException in Harness Chaos Engineering

The credentials supplied to the chaos pod do not have the required ECS permissions. Confirm the IAM policy attached to the user, role, or IRSA service account includes ecs:DescribeServices and ecs:UpdateService for the target cluster. When using ASSUME_ROLE_ARN, also confirm the source identity is trusted to assume the role.

ECS task scale leaves runningCount below REPLICAS at the end of the chaos window

The service could not reach the target desired count within the chaos window. The most common causes are: insufficient cluster capacity (CPU or memory) to schedule more tasks; the capacity provider could not scale the ASG fast enough; task placement constraints (distinctInstance, attribute, instanceType) cannot be satisfied; or container start failures are tripping the deployment circuit breaker. Inspect 'aws ecs describe-services' events and check the ASG scaling activity.

Autoscaling fights with ECS task scale

Application Auto Scaling continues to compute target capacity based on alarms while the fault is active. The fault holds the desired count via UpdateService, which overrides autoscaling for that window; autoscaling resumes when the fault ends. If you need to suppress autoscaling during the experiment, suspend the scalable target before the fault and resume after.

ECS task scale does not restore the original desired count

The fault stores the pre-fault desired count and calls UpdateService with that value at the end. If the restore fails (for example because of an IAM permission failure on the restore call only), the service is left at REPLICAS. Set the desired count manually with 'aws ecs update-service' and check the chaos pod logs for the cause.

ECS task stop: Stop specific tasks instead of changing service capacity.
ECS instance stop: Reduce capacity at the container-instance level.
ECS update container resource limit: Reduce CPU/memory limits to simulate resource starvation.
Common AWS fault tunables: Shared environment variables for AWS faults.

Use cases​

Prerequisites​

Supported environments​

Permissions required​

Authentication​

Fault tunables​

Fault execution in brief​

Expected behavior during fault execution​

Signals to watch​

Verify the fault execution effect​

Recovery and cleanup​

Limitations​

Troubleshooting​

Related faults​