Redaguoti

Bendrinti naudojant


Azure Chaos Studio fault and action library

This article lists the faults you can use in Chaos Studio, organized by the applicable resource type. To understand which role assignments are recommended for each resource type, see Supported resource types and role assignments for Azure Chaos Studio.

Agent-based faults

Agent-based faults are injected into Azure Virtual Machines or Virtual Machine Scale Set instances by installing the Chaos Studio Agent. Find the service-direct fault options for these resources below in the Virtual Machine and Virtual Machine Scale Set tables.

Applicable OS types Fault name Applicable scenarios
Windows, Linux CPU Pressure Compute capacity loss, resource pressure
Windows, Linux Kill Process Dependency disruption
Windows Pause Process Dependency disruption, service disruption
Windows1, Linux2 Network Disconnect Network disruption
Windows1, Linux2 Network Latency Network performance degradation
Windows1, Linux2 Network Packet Loss Network reliability issues
Windows, Linux2 Network Isolation Network disruption
Windows DNS Failure DNS resolution issues
Windows Network Disconnect (Via Firewall) Network disruption
Windows, Linux Physical Memory Pressure Memory capacity loss, resource pressure
Windows, Linux Stop Service Service disruption/restart
Windows Time Change Time synchronization issues
Windows Virtual Memory Pressure Memory capacity loss, resource pressure
Linux Arbitrary Stress-ng Stressor General system stress testing
Linux Linux DiskIO Pressure Disk I/O performance degradation
Windows DiskIO Pressure Disk I/O performance degradation

1 TCP/UDP packets only. 2 Outbound network traffic only.

App Service

This section applies to the Microsoft.Web/sites resource type. Learn more about App Service.

Fault name Applicable scenarios
Stop App Service Service disruption

Autoscale Settings

This section applies to the Microsoft.Insights/autoscaleSettings resource type. Learn more about Autoscale Settings.

Fault name Applicable scenarios
Disable Autoscale Compute capacity loss (when used with Virtual Machine Scale Set Shutdown)

Azure Kubernetes Service

This section applies to the Microsoft.ContainerService/managedClusters resource type. Learn more about Azure Kubernetes Service.

Fault name Applicable scenarios
AKS Chaos Mesh DNS Chaos DNS resolution issues
AKS Chaos Mesh HTTP Chaos Network disruption
AKS Chaos Mesh IO Chaos Disk degradation/pressure
AKS Chaos Mesh Kernel Chaos Kernel disruption
AKS Chaos Mesh Network Chaos Network disruption
AKS Chaos Mesh Pod Chaos Container disruption
AKS Chaos Mesh Stress Chaos System stress testing
AKS Chaos Mesh Time Chaos Time synchronization issues

Cloud Services (Classic)

This section applies to the Microsoft.ClassicCompute/domainNames resource type. Learn more about Cloud Services (Classic).

Fault name Applicable scenarios
Cloud Service Shutdown Compute loss

Clustered Cache for Redis

This section applies to the Microsoft.Cache/redis resource type. Learn more about Clustered Cache for Redis.

Fault name Applicable scenarios
Azure Cache for Redis (Reboot) Dependency disruption (caches)

Cosmos DB

This section applies to the Microsoft.DocumentDB/databaseAccounts resource type. Learn more about Cosmos DB.

Fault name Applicable scenarios
Cosmos DB Failover Database failover

Event Hubs

This section applies to the Microsoft.EventHub/namespaces resource type. Learn more about Event Hubs.

Fault name Applicable scenarios
Change Event Hub State Messaging infrastructure misconfiguration/disruption

Key Vault

This section applies to the Microsoft.KeyVault/vaults resource type. Learn more about Key Vault.

Fault name Applicable scenarios
Key Vault: Deny Access Certificate denial
Key Vault: Disable Certificate Certificate disruption
Key Vault: Increment Certificate Version Certificate version increment
Key Vault: Update Certificate Policy Certificate policy changes/misconfigurations

Network Security Groups

This section applies to the Microsoft.Network/networkSecurityGroups resource type. Learn more about network security groups.

Fault name Applicable scenarios
NSG Security Rule Network disruption (for many Azure services)

Service Bus

This section applies to the Microsoft.ServiceBus/namespaces resource type. Learn more about Service Bus.

Fault name Applicable scenarios
Change Queue State Messaging infrastructure misconfiguration/disruption
Change Subscription State Messaging infrastructure misconfiguration/disruption
Change Topic State Messaging infrastructure misconfiguration/disruption

Virtual Machines (service-direct)

This section applies to the Microsoft.Compute/virtualMachines resource type. Learn more about Virtual Machines.

Fault name Applicable scenarios
VM Redeploy Compute disruption, maintenance events
VM Shutdown Compute loss/disruption

Virtual Machine Scale Set

This section applies to the Microsoft.Compute/virtualMachineScaleSets resource type. Learn more about Virtual Machine Scale Sets.

Fault name Applicable scenarios
Virtual Machine Scale Set Shutdown Compute loss/disruption
Virtual Machine Scale Set Shutdown (2.0) Compute loss/disruption (by Availability Zone)

Orchestration actions

These actions are building blocks for constructing effective experiments. Use them in combination with other faults, such as running a load test while in parallel shutting down compute instances in a zone.

Action category Fault name
Load Start load test (Azure Load Testing)
Load Stop load test (Azure Load Testing)
Time delay Delay

Details: Agent-based faults

Network Disconnect

Property Value
Capability name NetworkDisconnect-1.1
Target type Microsoft-Agent
Supported OS types Windows, Linux (outbound traffic only)
Description Blocks network traffic for specified port range and network block. At least one destinationFilter or inboundDestinationFilter array must be provided.
Prerequisites Windows: The agent must run as administrator, which happens by default if installed as a VM extension.
Linux: The tc (Traffic Control) package is used for network faults. If it isn't already installed, the agent automatically attempts to install it from the default package manager.
Urn urn:csci:microsoft:agent:networkDisconnect/1.1
Fault type Continuous.
Parameters (key, value)
destinationFilters Delimited JSON array of packet filters defining which outbound packets to target. Maximum of 16.
inboundDestinationFilters Delimited JSON array of packet filters defining which inbound packets to target. Maximum of 16.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

The parameters destinationFilters and inboundDestinationFilters use the following array of packet filters.

Property Value
address IP address that indicates the start of the IP range.
subnetMask Subnet mask for the IP address range.
portLow (Optional) Port number of the start of the port range.
portHigh (Optional) Port number of the end of the port range.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:networkDisconnect/1.1",
      "parameters": [
        {
          "key": "destinationFilters",
          "value": "[ { \"address\": \"23.45.229.97\", \"subnetMask\": \"255.255.255.224\", \"portLow\": \"5000\", \"portHigh\": \"5200\" } ]"
        },
        {
          "key": "inboundDestinationFilters",
          "value": "[ { \"address\": \"23.45.229.97\", \"subnetMask\": \"255.255.255.224\", \"portLow\": \"5000\", \"portHigh\": \"5200\" } ]"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • The agent-based network faults currently only support IPv4 addresses.
  • The network disconnect fault only affects new connections. Existing active connections continue to persist. You can restart the service or process to force connections to break.
  • When running on Windows, the network disconnect fault currently only works with TCP or UDP packets.
  • When running on Linux, this fault can only affect outbound traffic, not inbound traffic. The fault can affect both inbound and outbound traffic on Windows environments (via the inboundDestinationFilters and destinationFilters parameters).

