Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Azure Resource Manager is the deployment and management service for Azure. It provides the control plane that you use to create, update, delete, and organize Azure resources. The Azure portal, Azure CLI, Azure PowerShell, Bicep, Resource Manager templates (ARM templates), Terraform, Azure SDK management libraries, and direct calls to the Resource Manager REST API all use Resource Manager to perform management operations.
When you use Azure, reliability is a shared responsibility. Microsoft provides a range of capabilities to support resiliency and recovery. You're responsible for understanding how those capabilities work within all of the services you use, and selecting the capabilities you need to meet your business objectives and uptime goals.
Resource Manager has built-in resilience that Microsoft manages. This article describes how the service behaves during transient faults, availability zone failures, region-wide failures, and service maintenance. It also describes patterns that help your management and deployment workflows tolerate interruptions and restore the intended configuration of your resources, including the use of infrastructure as code. The article also covers configuration recovery considerations and key service-level agreement (SLA) information.
Production deployment recommendations for reliability
For production workloads, follow these recommendations:
Select resource group locations deliberately. Choose a location close to where management operations originate. To reduce the impact of regional outages, place your resources and resource group in the same region. For more information, see Which location should I use for my resource group?.
Maintain your intended resource configuration and its deployment dependencies in source control by using infrastructure as code, such as Bicep files, ARM templates, or Terraform configurations, together with any required scripts and policies. Regularly test that you can reconcile or redeploy the configuration from this known source of truth. For more information, see Backup and restore.
Use Azure-provided management tools and SDKs: These tools include handling for many transient faults, and they select the appropriate global Resource Manager endpoint for the configured Azure environment. Verify the retry behavior of the tool or SDK that you use. If you call the REST API directly, use the global endpoint, typically
https://management.azure.com.Design applications so that normal operation doesn't depend on Resource Manager. In outage runbooks, treat Resource Manager interruptions separately from outages of the Azure services that run your workload.
Reliability architecture overview
This section describes how the service's architecture supports reliability. It introduces the logical architecture, including the resources and features that you deploy and use, and the physical architecture, including the underlying infrastructure that runs the service.
Logical architecture
Azure operations use two planes: the control plane to manage resources and the data plane to use the capabilities exposed by resources.
Control plane: Resource Manager is the Azure control plane. When you send a request to Resource Manager, the service authenticates and authorizes the request, applies management capabilities such as Azure role-based access control (Azure RBAC), Azure Policy, management locks, and Activity Logs, and then forwards the request to the target resource provider.
A resource provider supplies specific types of Azure resources and the operations for managing them. For example,
Microsoft.Computesupplies virtual machine resources. For more information, see Azure resource providers and types.Throughout this article, a management operation means a control plane operation performed through Resource Manager.
Data plane: Data plane requests go directly to a service-specific endpoint, such as a storage account or database endpoint, and don't pass through Resource Manager. If Resource Manager is degraded or unavailable, you might still be able to use the data plane of existing resources.
For more information, see Azure control plane and data plane.
Resource Manager usually isn't on the runtime path after you deploy a resource. It primarily coordinates deployment, configuration, and other management operations. You're responsible for making your management workflows resilient to interruptions, throttling, and conflicting operations. You're also responsible for configuring each Azure service to meet your workload's reliability requirements, because Resource Manager's resilience doesn't make the resources that it manages resilient.
Resource Manager scopes and locations
Resource Manager organizes Azure into four hierarchical levels of management scope:
| Scope | Purpose | Deployment scope |
|---|---|---|
| Management group | Organizes subscriptions and other management groups so that you can apply governance settings across them. | Global scope. A management group isn't associated with a specific Azure region. |
| Subscription | Contains resource groups and resources and provides a boundary for billing, access control, policies, and limits. | Global scope. A subscription isn't associated with a specific Azure region. |
| Resource group | Groups related resources, typically resources that share a lifecycle. Each resource group belongs to one subscription. | When you create a resource group, you select an Azure region where Azure Resource Manager stores metadata about the resources in the group. Resources can be deployed to different regions from their resource group. |
| Resource | Represents a manageable Azure item, such as a virtual machine, storage account, or virtual network. | Most resource types are deployed to a selected Azure region, but some resource types are nonregional. A resource's location can differ from its resource group's location. |
Resource Manager endpoint routing
If you work directly with the Resource Manager APIs, send management requests to the global endpoint for your Azure environment. In Azure global, this endpoint is https://management.azure.com. The global endpoint uses DNS-based traffic distribution and automatic failover to route requests through healthy Resource Manager infrastructure depending on the scope and operation type. For more information about how Resource Manager routes requests to regions, see Behavior when all regions are healthy.
