Chaos faults for Azure

Last updated on Jun 22, 2026

---

Introduction

Azure faults disrupt resources that run on Microsoft Azure: Virtual Machines (standalone or VMSS), Azure Kubernetes Service (AKS) node pools, managed disks, App Service web apps, and Service Bus queues. Each fault calls the Azure Resource Manager API to inject the disruption, then reverses it cleanly at the end of the configured duration. Go to Azure authentication methods to set up service principals, workload identity, or managed identity, and to Azure fault permissions to grant the right RBAC roles.

Azure AKS node down

Deallocate a percentage of AKS worker VMs (selected by node pool and zone) for a configurable duration, then start them again.

AKSnode downkubernetes

Azure disk loss

Detach one or more managed data disks from their attached VMs for a configurable duration, then reattach them on the same LUN.

disk loss

Azure instance CPU hog

Drive CPU utilization to a configurable target on one or more Azure VMs for a configurable duration via the VM run-command extension.

CPU hog

Azure instance IO stress

Drive sustained disk read/write IO on one or more Azure VMs for a configurable duration via the VM run-command extension.

I/O stress

Azure instance memory hog

Consume a configurable amount of memory on one or more Azure VMs for a configurable duration via the VM run-command extension.

memory hog

Azure instance stop

Stop one or more Azure VMs (or VMSS instances) for a configurable duration, then start them again.

azure

Page 1 of 2

Azure AKS node down

Back to top

Azure AKS node down resolves the VMSS instances backing an AKS cluster (filtered by TARGET_NODE_POOL_NAMES and TARGET_ZONES), picks NODE_AFFECTED_PERCENTAGE of them, deallocates them for TOTAL_CHAOS_DURATION seconds, then starts them again.

Use cases

• Test that the scheduler reschedules pods inside the SLA when a percentage of AKS nodes disappear.
• Validate PodDisruptionBudgets across a node-down window.
• Test cluster autoscaler reaction time when pods become unschedulable.
• Simulate zone-level failures by targeting a single zone.

View details

Azure disk loss

Back to top

Azure disk loss detaches one or more managed data disks (VIRTUAL_DISK_NAMES) from their attached VMs for TOTAL_CHAOS_DURATION seconds, then reattaches them on the same LUN. OS disks are excluded by design.

Use cases

• Test how stateful workloads (databases, file servers) handle a brief storage outage.
• Validate that filesystems remount cleanly when the disk returns.
• Confirm DR snapshot strategies cover sudden volume loss.

View details

Azure instance CPU hog

Back to top

Azure instance CPU hog drives CPU utilization to CPU_LOAD percent (or saturates CPU_CORES cores) on the target VMs for TOTAL_CHAOS_DURATION seconds via the Azure VM run-command extension.

Use cases

• Test that application latency stays inside the SLA when compute headroom shrinks.
• Validate VMSS autoscale, AKS HPA, or App Service autoscale reaction time.
• Confirm critical processes keep getting CPU time during saturation.

View details

Azure instance IO stress

Back to top

Azure instance IO stress drives sustained disk read/write IO on the volume mounted at VOLUME_MOUNT_PATH of the target VMs for TOTAL_CHAOS_DURATION seconds via the Azure VM run-command extension.

Use cases

• Test that application latency degrades gracefully when storage is saturated.
• Validate Premium SSD burst credit and write-path back-off behavior.
• Confirm monitoring alerts on disk metrics fire inside the SLA.

View details

Azure instance memory hog

Back to top

Azure instance memory hog consumes MEMORY_CONSUMPTION MB (or MEMORY_PERCENTAGE percent of RAM) through NUMBER_OF_WORKERS workers on the target VMs for TOTAL_CHAOS_DURATION seconds via the Azure VM run-command extension.

Use cases

• Test that the OOM killer targets the right process under memory pressure.
• Validate GC pause behavior on JVM/CLR workloads.
• Confirm VMSS autoscale reaction time on memory metrics.

View details

Azure instance stop

Back to top

Azure instance stop deallocates one or more VMs listed in AZURE_INSTANCE_NAMES (or VMSS instance IDs when SCALE_SET=enable) for TOTAL_CHAOS_DURATION seconds, then starts them again.

Use cases

• Test how the workload behaves when a VM disappears.
• Validate VMSS auto-healing and AKS node-down handling.

View details

Azure Service Bus queue state change

Back to top

Azure Service Bus queue state change sets the status of one or more queues to disabled, sendDisabled, or receiveDisabled (via ACTION) for TOTAL_CHAOS_DURATION seconds, then restores Active.

Use cases

• Test that producers retry cleanly when sends are disabled.
• Validate consumer behavior when receives are disabled and the queue builds up.
• Confirm dead-letter handling under sustained chaos.

View details

Azure web app access restrict

Back to top

Azure web app access restrict adds an Access Restriction rule (RULE_NAME) to one or more App Service web apps with ACTION against IP_ADDRESS_BLOCK for TOTAL_CHAOS_DURATION seconds, then removes the rule.

Use cases

• Test that Traffic Manager / Front Door reroute traffic when an Access Restriction blocks the web app.
• Validate runbooks for removing a stale or misconfigured Access Restriction.

View details

Azure web app stop

Back to top

Azure web app stop calls the App Service stop API on one or more web apps for TOTAL_CHAOS_DURATION seconds, then starts them again.

Use cases

• Test how clients behave when an App Service web app is unavailable.
• Validate Traffic Manager / Front Door failover SLA.
• Confirm dependent services degrade gracefully and recover.

View details