Network Disconnect (Via Firewall)

Property Value
Capability name NetworkDisconnectViaFirewall-1.0
Target type Microsoft-Agent
Supported OS types Windows
Description Applies a Windows Firewall rule to block outbound traffic for specified port range and network block.
Prerequisites Agent must run as administrator. If the agent is installed as a VM extension, it runs as administrator by default.
Urn urn:csci:microsoft:agent:networkDisconnectViaFirewall/1.0
Fault type Continuous.
Parameters (key, value)
destinationFilters Delimited JSON array of packet filters that define which outbound packets to target for fault injection.
address IP address that indicates the start of the IP range.
subnetMask Subnet mask for the IP address range.
portLow (Optional) Port number of the start of the port range.
portHigh (Optional) Port number of the end of the port range.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:networkDisconnectViaFirewall/1.0",
      "parameters": [
        {
          "key": "destinationFilters",
          "value": "[ { \"Address\": \"23.45.229.97\", \"SubnetMask\": \"255.255.255.224\", \"PortLow\": \"5000\", \"PortHigh\": \"5200\" } ]"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • The agent-based network faults currently only support IPv4 addresses.
  • This fault currently only affects new connections. Existing active connections are unaffected. You can restart the service or process to force connections to break.

Network Latency

Property Value
Capability name NetworkLatency-1.1
Target type Microsoft-Agent
Supported OS types Windows, Linux (outbound traffic only)
Description Increases network latency for a specified port range and network block. At least one destinationFilter or inboundDestinationFilter array must be provided.
Prerequisites Windows: The agent must run as administrator, which happens by default if installed as a VM extension.
Linux: The tc (Traffic Control) package is used for network faults. If it isn't already installed, the agent automatically attempts to install it from the default package manager.
Urn urn:csci:microsoft:agent:networkLatency/1.1
Fault type Continuous.
Parameters (key, value)
latencyInMilliseconds Amount of latency to be applied in milliseconds.
destinationFilters Delimited JSON array of packet filters defining which outbound packets to target. Maximum of 16.
inboundDestinationFilters Delimited JSON array of packet filters defining which inbound packets to target. Maximum of 16.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

The parameters destinationFilters and inboundDestinationFilters use the following array of packet filters.

Property Value
address IP address that indicates the start of the IP range.
subnetMask Subnet mask for the IP address range.
portLow (Optional) Port number of the start of the port range.
portHigh (Optional) Port number of the end of the port range.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:networkLatency/1.1",
      "parameters": [
        {
          "key": "destinationFilters",
          "value": "[ { \"address\": \"23.45.229.97\", \"subnetMask\": \"255.255.255.224\", \"portLow\": \"5000\", \"portHigh\": \"5200\" } ]"
        },
        {
          "key": "inboundDestinationFilters",
          "value": "[ { \"address\": \"23.45.229.97\", \"subnetMask\": \"255.255.255.224\", \"portLow\": \"5000\", \"portHigh\": \"5200\" } ]"
        },
        {
          "key": "latencyInMilliseconds",
          "value": "100",
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • The agent-based network faults currently only support IPv4 addresses.
  • When running on Linux, the network latency fault can only affect outbound traffic, not inbound traffic. The fault can affect both inbound and outbound traffic on Windows environments (via the inboundDestinationFilters and destinationFilters parameters).
  • When running on Windows, the network latency fault currently only works with TCP or UDP packets.
  • This fault currently only affects new connections. Existing active connections are unaffected. You can restart the service or process to force connections to break.

Network Packet Loss

Property Value
Capability name NetworkPacketLoss-1.0
Target type Microsoft-Agent
Supported OS types Windows, Linux (outbound traffic only)
Description Introduces packet loss for outbound traffic at a specified rate, between 0.0 (no packets lost) and 1.0 (all packets lost). This action can help simulate scenarios like network congestion or network hardware issues.
Prerequisites Windows: The agent must run as administrator, which happens by default if installed as a VM extension.
Linux: The tc (Traffic Control) package is used for network faults. If it isn't already installed, the agent automatically attempts to install it from the default package manager.
Urn urn:csci:microsoft:agent:networkPacketLoss/1.0
Fault type Continuous.
Parameters (key, value)
packetLossRate The rate at which packets matching the destination filters will be lost, ranging from 0.0 to 1.0.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.
destinationFilters Delimited JSON array of packet filters (parameters below) that define which outbound packets to target for fault injection. Maximum of three.
address IP address that indicates the start of the IP range.
subnetMask Subnet mask for the IP address range.
portLow (Optional) Port number of the start of the port range.
portHigh (Optional) Port number of the end of the port range.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:networkPacketLoss/1.0",
      "parameters": [
            {
                "key": "destinationFilters",
                "value": "[{\"address\":\"23.45.229.97\",\"subnetMask\":\"255.255.255.224\",\"portLow\":5000,\"portHigh\":5200}]"
            },
            {
                "key": "packetLossRate",
                "value": "0.5"
            },
            {
                "key": "virtualMachineScaleSetInstances",
                "value": "[0,1,2]"
            }
        ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • The agent-based network faults currently only support IPv4 addresses.
  • When running on Windows, the network packet loss fault currently only works with TCP or UDP packets.
  • When running on Linux, this fault can only affect outbound traffic, not inbound traffic. The fault can affect both inbound and outbound traffic on Windows environments (via the inboundDestinationFilters and destinationFilters parameters).
  • This fault currently only affects new connections. Existing active connections are unaffected. You can restart the service or process to force connections to break.

Network Isolation

Property Value
Capability name NetworkIsolation-1.0
Target type Microsoft-Agent
Supported OS types Windows, Linux (outbound only)
Description Fully isolate the virtual machine from network connections by dropping all IP-based inbound (on Windows) and outbound (on Windows and Linux) packets for the specified duration. At the end of the duration, network connections will be re-enabled. Because the agent depends on network traffic, this action cannot be canceled and will run to the specified duration.
Prerequisites Windows: The agent must run as administrator, which happens by default if installed as a VM extension.
Linux: The tc (Traffic Control) package is used for network faults. If it isn't already installed, the agent automatically attempts to install it from the default package manager.
Urn urn:csci:microsoft:agent:networkIsolation/1.0
Fault type Continuous.
Parameters (key, value)
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode, optional otherwise. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:networkIsolation/1.0",
      "parameters": [],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • Because the agent depends on network traffic, this action cannot be canceled and will run to the specified duration. Use with caution.
  • This fault currently only affects new connections. Existing active connections are unaffected. You can restart the service or process to force connections to break.
  • When running on Linux, this fault can only affect outbound traffic, not inbound traffic. The fault can affect both inbound and outbound traffic on Windows environments.

DNS Failure

Property Value
Capability name DnsFailure-1.0
Target type Microsoft-Agent
Supported OS types Windows
Description Substitutes DNS lookup request responses with a specified error code. DNS lookup requests that are substituted must:
  • Originate from the VM.
  • Match the defined fault parameters.
DNS lookups that aren't made by the Windows DNS client aren't affected by this fault.
Prerequisites None.
Urn urn:csci:microsoft:agent:dnsFailure/1.0
Fault type Continuous.
Parameters (key, value)
hosts Delimited JSON array of host names to fail DNS lookup request for.

