Windows EC2 process kill

Last updated on Jun 5, 2026

Windows EC2 process kill is an AWS chaos fault that kills one or more processes (selected by PROCESS_IDS PIDs or PROCESS_NAMES names, optionally with FORCE=enable for an immediate hard kill) on one or more Windows EC2 instances for a configurable duration. The fault repeatedly kills matching processes during TOTAL_CHAOS_DURATION so that respawns are also targeted; at the end of the fault, killing stops and the host's service manager (or supervisor) is left to restore the process as it normally would.

Use this fault to test how a Windows-hosted workload behaves when its backing process dies: whether the service auto-restarts cleanly (Windows service recovery), whether dependent processes detect and respond to the loss, and whether monitoring detects the process disappearance quickly.

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:

Service recovery: When a Windows service is killed, does Service Control Manager restart it within the configured recovery delay?
Watchdog behaviour: Does a custom watchdog or supervisor (NSSM, FireDaemon, custom) respawn the process?
Cluster failover: For clustered Windows applications (SQL Server AlwaysOn, MSCS), does a peer take over correctly when the local process dies?
Monitoring SLA: Does the process-down alert fire within the SLA, and does the runbook accurately describe the recovery?
Multi-process coordination: For applications composed of multiple processes that share state via named pipes or shared memory, do peers detect and recover from the loss?

Prerequisites

Kubernetes version: 1.21 or later for the chaos infrastructure cluster. Go to What's supported to confirm distribution support.
Target Windows EC2 instances: Each instance is in the running state and registered with AWS Systems Manager.
Selection criteria: Either EC2_INSTANCE_ID is set or EC2_INSTANCE_TAG is set.
Process selection: Either PROCESS_IDS or PROCESS_NAMES is set; if both are empty, the fault has nothing to kill.
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.

Supported environments

Platform	Support status
Windows Server 2016 / 2019 / 2022 EC2	Supported
Linux EC2 instances	Not supported (use EC2 process kill)
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": [
        "ec2:DescribeInstances",
        "ec2:DescribeInstanceStatus"
      ],
      "Resource": "*"
    },
    {
      "Effect": "Allow",
      "Action": [
        "ssm:SendCommand",
        "ssm:CancelCommand",
        "ssm:GetCommandInvocation",
        "ssm:DescribeInstanceInformation"
      ],
      "Resource": "*"
    },
    {
      "Effect": "Allow",
      "Action": [
        "ec2messages:AcknowledgeMessage",
        "ec2messages:DeleteMessage",
        "ec2messages:FailMessage",
        "ec2messages:GetEndpoint",
        "ec2messages:GetMessages",
        "ec2messages:SendReply"
      ],
      "Resource": "*"
    }
  ]
}

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 Windows EC2 process kill to an experiment in Chaos Studio. Defaults are shown for reference.

Required parameters

Tunable	Description	Default
`REGION`	AWS region that hosts the target instances (for example `us-east-1`).	(required)
`INSTANCE_AFFECTED_PERC`	Percentage of matching instances (by tag) to affect. `0` corresponds to one instance.	`0`

Targeting parameters

Tunable	Description	Default
`EC2_INSTANCE_ID`	Comma-separated list of EC2 instance IDs to affect.	`""`
`EC2_INSTANCE_TAG`	Tag in `Key:Value` form matching candidate instances when `EC2_INSTANCE_ID` is empty.	`""`
`PROCESS_IDS`	Comma-separated list of process IDs (PIDs) to kill on each target host.	`""`
`PROCESS_NAMES`	Comma-separated list of process executable names (for example `w3wp.exe`, `sqlservr`) to kill on each target host.	`""`

Chaos parameters

Tunable	Description	Default
`FORCE`	`enable` to kill processes immediately (equivalent to `Stop-Process -Force`); `disable` to signal termination and allow graceful shutdown.	`disable`
`TOTAL_CHAOS_DURATION`	Duration of the fault in seconds. The fault re-issues the kill command periodically so that respawned processes are killed too.	`60`
`INSTALL_DEPENDENCIES`	Install supporting tooling on the Windows host if missing.	`true`
`SEQUENCE`	Order in which multiple instances are affected: `parallel` kills on all selected instances at once; `serial` does so 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

Resolves the target instance list from EC2_INSTANCE_ID (or by EC2_INSTANCE_TAG), dispatches a PowerShell kill command via AWS Systems Manager Run Command to each selected Windows instance for the processes named in PROCESS_IDS and PROCESS_NAMES, and continues killing matching processes for TOTAL_CHAOS_DURATION seconds.