Azure tools and SDKs use the appropriate global endpoint for the configured Azure environment. You generally specify the endpoint yourself only when you call the REST APIs directly.
Physical architecture
Resource Manager is a distributed platform service. Microsoft deploys and operates the physical infrastructure that underpins Resource Manager, including regional service instances. Microsoft is also responsible for the service's availability zone and region failover behavior and service maintenance. You don't deploy or configure Resource Manager infrastructure.
Resource Manager stores resource-group metadata and other control-plane metadata in multiple Microsoft-managed data stores, which this article refers to as backing stores. Different management features and resource types can depend on different backing stores, so replication and recovery behavior can vary.
Resilience to transient faults
Transient faults are short, intermittent failures in components. They occur frequently in a distributed environment like the cloud, and they're a normal part of operations. Transient faults correct themselves after a short period of time. It's important that your applications can handle transient faults, usually by retrying affected requests.
All cloud-hosted applications should follow the Azure transient fault handling guidance when they communicate with any cloud-hosted APIs, databases, and other components. For more information, see Recommendations for handling transient faults.
You should expect occasional intermittent failures from Resource Manager, resource providers, or dependencies involved in management operations. Most Microsoft-provided tools and SDKs retry some transient failures, but behavior varies by tool and operation. Verify the retry behavior of the client that you use before you add another retry layer.
Handle failures based on the condition and the state of the operation:
Retryable failures: When you call the Resource Manager REST APIs directly, such as in custom automation that you build yourself, implement retry handling. Before you retry an operation, determine whether it's safe to repeat. Where possible, design operations to be idempotent, which means that repeating an operation has the same intended effect as running it once. Use bounded retries only when the failure is likely to be transient. Typical retry candidates include
408responses and5xxresponses.Use exponential backoff, which progressively increases the delay between retries, and add jitter, which introduces random variation into each delay to help prevent simultaneous retries from multiple clients. When the response includes a
Retry-Afterheader, wait at least the specified interval before you retry.Throttling: Resource Manager and individual resource providers apply separate request limits to protect platform stability. When you receive a
429response, honor theRetry-Afterheader and reduce your request rate. If your automation performs high-frequency discovery, deployment, or policy operations, monitor for throttling responses and shape the request rate accordingly. For more information, see Understand how Resource Manager throttles requests.Interrupted asynchronous operations: Many management operations run asynchronously. If the client disconnects or a transient failure occurs before the operation finishes, don't assume that the operation failed or terminated. Check the operation status or final resource state before you resubmit the request. If you use the Resource Manager API directly in your code, follow the
Azure-AsyncOperationorLocationURL returned by the operation, and honor theRetry-Afterheader when you poll it. For more information, see Track asynchronous Azure operations.Conflicting operations: Resource Manager detects conflicting concurrent updates to the same resource and returns a
409response for the later operation. Don't retry the request immediately. After the other operation finishes, re-read the resource state and decide whether another update is still necessary. For more information, see Resolve concurrent operations.
Resilience to availability zone failures
Availability zones are physically separate groups of datacenters within an Azure region. When one zone fails, services can fail over to one of the remaining zones.
Resource Manager uses availability zones for resilience in many Azure regions. If a zone becomes unavailable in one of these regions, Resource Manager can reroute management operations to another zone or region.
Availability zone support for Resource Manager is automatic. You don't need to enable or configure it.
Important
Resource Manager zone resilience only protects management operations. It doesn't make resource providers or resources automatically resilient to zone failures, and you might need to configure the resources separately. For example, to protect Azure Blob Storage data against a zone failure, select a zone-redundant storage option such as ZRS, GZRS, or RA-GZRS where supported. Refer to each service's reliability documentation to understand how to configure zone resiliency for that service.
Requirements
Region support: Resource Manager uses availability zones for resilience in selected Azure regions. A region might use availability zones, a backup region, or both. Microsoft determines and manages the resilience arrangement for each region.
Cost
There's no separate charge for Resource Manager availability zone support.
Configure availability zone support
You don't need to enable or configure availability zone support for Resource Manager.
Behavior when all zones are healthy
This section describes what to expect when all availability zones are operational.
Cross-zone operation: Resource Manager selects the infrastructure that processes each management request. If an availability zone becomes unavailable, Resource Manager can reroute management operations to healthy infrastructure in another zone or region.