This property accepts wildcards (*), but only for the first subdomain in an address and only applies to the subdomain for which they're specified. For example:
  • *.microsoft.com is supported.
  • subdomain.*.microsoft isn't supported.
  • *.microsoft.com doesn't work for multiple subdomains in an address, such as subdomain1.subdomain2.microsoft.com.
dnsFailureReturnCode DNS error code to be returned to the client for the lookup failure (FormErr, ServFail, NXDomain, NotImp, Refused, XDomain, YXRRSet, NXRRSet, NotAuth, NotZone). For more information on DNS return codes, see the IANA website.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:dnsFailure/1.0",
      "parameters": [
        {
          "key": "hosts",
          "value": "[ \"www.bing.com\", \"msdn.microsoft.com\" ]"
        },
        {
          "key": "dnsFailureReturnCode",
          "value": "ServFail"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • The DNS Failure fault requires Windows 2019 RS5 or newer.
  • DNS Cache is ignored during the duration of the fault for the host names defined in the fault.

CPU Pressure

Property Value
Capability name CPUPressure-1.0
Target type Microsoft-Agent
Supported OS types Windows, Linux
Description Adds CPU pressure, up to the specified value, on the VM where this fault is injected during the fault action. The artificial CPU pressure is removed at the end of the duration or if the experiment is canceled. On Windows, the % Processor Utility performance counter is used at fault start to determine current CPU percentage, which is subtracted from the pressureLevel defined in the fault so that % Processor Utility hits approximately the pressureLevel defined in the fault parameters.
Prerequisites Linux: The stress-ng utility needs to be installed. Installation happens automatically as part of agent installation, using the default package manager, on several operating systems including Debian-based (like Ubuntu), Red Hat Enterprise Linux, and OpenSUSE. For other distributions, including Azure Linux, you must install stress-ng manually. For more information, see the upstream project repository.
Windows: None.
Urn urn:csci:microsoft:agent:cpuPressure/1.0
Fault type Continuous.
Parameters (key, value)
pressureLevel An integer between 1 and 95 that indicates how much CPU pressure (%) is applied to the VM in terms of % CPU Usage
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:cpuPressure/1.0",
      "parameters": [
        {
          "key": "pressureLevel",
          "value": "95"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

Known issues on Linux:

  • The stress effect might not be terminated correctly if AzureChaosAgent is unexpectedly killed.

Physical Memory Pressure

Property Value
Capability name PhysicalMemoryPressure-1.0
Target type Microsoft-Agent
Supported OS types Windows, Linux
Description Adds physical memory pressure, up to the specified value, on the VM where this fault is injected during the fault action. The artificial physical memory pressure is removed at the end of the duration or if the experiment is canceled.
Prerequisites Linux: The stress-ng utility needs to be installed. Installation happens automatically as part of agent installation, using the default package manager, on several operating systems including Debian-based (like Ubuntu), Red Hat Enterprise Linux, and OpenSUSE. For other distributions, including Azure Linux, you must install stress-ng manually. For more information, see the upstream project repository.
Windows: None.
Urn urn:csci:microsoft:agent:physicalMemoryPressure/1.0
Fault type Continuous.
Parameters (key, value)
pressureLevel An integer between 1 and 95 that indicates how much physical memory pressure (%) is applied to the VM.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:physicalMemoryPressure/1.0",
      "parameters": [
        {
          "key": "pressureLevel",
          "value": "95"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

Currently, the Windows agent doesn't reduce memory pressure when other applications increase their memory usage. If the overall memory usage exceeds 100%, the Windows agent might crash.

Virtual Memory Pressure

Property Value
Capability name VirtualMemoryPressure-1.0
Target type Microsoft-Agent
Supported OS types Windows
Description Adds virtual memory pressure, up to the specified value, on the VM where this fault is injected during the fault action. The artificial virtual memory pressure is removed at the end of the duration or if the experiment is canceled.
Prerequisites None.
Urn urn:csci:microsoft:agent:virtualMemoryPressure/1.0
Fault type Continuous.
Parameters (key, value)
pressureLevel An integer between 1 and 95 that indicates how much physical memory pressure (%) is applied to the VM.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:virtualMemoryPressure/1.0",
      "parameters": [
        {
          "key": "pressureLevel",
          "value": "95"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Disk IO Pressure

Property Value
Capability name DiskIOPressure-1.1
Target type Microsoft-Agent
Supported OS types Windows
Description Uses the diskspd utility to add disk pressure to a Virtual Machine. Pressure is added to the primary disk by default, or the disk specified with the targetTempDirectory parameter. This fault has five different modes of execution. The artificial disk pressure is removed at the end of the duration or if the experiment is canceled.
Prerequisites None.
Urn urn:csci:microsoft:agent:diskIOPressure/1.1
Fault type Continuous.
Parameters (key, value)
pressureMode The preset mode of disk pressure to add to the primary storage of the VM. Must be one of the PressureModes in the following table.
targetTempDirectory (Optional) The directory to use for applying disk pressure. For example, D:/Temp. If the parameter is not included, pressure is added to the primary disk.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Pressure modes

PressureMode Description
PremiumStorageP10IOPS numberOfThreads = 1
randomBlockSizeInKB = 64
randomSeed = 10
numberOfIOperThread = 25
sizeOfBlocksInKB = 8
sizeOfWriteBufferInKB = 64
fileSizeInGB = 2
percentOfWriteActions = 50
PremiumStorageP10Throttling
numberOfThreads = 2
randomBlockSizeInKB = 64
randomSeed = 10
numberOfIOperThread = 25
sizeOfBlocksInKB = 64
sizeOfWriteBufferInKB = 64
fileSizeInGB = 1
percentOfWriteActions = 50
PremiumStorageP50IOPS numberOfThreads = 32
randomBlockSizeInKB = 64
randomSeed = 10
numberOfIOperThread = 32
sizeOfBlocksInKB = 8
sizeOfWriteBufferInKB = 64
fileSizeInGB = 1
percentOfWriteActions = 50
PremiumStorageP50Throttling numberOfThreads = 2
randomBlockSizeInKB = 1024
randomSeed = 10
numberOfIOperThread = 2
sizeOfBlocksInKB = 1024
sizeOfWriteBufferInKB = 1024
fileSizeInGB = 20
percentOfWriteActions = 50
Default numberOfThreads = 2
randomBlockSizeInKB = 64
randomSeed = 10
numberOfIOperThread = 2
sizeOfBlocksInKB = 64
sizeOfWriteBufferInKB = 64
fileSizeInGB = 1
percentOfWriteActions = 50

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:diskIOPressure/1.1",
      "parameters": [
        {
          "key": "pressureMode",
          "value": "PremiumStorageP10IOPS"
        },
        {
          "key": "targetTempDirectory",
          "value": "C:/temp/"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Linux Disk IO Pressure