Expected behavior during fault execution

Matching processes terminate immediately (or after Stop-Process signals when FORCE=disable).
Windows services configured for automatic restart re-launch; custom supervisors respawn the process. The fault kills the respawned processes too.
Dependent processes that maintain pipes, sockets, or shared memory with the killed process see disconnects.
CloudWatch process metrics (if collected by the CW Agent) show the gap.

When the fault ends

The chaos pod stops killing processes. Service Control Manager (or your supervisor) is allowed to restore the process and keep it running; no further intervention is needed unless the application requires manual recovery.

Signals to watch

Attach resilience probes to assert each layer:

Process presence: Use a command probe that runs Get-Process -Name <name> and asserts the absence (during the fault) and presence (after recovery).
Service status: Use a command probe that runs Get-Service -Name <name> and asserts on Status.
Application errors: Use a Prometheus probe on application or dependent-service error counters.
Cluster failover: For clustered workloads, monitor cluster role ownership.

Verify the fault execution effect

While the experiment is running, confirm the process is killed and recovers:

Check process presence from the host (via SSM Session Manager).
```
Get-Process -Name <process-name> -ErrorAction SilentlyContinue
```
During the fault the command should return nothing (or transient entries from respawn attempts). After recovery the process should be back.
Check Windows service state.
```
Get-Service -Name <service-name>
```
For a service backing the killed process, Status should oscillate between Running and Stopped during the fault.

Inspect SSM command status.

aws ssm list-command-invocations --region <region> --details --filters "key=Status,value=InProgress"

Recovery and cleanup

End of duration: The chaos pod stops killing processes; the host's service manager restores the process.
Abort the experiment: Stopping the experiment from Chaos Studio cancels the SSM command.
Manual recovery: If a critical process does not auto-recover (no Windows service, no supervisor), start it manually with Start-Service or by launching the executable.
Workload recovery: Long-lived connections that depended on the killed process are not re-established by this fault; clients must reconnect.

Limitations

Windows-only.
SSM-managed hosts only.
Auto-restart dependency: The fault does not restart the process; it relies on the host's service manager or supervisor. Processes without an auto-restart configuration stay dead.
No protection against kernel processes: Killing critical Windows system processes (csrss.exe, wininit.exe) will likely BSOD the instance.
Cross-region targeting: A single experiment targets one region (the value of REGION).

Troubleshooting

Windows EC2 process kill fails with AccessDeniedException in Harness Chaos Engineering

The credentials supplied to the chaos pod do not have the required EC2 or SSM permissions. Confirm the IAM policy attached to the user, role, or IRSA service account includes ec2:DescribeInstances, ssm:SendCommand, ssm:GetCommandInvocation, and the ec2messages:* actions used by SSM.

Process keeps running despite the kill command

The most common causes are: the process is protected by Windows (such as a critical system process); the supervisor respawns it faster than the kill cadence (set FORCE=enable for an immediate hard kill); or PROCESS_NAMES does not match the process executable name (Get-Process is case-insensitive but expects the bare name without .exe). Verify the exact process name with 'Get-Process | Select Name, Id' on the host.

Instance BSOD or becomes unresponsive after the chaos window

The most common cause is killing a Windows system process or a critical driver. Avoid PROCESS_NAMES like 'csrss', 'wininit', 'lsass', 'services'. Use a Windows test instance, list the safe-to-kill processes explicitly in PROCESS_NAMES (for example w3wp.exe for IIS application pools), and validate the experiment scope before running in production.

Service does not auto-restart after the chaos window

Windows services restart on failure only if the recovery options are configured. In Services.msc (or via 'sc.exe failure'), set the first/second/third failures to 'Restart the Service' on the affected service. Without recovery configured, the service stays Stopped after this fault ends.

Windows EC2 CPU hog: Stress CPU on Windows instances without killing processes.
Windows EC2 memory hog: Stress memory on Windows instances.
EC2 process kill: Kill processes on Linux EC2 instances.
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​