Monday, 22 July 2024

Create a chaos experiment that uses an agent-based fault with the Azure portal

 

Create a chaos experiment that uses an agent-based fault with the Azure portal

You can use these same steps to set up and run an experiment for any agent-based fault. An agent-based fault requires setup and installation of the chaos agent. A service-direct fault runs directly against an Azure resource without any need for instrumentation.

Prerequisites

Enable Chaos Studio on your virtual machine

Chaos Studio can't inject faults against a VM unless that VM was added to Chaos Studio first. To add a VM to Chaos Studio, create a target and capabilities on the resource. Then you install the chaos agent.

Virtual machines have two target types. One target type enables service-direct faults (where no agent is required). Another target type enables agent-based faults (which requires the installation of an agent). The chaos agent is an application installed on your VM as a VM extension. You use it to inject faults in the guest operating system.

Enable the chaos target, capabilities, and agent

 Important

Prior to finishing the next steps, you must create a user-assigned managed identity. Then you assign it to the target VM or virtual machine scale set.

  1. Open the Azure portal.

  2. Search for Chaos Studio in the search bar.

  3. Select Targets and move to your VM.

    Screenshot that shows the Targets view in the Azure portal.

  4. Select the checkbox next to your VM and select Enable targets. Then select Enable agent-based targets from the dropdown menu.

    Screenshot that shows enabling targets in the Azure portal.

  5. Select the Managed Identity to use to authenticate the chaos agent and optionally enable Application Insights to see experiment events and agent logs.

    Screenshot that shows selecting a managed identity.

  6. Select Review + Enable > Enable.

    Screenshot that shows reviewing agent-based target enablement.

  7. After a few minutes, a notification appears that indicates that the resources selected were successfully enabled. The Azure portal adds the user-assigned identity to the VM. The portal enables the agent target and capabilities and installs the chaos agent as a VM extension.

    Screenshot that shows a notification that shows the target was successfully enabled.

  8. If you're enabling a virtual machine scale set, upgrade instances to the latest model by going to the virtual machine scale set resource pane. Select Instances, and then select all instances. Select Upgrade if you're not on the latest model.

You've now successfully added your Linux VM to Chaos Studio. In the Targets view, you can also manage the capabilities enabled on this resource. Select the Manage actions link next to a resource to display the capabilities enabled for that resource.

Create an experiment

Now you can create your experiment. A chaos experiment defines the actions you want to take against target resources. The actions are organized and run in sequential steps. The chaos experiment also defines the actions you want to take against branches, which run in parallel.

  1. Select the Experiments tab in Chaos Studio. In this view, you can see and manage all your chaos experiments. Select Create > New experiment.

    Screenshot that shows the Experiments view in the Azure portal.

  2. Fill in the SubscriptionResource Group, and Location where you want to deploy the chaos experiment. Give your experiment a name. Select Next: Experiment designer.

    Screenshot that shows adding basic experiment details.

  3. You're now in the Chaos Studio experiment designer. You can build your experiment by adding steps, branches, and faults. Give a friendly name to your Step and Branch. Then select Add action > Add fault.

    Screenshot that shows the experiment designer.

  4. Select CPU Pressure from the dropdown list. Fill in Duration with the number of minutes to apply pressure. Fill in pressureLevel with the % of CPU utilization pressure that you want to apply. Leave virtualMachineScaleSetInstances blank. Select Next: Target resources.

    Screenshot that shows fault properties.

  5. Select your VM and select Next.

    Screenshot that shows adding a target.

  6. Verify that your experiment looks correct. Then select Review + create > Create.

    Screenshot that shows reviewing and creating the experiment.

Give the experiment permission to your virtual machine

When you create a chaos experiment, Chaos Studio creates a system-assigned managed identity that executes faults against your target resources. This identity must be given appropriate permissions to the target resource for the experiment to run successfully.

  1. Go to your VM and select Access control (IAM).

    Screenshot that shows the virtual machine Overview page.

  2. Select Add > Add role assignment.

    Screenshot that shows Access control overview.

  3. Search for Reader and select the role. Select Next.

    Screenshot that shows assigning the virtual machine Contributor role.

  4. Choose Select members and search for your experiment name. Select your experiment and choose Select. If there are multiple experiments in the same tenant with the same name, your experiment name is truncated with random characters added.

    Screenshot that shows adding the experiment to a role.

  5. Select Review + assign > Review + assign.

Run your experiment

You're now ready to run your experiment. To see the impact, we recommend that you open an Azure Monitor metrics chart with your VM's CPU pressure in a separate browser tab.

  1. In the Experiments view, select your experiment. Select Start > OK.

    Screenshot that shows starting the experiment.

  2. After the Status changes to Running, under History, select Details for the latest run to see details for the running experiment.

No comments:

Post a Comment