Property Value
Capability name LinuxDiskIOPressure-1.1
Target type Microsoft-Agent
Supported OS types Linux
Description Uses stress-ng to apply pressure to the disk. One or more worker processes are spawned that perform I/O processes with temporary files. Pressure is added to the primary disk by default, or the disk specified with the targetTempDirectory parameter. For information on how pressure is applied, see the stress-ng article.
Prerequisites Linux: The stress-ng utility needs to be installed. Installation happens automatically as part of agent installation, using the default package manager, on several operating systems including Debian-based (like Ubuntu), Red Hat Enterprise Linux, and OpenSUSE. For other distributions, including Azure Linux, you must install stress-ng manually. For more information, see the upstream project repository.
Urn urn:csci:microsoft:agent:linuxDiskIOPressure/1.1
Fault type Continuous.
Parameters (key, value)
workerCount Number of worker processes to run. Setting workerCount to 0 generates as many worker processes as there are number of processors.
fileSizePerWorker Size of the temporary file that a worker performs I/O operations against. Integer plus a unit in bytes (b), kilobytes (k), megabytes (m), or gigabytes (g) (for example, 4m for 4 megabytes and 256g for 256 gigabytes).
blockSize Block size to be used for disk I/O operations, greater than 1 byte and less than 4 megabytes (maximum value is 4095k). Integer plus a unit in bytes, kilobytes, or megabytes (for example, 512k for 512 kilobytes).
targetTempDirectory (Optional) The directory to use for applying disk pressure. For example, /tmp/. If the parameter is not included, pressure is added to the primary disk.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

These sample values produced ~100% disk pressure when tested on a Standard_D2s_v3 virtual machine with Premium SSD LRS. A large fileSizePerWorker and smaller blockSize help stress the disk fully.

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:linuxDiskIOPressure/1.1",
      "parameters": [
        {
          "key": "workerCount",
          "value": "4"
        },
        {
          "key": "fileSizePerWorker",
          "value": "2g"
        },
        {
          "key": "blockSize",
          "value": "64k"
        },
        {
          "key": "targetTempDirectory",
          "value": "/tmp/"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Stop Service

Property Value
Capability name StopService-1.0
Target type Microsoft-Agent
Supported OS types Windows, Linux
Description Stops a Windows service or a Linux systemd service during the fault. Restarts it at the end of the duration or if the experiment is canceled.
Prerequisites None.
Urn urn:csci:microsoft:agent:stopService/1.0
Fault type Continuous.
Parameters (key, value)
serviceName Name of the Windows service or Linux systemd service you want to stop.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:stopService/1.0",
      "parameters": [
        {
          "key": "serviceName",
          "value": "nvagent"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • Windows: Display names for services aren't supported. Use sc.exe query in the command prompt to explore service names.
  • Linux: Other service types besides systemd, like sysvinit, aren't supported.

Kill Process

Property Value
Capability name KillProcess-1.0
Target type Microsoft-Agent
Supported OS types Windows, Linux
Description Kills all the running instances of a process that matches the process name sent in the fault parameters. Within the duration set for the fault action, a process is killed repetitively based on the value of the kill interval specified. This fault is a destructive fault where system admin would need to manually recover the process if self-healing is configured for it. Note that this fault will error when used on an empty name process, when used with an unspecified interval, or when we cannot find the target process name that we want to kill.
Prerequisites None.
Urn urn:csci:microsoft:agent:killProcess/1.0
Fault type Continuous.
Parameters (key, value)
processName Name of a process to continuously kill (without the .exe). The process does not need to be running when the fault begins executing.
killIntervalInMilliseconds Amount of time the fault waits in between successive kill attempts in milliseconds.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:killProcess/1.0",
      "parameters": [
        {
          "key": "processName",
          "value": "myapp"
        },
        {
          "key": "killIntervalInMilliseconds",
          "value": "1000"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Pause Process

Property Value
Capability name PauseProcess-1.0
Target type Microsoft-Agent
Supported OS types Windows
Description Pauses (suspends) the specified processes for the specified duration. If there are multiple processes with the same name, this fault suspends all of those processes. Within the fault's duration, the processes are paused repetitively at the specified interval. At the end of the duration or if the experiment is canceled, the processes will resume.
Prerequisites None.
Urn urn:csci:microsoft:agent:pauseProcess/1.0
Fault type Continuous.
Parameters (key, value)
processNames Delimited JSON array of process names defining which processes are to be paused. Maximum of 4. The process name can optionally include the ".exe" extension.
pauseIntervalInMilliseconds Amount of time the fault waits between successive pausing attempts, in milliseconds.
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:pauseProcess/1.0",
      "parameters": [
        {
          "key": "processNames",
          "value": "[ \"test-0\", \"test-1.exe\" ]"
        },
        {
          "key": "pauseIntervalInMilliseconds",
          "value": "1000"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Limitations

Currently, a maximum of 4 process names can be listed in the processNames parameter.

Time Change

Property Value
Capability name TimeChange-1.0
Target type Microsoft-Agent
Supported OS types Windows
Description Changes the system time of the virtual machine and resets the time at the end of the experiment or if the experiment is canceled.
Prerequisites None.
Urn urn:csci:microsoft:agent:timeChange/1.0
Fault type Continuous.
Parameters (key, value)
dateTime A DateTime string in ISO8601 format. If YYYY-MM-DD values are missing, they're defaulted to the current day when the experiment runs. If Thh:mm:ss values are missing, the default value is 12:00:00 AM. If a 2-digit year is provided (YY), it's converted to a 4-digit year (YYYY) based on the current century. If the timezone <Z> is missing, the default offset is the local timezone. <Z> must always include a sign symbol (negative or positive).
virtualMachineScaleSetInstances An array of instance IDs when you apply this fault to a virtual machine scale set. Required for virtual machine scale sets in uniform orchestration mode. Learn more about instance IDs.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:timeChange/1.0",
      "parameters": [
        {
          "key": "dateTime",
          "value": "2038-01-01T03:14:07"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Arbitrary Stress-ng Stressor

Property Value
Capability name StressNg-1.0
Target type Microsoft-Agent
Supported OS types Linux
Description Runs any stress-ng command by passing arguments directly to stress-ng. Useful when one of the predefined faults for stress-ng doesn't meet your needs.
Prerequisites Linux: The stress-ng utility needs to be installed. Installation happens automatically as part of agent installation, using the default package manager, on several operating systems including Debian-based (like Ubuntu), Red Hat Enterprise Linux, and OpenSUSE. For other distributions, including Azure Linux, you must install stress-ng manually. For more information, see the upstream project repository.
Urn urn:csci:microsoft:agent:stressNg/1.0
Fault type Continuous.
Parameters (key, value)
stressNgArguments One or more arguments to pass to the stress-ng process. For information on possible stress-ng arguments, see the stress-ng article. NOTE: Do NOT include the "-t " argument because it will cause an error. Experiment length is defined directly in the Azure chaos experiment UI, NOT in the stressNgArguments.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:agent:stressNg/1.0",
      "parameters": [
        {
          "key": "stressNgArguments",
          "value": "--random 64"
        },
        {
          "key": "virtualMachineScaleSetInstances",
          "value": "[0,1,2]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Details: Service-direct faults

Stop App Service

Property Value
Capability name Stop-1.0
Target type Microsoft-AppService
Description Stops the targeted App Service applications, then restarts them at the end of the fault duration. This action applies to resources of the "Microsoft.Web/sites" type, including App Service, API Apps, Mobile Apps, and Azure Functions.
Prerequisites None.
Urn urn:csci:microsoft:appService:stop/1.0
Fault type Continuous.
Parameters (key, value) None.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:appService:stop/1.0",
      "duration": "PT10M",
      "parameters":[],
      "selectorid": "myResources"
    }
  ]
}

Disable Autoscale

Property Value
Capability name DisableAutoscale
Target type Microsoft-AutoscaleSettings
Description Disables the autoscale service. When autoscale is disabled, resources such as virtual machine scale sets, web apps, service bus, and more aren't automatically added or removed based on the load of the application.
Prerequisites The autoScalesetting resource that's enabled on the resource must be onboarded to Chaos Studio.
Urn urn:csci:microsoft:autoscalesettings:disableAutoscale/1.0
Fault type Continuous.
Parameters (key, value)
enableOnComplete Boolean. Configures whether autoscaling is reenabled after the action is done. Default is true.

Sample JSON

{
  "name": "BranchOne", 
  "actions": [ 
    { 
    "type": "continuous", 
    "name": "urn:csci:microsoft:autoscaleSetting:disableAutoscale/1.0", 
    "parameters": [ 
     { 
      "key": "enableOnComplete", 
      "value": "true" 
      }                 
  ],                                 
   "duration": "PT2M", 
   "selectorId": "Selector1",           
  } 
 ] 
} 

AKS Chaos Mesh Network Chaos

Property Value
Capability name NetworkChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes a network fault available through Chaos Mesh to run against your Azure Kubernetes Service (AKS) cluster. Useful for re-creating AKS incidents that result from network outages, delays, duplications, loss, and corruption.
Prerequisites The AKS cluster must have Chaos Mesh deployed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:networkChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the NetworkChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:networkChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"action\":\"delay\",\"mode\":\"one\",\"selector\":{\"namespaces\":[\"default\"]},\"delay\":{\"latency\":\"200ms\",\"correlation\":\"100\",\"jitter\":\"0ms\"}}}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

AKS Chaos Mesh Pod Chaos

Property Value
Capability name PodChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes a pod fault available through Chaos Mesh to run against your AKS cluster. Useful for re-creating AKS incidents that are a result of pod failures or container issues.
Prerequisites The AKS cluster must have Chaos Mesh deployed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:podChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the PodChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:podChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"action\":\"pod-failure\",\"mode\":\"all\",\"selector\":{\"namespaces\":[\"default\"]}}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

AKS Chaos Mesh Stress Chaos

Property Value
Capability name StressChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes a stress fault available through Chaos Mesh to run against your AKS cluster. Useful for re-creating AKS incidents because of stresses over a collection of pods, for example, due to high CPU or memory consumption.
Prerequisites The AKS cluster must have Chaos Mesh deployed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:stressChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the StressChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:stressChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"mode\":\"one\",\"selector\":{\"namespaces\":[\"default\"]},\"stressors\":{\"cpu\":{\"workers\":1,\"load\":50},\"memory\":{\"workers\":4,\"size\":\"256MB\"}}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

AKS Chaos Mesh IO Chaos

Property Value
Capability name IOChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes an IO fault available through Chaos Mesh to run against your AKS cluster. Useful for re-creating AKS incidents because of IO delays and read/write failures when you use IO system calls such as open, read, and write.
Prerequisites The AKS cluster must have Chaos Mesh deployed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:IOChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the IOChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:IOChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"action\":\"latency\",\"mode\":\"one\",\"selector\":{\"app\":\"etcd\"},\"volumePath\":\"\/var\/run\/etcd\",\"path\":\"\/var\/run\/etcd\/**\/*\",\"delay\":\"100ms\",\"percent\":50}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

AKS Chaos Mesh Time Chaos

Property Value
Capability name TimeChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes a change in the system clock on your AKS cluster by using Chaos Mesh. Useful for re-creating AKS incidents that result from distributed systems falling out of sync, missing/incorrect leap year/leap second logic, and more.
Prerequisites The AKS cluster must have Chaos Mesh deployed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:timeChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the TimeChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:timeChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"mode\":\"one\",\"selector\":{\"namespaces\":[\"default\"]},\"timeOffset\":\"-10m100ns\"}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

AKS Chaos Mesh Kernel Chaos

Property Value
Capability name KernelChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes a kernel fault available through Chaos Mesh to run against your AKS cluster. Useful for re-creating AKS incidents because of Linux kernel-level errors, such as a mount failing or memory not being allocated.
Prerequisites The AKS cluster must have Chaos Mesh deployed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:kernelChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the KernelChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:kernelChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"mode\":\"one\",\"selector\":{\"namespaces\":[\"default\"]},\"failKernRequest\":{\"callchain\":[{\"funcname\":\"__x64_sys_mount\"}],\"failtype\":0}}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

AKS Chaos Mesh HTTP Chaos

Property Value
Capability name HTTPChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes an HTTP fault available through Chaos Mesh to run against your AKS cluster. Useful for re-creating incidents because of HTTP request and response processing failures, such as delayed or incorrect responses.
Prerequisites The AKS cluster must have Chaos Mesh deployed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:httpChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the HTTPChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:httpChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"mode\":\"all\",\"selector\":{\"namespaces\":[\"default\"]},\"target\":\"Request\",\"port\":80,\"method\":\"GET\",\"path\":\"/api\",\"abort\":true}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

AKS Chaos Mesh DNS Chaos

Property Value
Capability name DNSChaos-2.2
Target type Microsoft-AzureKubernetesServiceChaosMesh
Supported node pool OS types Linux
Description Causes a DNS fault available through Chaos Mesh to run against your AKS cluster. Useful for re-creating incidents because of DNS failures.
Prerequisites The AKS cluster must have Chaos Mesh deployed and the DNS service must be installed.
Urn urn:csci:microsoft:azureKubernetesServiceChaosMesh:dnsChaos/2.2
Parameters (key, value)
jsonSpec A JSON-formatted Chaos Mesh spec that uses the DNSChaos kind. You can use a YAML-to-JSON converter like Convert YAML To JSON to convert the Chaos Mesh YAML to JSON and minify it. Use single-quotes within the JSON or escape the quotes with a backslash character. Only include the YAML under the jsonSpec property. Don't include information like metadata and kind. Specifying duration within the jsonSpec isn't necessary, but it's used if available.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:azureKubernetesServiceChaosMesh:dnsChaos/2.2",
      "parameters": [
        {
            "key": "jsonSpec",
            "value": "{\"action\":\"random\",\"mode\":\"all\",\"patterns\":[\"google.com\",\"chaos-mesh.*\",\"github.?om\"],\"selector\":{\"namespaces\":[\"default\"]}}"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

Cloud Services (Classic) Shutdown

Property Value
Capability name Shutdown-1.0
Target type Microsoft-DomainName
Description Stops a deployment during the fault. Restarts the deployment at the end of the fault duration or if the experiment is canceled.
Prerequisites None.
Urn urn:csci:microsoft:domainName:shutdown/1.0
Fault type Continuous.
Parameters None.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:domainName:shutdown/1.0",
      "parameters": [],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Azure Cache for Redis (Reboot)

Property Value
Capability name Reboot-1.0
Target type Microsoft-AzureClusteredCacheForRedis
Description Causes a forced reboot operation to occur on the target to simulate a brief outage.
Prerequisites N/A
Urn urn:csci:microsoft:azureClusteredCacheForRedis:reboot/1.0
Fault type Discrete.
Parameters (key, value)
rebootType The node types where the reboot action is to be performed, which can be specified as PrimaryNode, SecondaryNode, or AllNodes.
shardId The ID of the shard to be rebooted. Only relevant for Premium tier caches.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:azureClusteredCacheForRedis:reboot/1.0",
      "parameters": [
        {
          "key": "RebootType",
          "value": "AllNodes"
        },
        {
          "key": "ShardId",
          "value": "0"
        }
      ],
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • The reboot fault causes a forced reboot to better simulate an outage event, which means there's the potential for data loss to occur.
  • The reboot fault is a discrete fault type. Unlike continuous faults, it's a one-time action and has no duration.

Cosmos DB Failover

Property Value
Capability name Failover-1.0
Target type Microsoft-CosmosDB
Description Causes an Azure Cosmos DB account with a single write region to fail over to a specified read region to simulate a write region outage.
Prerequisites None.
Urn urn:csci:microsoft:cosmosDB:failover/1.0
Fault type Continuous.
Parameters (key, value)
readRegion The read region that should be promoted to write region during the failover, for example, East US 2.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:cosmosDB:failover/1.0",
      "parameters": [
        {
          "key": "readRegion",
          "value": "West US 2"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Change Event Hub State

Property Value
Capability name ChangeEventHubState-1.0
Target type Microsoft-EventHub
Description Sets individual event hubs to the desired state within an Azure Event Hubs namespace. You can affect specific event hub names or use “*” to affect all within the namespace. This action can help test your messaging infrastructure for maintenance or failure scenarios. This is a discrete fault, so the entity will not be returned to the starting state automatically.
Prerequisites An Azure Event Hubs namespace with at least one event hub entity.
Urn urn:csci:microsoft:eventHub:changeEventHubState/1.0
Fault type Discrete.
Parameters (key, value)
desiredState The desired state for the targeted event hubs. The possible states are Active, Disabled, and SendDisabled.
eventHubs A comma-separated list of the event hub names within the targeted namespace. Use "*" to affect all entities within the namespace.

Sample JSON

{
  "name": "Branch1",
    "actions": [
        {
            "selectorId": "Selector1",
            "type": "discrete",
            "parameters": [
                {
                    "key": "eventhubs",
                    "value": "[\"*\"]"
                },
                {
                    "key": "desiredState",
                    "value": "Disabled"
                }
            ],
            "name": "urn:csci:microsoft:eventHub:changeEventHubState/1.0"
        }
    ]
}

Key Vault: Deny Access

Property Value
Capability name DenyAccess-1.0
Target type Microsoft-KeyVault
Description Blocks all network access to a key vault by temporarily modifying the key vault network rules. This action prevents an application dependent on the key vault from accessing secrets, keys, and/or certificates. If the key vault allows access to all networks, this setting is changed to only allow access from selected networks. No virtual networks are in the allowed list at the start of the fault. All networks are allowed access at the end of the fault duration. If the key vault is set to only allow access from selected networks, any virtual networks in the allowed list are removed at the start of the fault. They're restored at the end of the fault duration.
Prerequisites The target key vault can't have any firewall rules and must not be set to allow Azure services to bypass the firewall. If the target key vault is set to only allow access from selected networks, there must be at least one virtual network rule. The key vault can't be in recover mode.
Urn urn:csci:microsoft:keyVault:denyAccess/1.0
Fault type Continuous.
Parameters (key, value) None.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:keyvault:denyAccess/1.0",
      "parameters": [],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Key Vault: Disable Certificate

Property Value
Capability name DisableCertificate-1.0
Target type Microsoft-KeyVault
Description By using certificate properties, the fault disables the certificate for a specific duration (provided by the user). It enables the certificate after this fault duration.
Prerequisites None.
Urn urn:csci:microsoft:keyvault:disableCertificate/1.0
Fault type Continuous.
Parameters (key, value)
certificateName Name of Azure Key Vault certificate on which the fault is executed.
version Certificate version that should be disabled. If not specified, the latest version is disabled.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:keyvault:disableCertificate/1.0",
      "parameters": [
        {
            "key": "certificateName",
            "value": "<name of AKV certificate>"
        },
        {
            "key": "version",
            "value": "<certificate version>"
        }

],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Key Vault: Increment Certificate Version

Property Value
Capability name IncrementCertificateVersion-1.0
Target type Microsoft-KeyVault
Description Generates a new certificate version and thumbprint by using the Key Vault Certificate client library. Current working certificate is upgraded to this version. Certificate version is not reverted after the fault duration.
Prerequisites None.
Urn urn:csci:microsoft:keyvault:incrementCertificateVersion/1.0
Fault type Discrete.
Parameters (key, value)
certificateName Name of Azure Key Vault certificate on which the fault is executed.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:keyvault:incrementCertificateVersion/1.0",
      "parameters": [
        {
            "key": "certificateName",
            "value": "<name of AKV certificate>"
        }
    ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Key Vault: Update Certificate Policy

Property Value
Capability name UpdateCertificatePolicy-1.0
Target type Microsoft-KeyVault
Description Certificate policies (for example, certificate validity period, certificate type, key size, or key type) are updated based on user input and reverted after the fault duration.
Prerequisites None.
Urn urn:csci:microsoft:keyvault:updateCertificatePolicy/1.0
Fault type Continuous.
Parameters (key, value)
certificateName Name of Azure Key Vault certificate on which the fault is executed.
version Certificate version that should be updated. If not specified, the latest version is updated.
enabled Boolean. Value that indicates if the new certificate version is enabled.
validityInMonths Validity period of the certificate in months.
certificateTransparency Indicates whether the certificate should be published to the certificate transparency list when created.
certificateType Certificate type.
contentType Content type of the certificate. For example, it's Pkcs12 when the certificate contains raw PFX bytes or Pem when it contains ASCII PEM-encoded bytes. Pkcs12 is the default value assumed.
keySize Size of the RSA key: 2048, 3072, or 4096.
exportable Boolean. Value that indicates if the certificate key is exportable from the vault or secure certificate store.
reuseKey Boolean. Value that indicates if the certificate key should be reused when the certificate is rotated.
keyType Type of backing key generated when new certificates are issued, such as RSA or EC.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:keyvault:updateCertificatePolicy/1.0",
      "parameters": [
        {
            "key": "certificateName",
            "value": "<name of AKV certificate>"
        },
        {
            "key": "version",
            "value": "<certificate version>"
        },
        {
            "key": "enabled",
            "value": "True"
        },
        {
            "key": "validityInMonths",
            "value": "12"
        },
        {
            "key": "certificateTransparency",
            "value": "True"
        },
        {
            "key": "certificateType",
            "value": "<certificate type>"
        },
        {
            "key": "contentType",
            "value": "Pem"
        },
        {
            "key": "keySize",
            "value": "4096"
        },
                {
            "key": "exportable",
            "value": "True"
        },
        {
            "key": "reuseKey",
            "value": "False"
        },
        {
            "key": "keyType",
            "value": "RSA"
        }

     ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

NSG Security Rule

Property Value
Capability name SecurityRule-1.0, SecurityRule-1.1
Target type Microsoft-NetworkSecurityGroup
Description Enables manipulation or rule creation in an existing Azure network security group (NSG) or set of Azure NSGs, assuming the rule definition is applicable across security groups. Useful for:
  • Simulating an outage of a downstream or cross-region dependency/nondependency.
  • Simulating an event that's expected to trigger a logic to force a service failover.
  • Simulating an event that's expected to trigger an action from a monitoring or state management service.
  • Using as an alternative for blocking or allowing network traffic where Chaos Agent can't be deployed.
Prerequisites None.
Urn urn:csci:microsoft:networkSecurityGroup:securityRule/1.0, urn:csci:microsoft:networkSecurityGroup:securityRule/1.1
Fault type Continuous.
Parameters (key, value)
name A unique name for the security rule that's created. The fault fails if another rule already exists on the NSG with the same name. Must begin with a letter or number. Must end with a letter, number, or underscore. May contain only letters, numbers, underscores, periods, or hyphens.
protocol Protocol for the security rule. Must be Any, TCP, UDP, or ICMP.
sourceAddresses A string that represents a JSON-delimited array of CIDR-formatted IP addresses. Can also be a service tag name for an inbound rule, for example, AppService. An asterisk * can also be used to match all source IPs.
destinationAddresses A string that represents a JSON-delimited array of CIDR-formatted IP addresses. Can also be a service tag name for an outbound rule, for example, AppService. An asterisk * can also be used to match all destination IPs.
action Security group access type. Must be either Allow or Deny.
destinationPortRanges A string that represents a JSON-delimited array of single ports and/or port ranges, such as 80 or 1024-65535.
sourcePortRanges A string that represents a JSON-delimited array of single ports and/or port ranges, such as 80 or 1024-65535.
priority A value between 100 and 4096 that's unique for all security rules within the NSG. The fault fails if another rule already exists on the NSG with the same priority.
direction Direction of the traffic affected by the security rule. Must be either Inbound or Outbound.

Sample JSON

{ 
  "name": "branchOne", 
  "actions": [ 
    { 
      "type": "continuous", 
      "name": "urn:csci:microsoft:networkSecurityGroup:securityRule/1.0", 
      "parameters": [ 
          { 
              "key": "name", 
              "value": "Block_SingleHost_to_Networks" 

          }, 
          { 
              "key": "protocol", 
              "value": "Any" 
          }, 
          { 
              "key": "sourceAddresses", 
              "value": "[\"10.1.1.128/32\"]"
          }, 
          { 
              "key": "destinationAddresses", 
              "value": "[\"10.20.0.0/16\",\"10.30.0.0/16\"]"
          }, 
          { 
              "key": "access", 
              "value": "Deny" 
          }, 
          { 
              "key": "destinationPortRanges", 
              "value": "[\"80-8080\"]"
          }, 
          { 
              "key": "sourcePortRanges", 
              "value": "[\"*\"]"
          }, 
          { 
              "key": "priority", 
              "value": "100" 
          }, 
          { 
              "key": "direction", 
              "value": "Outbound" 
          } 
      ], 
      "duration": "PT10M", 
      "selectorid": "myResources" 
    } 
  ] 
} 

Limitations

  • The fault can only be applied to an existing NSG.
  • When an NSG rule that's intended to deny traffic is applied, existing connections won't be broken until they've been idle for 4 minutes. One workaround is to add another branch in the same step that uses a fault that would cause existing connections to break when the NSG fault is applied. For example, killing the process, temporarily stopping the service, or restarting the VM would cause connections to reset.
  • Rules are applied at the start of the action. Any external changes to the rule during the duration of the action cause the experiment to fail.
  • Creating or modifying Application Security Group rules isn't supported.
  • Priority values must be unique on each NSG targeted. Attempting to create a new rule that has the same priority value as another causes the experiment to fail.
  • The NSG Security Rule version 1.1 fault supports an additional flushConnection parameter. This functionality has an active known issue: if flushConnection is enabled, the fault may result in a "FlushingNetworkSecurityGroupConnectionIsNotEnabled" error. To avoid this error temporarily, disable the flushConnection parameter or use the NSG Security Rule version 1.0 fault.

Service Bus: Change Queue State

Property Value
Capability name ChangeQueueState-1.0
Target type Microsoft-ServiceBus
Description Sets Queue entities within a Service Bus namespace to the desired state. You can affect specific entity names or use “*” to affect all. This action can help test your messaging infrastructure for maintenance or failure scenarios. This is a discrete fault, so the entity will not be returned to the starting state automatically.
Prerequisites A Service Bus namespace with at least one Queue entity.
Urn urn:csci:microsoft:serviceBus:changeQueueState/1.0
Fault type Discrete.
Parameters (key, value)
desiredState The desired state for the targeted queues. The possible states are Active, Disabled, SendDisabled, and ReceiveDisabled.
queues A comma-separated list of the queue names within the targeted namespace. Use "*" to affect all queues within the namespace.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:serviceBus:changeQueueState/1.0",
      "parameters":[
          {
            "key": "desiredState",
            "value": "Disabled"
          },
          {
            "key": "queues",
            "value": "samplequeue1,samplequeue2"
          }
      ],
      "selectorid": "myServiceBusSelector"
    }
  ]
}

Limitations

  • A maximum of 1000 queue entities can be passed to this fault.

Service Bus: Change Subscription State

Property Value
Capability name ChangeSubscriptionState-1.0
Target type Microsoft-ServiceBus
Description Sets Subscription entities within a Service Bus namespace and Topic to the desired state. You can affect specific entity names or use “*” to affect all. This action can help test your messaging infrastructure for maintenance or failure scenarios. This is a discrete fault, so the entity will not be returned to the starting state automatically.
Prerequisites A Service Bus namespace with at least one Subscription entity.
Urn urn:csci:microsoft:serviceBus:changeSubscriptionState/1.0
Fault type Discrete.
Parameters (key, value)
desiredState The desired state for the targeted subscriptions. The possible states are Active and Disabled.
topic The parent topic containing one or more subscriptions to affect.
subscriptions A comma-separated list of the subscription names within the targeted namespace. Use "*" to affect all subscriptions within the namespace.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:serviceBus:changeSubscriptionState/1.0",
      "parameters":[
          {
            "key": "desiredState",
            "value": "Disabled"
          },
          {
            "key": "topic",
            "value": "topic01"
          },
          {
            "key": "subscriptions",
            "value": "*"
          }
      ],
      "selectorid": "myServiceBusSelector"
    }
  ]
}

Limitations

  • A maximum of 1000 subscription entities can be passed to this fault.

Service Bus: Change Topic State

Property Value
Capability name ChangeTopicState-1.0
Target type Microsoft-ServiceBus
Description Sets the specified Topic entities within a Service Bus namespace to the desired state. You can affect specific entity names or use “*” to affect all. This action can help test your messaging infrastructure for maintenance or failure scenarios. This is a discrete fault, so the entity will not be returned to the starting state automatically.
Prerequisites A Service Bus namespace with at least one Topic entity.
Urn urn:csci:microsoft:serviceBus:changeTopicState/1.0
Fault type Discrete.
Parameters (key, value)
desiredState The desired state for the targeted topics. The possible states are Active and Disabled.
topics A comma-separated list of the topic names within the targeted namespace. Use "*" to affect all topics within the namespace.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:serviceBus:changeTopicState/1.0",
      "parameters":[
          {
            "key": "desiredState",
            "value": "Disabled"
          },
          {
            "key": "topics",
            "value": "*"
          }
      ],
      "selectorid": "myServiceBusSelector"
    }
  ]
}

Limitations

  • A maximum of 1000 topic entities can be passed to this fault.

VM Redeploy

Property Value
Capability name Redeploy-1.0
Target type Microsoft-VirtualMachine
Description Redeploys a VM by shutting it down, moving it to a new node in the Azure infrastructure, and powering it back on. This helps validate your workload's resilience to maintenance events.
Prerequisites None.
Urn urn:csci:microsoft:virtualMachine:redeploy/1.0
Fault type Discrete.
Parameters (key, value) None.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:virtualMachine:redeploy/1.0",
      "parameters":[],
      "selectorid": "myResources"
    }
  ]
}

Limitations

  • The Virtual Machine Redeploy operation is throttled within an interval of 10 hours. If your experiment fails with a "Too many redeploy requests" error, wait for 10 hours to retry the experiment.

Virtual Machine Shutdown

Property Value
Capability name Shutdown-1.0
Target type Microsoft-VirtualMachine
Supported OS types Windows, Linux.
Description Shuts down a VM for the duration of the fault. Restarts it at the end of the experiment or if the experiment is canceled. Only Azure Resource Manager VMs are supported.
Prerequisites None.
Urn urn:csci:microsoft:virtualMachine:shutdown/1.0
Fault type Continuous.
Parameters (key, value)
abruptShutdown (Optional) Boolean that indicates if the VM should be shut down gracefully or abruptly (destructive).

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:virtualMachine:shutdown/1.0",
      "parameters": [
        {
          "key": "abruptShutdown",
          "value": "false"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Virtual Machine Scale Set Shutdown

This fault has two available versions that you can use, Version 1.0 and Version 2.0. The main difference is that Version 2.0 allows you to filter by availability zones, only shutting down instances within a specified zone or zones.

Virtual Machine Scale Set Shutdown Version 1.0

Property Value
Capability name Version 1.0
Target type Microsoft-VirtualMachineScaleSet
Supported OS types Windows, Linux.
Description Shuts down or kills a virtual machine scale set instance during the fault and restarts the VM at the end of the fault duration or if the experiment is canceled.
Prerequisites None.
Urn urn:csci:microsoft:virtualMachineScaleSet:shutdown/1.0
Fault type Continuous.
Parameters (key, value)
abruptShutdown (Optional) Boolean that indicates if the virtual machine scale set instance should be shut down gracefully or abruptly (destructive).
instances A string that's a delimited array of virtual machine scale set instance IDs to which the fault is applied.
Version 1.0 sample JSON
{
  "name": "branchOne",
  "actions": [
    {
      "type": "continuous",
      "name": "urn:csci:microsoft:virtualMachineScaleSet:shutdown/1.0",
      "parameters": [
        {
          "key": "abruptShutdown",
          "value": "true"
        },
        {
          "key": "instances",
          "value": "[\"1\",\"3\"]"
        }
      ],
      "duration": "PT10M",
      "selectorid": "myResources"
    }
  ]
}

Virtual Machine Scale Set Shutdown Version 2.0

Property Value
Capability name Shutdown-2.0
Target type Microsoft-VirtualMachineScaleSet
Supported OS types Windows, Linux.
Description Shuts down or kills a virtual machine scale set instance during the fault. Restarts the VM at the end of the fault duration or if the experiment is canceled. Supports dynamic targeting.
Prerequisites None.
Urn urn:csci:microsoft:virtualMachineScaleSet:shutdown/2.0
Fault type Continuous.
filter (Optional) Available starting with Version 2.0. Used to filter the list of targets in a selector. Currently supports filtering on a list of zones. The filter is only applied to virtual machine scale set resources within a zone:
  • If no filter is specified, this fault shuts down all instances in the virtual machine scale set.
  • The experiment targets all virtual machine scale set instances in the specified zones.
  • If a filter results in no targets, the experiment fails.
Parameters (key, value)
abruptShutdown (Optional) Boolean that indicates if the virtual machine scale set instance should be shut down gracefully or abruptly (destructive).
Version 2.0 sample JSON snippets

The following snippets show how to configure both dynamic filtering and the shutdown 2.0 fault.

Configure a filter for dynamic targeting:

{
  "type": "List",
  "id": "myResources",
  "targets": [
    {
      "id": "<targetResourceId>",
      "type": "ChaosTarget"
    }
  ],
  "filter": {
    "type": "Simple",
    "parameters": {
      "zones": [
        "1"
      ]
    }
  }
}

Configure the shutdown fault:

{
  "name": "branchOne",
  "actions": [
    {
      "name": "urn:csci:microsoft:virtualMachineScaleSet:shutdown/2.0",
      "type": "continuous",
      "selectorId": "myResources",
      "duration": "PT10M",
      "parameters": [
        {
          "key": "abruptShutdown",
          "value": "false"
        }
      ]
    }
  ]
}

Limitations

Currently, only virtual machine scale sets configured with the Uniform orchestration mode are supported. If your virtual machine scale set uses Flexible orchestration, you can use the Azure Resource Manager virtual machine shutdown fault to shut down selected instances.

Details: Orchestration actions

Delay

Property Value
Fault provider N/A
Supported OS types N/A
Description Adds a time delay before, between, or after other experiment actions. This action isn't a fault and is used to synchronize actions within an experiment. Use this action to wait for the impact of a fault to appear in a service, or wait for an activity outside of the experiment to complete. For example, your experiment could wait for autohealing to occur before injecting another fault.
Prerequisites N/A
Urn urn:csci:microsoft:chaosStudio:timedDelay/1.0
Duration The duration of the delay in ISO 8601 format (for example, PT10M).

Sample JSON

{
  "name": "branchOne",
  "actions": [ 
    {
      "type": "delay",
      "name": "urn:csci:microsoft:chaosStudio:timedDelay/1.0",
      "duration": "PT10M"
    }
  ] 
}

Start Load Test (Azure Load Testing)

Property Value
Capability name Start-1.0
Target type Microsoft-AzureLoadTest
Description Starts a load test (from Azure Load Testing) based on the provided load test ID.
Prerequisites A load test with a valid load test ID must be created in the Azure Load Testing service.
Urn urn:csci:microsoft:azureLoadTest:start/1.0
Fault type Discrete.
Parameters (key, value)
testID The ID of a specific load test created in the Azure Load Testing service.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:azureLoadTest:start/1.0",
      "parameters": [
        {
            "key": "testID",
            "value": "0"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}

Stop Load Test (Azure Load Testing)

Property Value
Capability name Stop-1.0
Target type Microsoft-AzureLoadTest
Description Stops a load test (from Azure Load Testing) based on the provided load test ID.
Prerequisites A load test with a valid load test ID must be created in the Azure Load Testing service.
Urn urn:csci:microsoft:azureLoadTest:stop/1.0
Fault type Discrete.
Parameters (key, value)
testID The ID of a specific load test created in the Azure Load Testing service.

Sample JSON

{
  "name": "branchOne",
  "actions": [
    {
      "type": "discrete",
      "name": "urn:csci:microsoft:azureLoadTest:stop/1.0",
      "parameters": [
        {
            "key": "testID",
            "value": "0"
        }
    ],
      "selectorid": "myResources"
    }
  ]
}