Cross-zone data replication: Resource Manager stores control-plane metadata in multiple backing stores. Depending on the management feature or resource type, a store might replicate metadata across availability zones, regions, or both.
Behavior during a zone failure
This section describes what to expect when an availability zone experiences an outage.
- Detection and response: Microsoft detects the zone failure and manages rerouting and failover for management operations. You don't need to initiate a zone failover.
- Notification: Microsoft doesn't automatically notify you when a zone is down. However, you can use Azure Service Health to understand the overall health of the service, including any zone failures, and you can set up Service Health alerts to notify you of problems.
Active requests: In-flight resource management requests might fail, time out, or return a retryable response while the platform reroutes traffic. Follow transient fault handling guidance to safely retry the request after you confirm the current state of the target resource or asynchronous operation.
Expected data loss: Microsoft doesn't publish a uniform expected amount of control-plane metadata loss for a zone outage. Resource Manager uses multiple backing stores, and metadata availability and consistency during failover vary by management feature and resource type. Management operations might temporarily return older values for recent changes, and some operations might remain unavailable until their backing store recovers.
This behavior doesn't apply to data stored by the resources you manage. The data protection behavior of each resource depends on the service and how you configure it.
Expected downtime: Resource Manager request processing might be interrupted for a few minutes while traffic is rerouted to another zone or region. Individual management operations might remain unavailable for longer if they depend on a backing store that hasn't recovered.
If you're deploying or managing a resource that depends on a single availability zone, that resource can still be affected by the zone outage even when the Resource Manager service remains available. Refer to each service's reliability documentation to understand how to configure zone resiliency for that service.
Traffic rerouting: Resource Manager routes management requests to healthy platform infrastructure in the surviving zones or another region.
Zone recovery
When the failed availability zone recovers, Microsoft restores Resource Manager infrastructure and normal request routing. You don't need to initiate recovery. Check the outcome of management operations that were interrupted during the outage, and retry or reconcile them as necessary.
Test for zone failures
Microsoft manages Resource Manager's response to zone failures. Test that your management automation handles interrupted or failed operations by following the resilience to transient faults guidance.
Resilience to region-wide failures
Resource Manager is distributed across Azure regions. The global Resource Manager endpoint uses DNS-based traffic distribution and automatic failover to route requests through healthy Resource Manager infrastructure, depending on the operation's scope.
Resource Manager uses availability zones and backup regions as complementary layers of resilience. In regions where Resource Manager uses availability zones, it can reroute request processing away from an affected zone. Many Azure regions also have a designated Resource Manager backup region, which provides a cross-region recovery path. Some regions use both approaches. If a region failure occurs and the affected region has a backup region, Microsoft reroutes request processing for resource groups and the resources that they contain to the backup region.
Management groups and subscriptions are global scopes that aren't associated with a specific Azure region, so resource-group failover behavior doesn't apply to operations at those scopes.
Important
Resource Manager resilience applies to the Resource Manager portion of a management operation. After Resource Manager forwards a request to the target resource provider, the operation also depends on that resource provider and the target resource. Resource Manager doesn't fail over the resources that you manage or their data.
Microsoft-managed regional resilience for resource group and resource changes
When you create a resource group, you select the region where Resource Manager stores its configuration metadata. Resource Manager routes management operations that target the resource group through that location to help maintain a consistent state.
If a resource group's region has a backup region and the primary region is temporarily unavailable, Resource Manager automatically reroutes request processing to the backup region. This process is transparent when you use the global Resource Manager endpoint, and you don't initiate or configure the failover. The successful completion of an operation also depends on the backing store used by the management feature or resource type and on the target resource provider.
Resource Manager selects backup regions based on geographic proximity, data residency requirements, traffic patterns, and available capacity. These assignments are specific to Resource Manager and can differ from Azure region pairs. A backup region can serve more than one primary region. Microsoft doesn't provide a list of Resource Manager backup regions.
Requirements
Region support: Microsoft manages Resource Manager regional resilience and backup-region assignments. You can't view or change the resilience arrangement for a specific Azure region. Cross-region recovery behavior varies by management feature and resource type.
Cost
There's no separate charge for Resource Manager regional resilience or failover.
Configure multiregion support
Resource Manager regional resilience is built in and managed by Microsoft. You don't enable, disable, or configure it, and you can't select a backup region or initiate cross-region failover.
Behavior when all regions are healthy
This section describes what to expect when all regions are operational.
Cross-region operation: Region routing behavior depends on the scope of the operation:
For operations that target a resource group: Resource Manager sends the operation to the resource group's region, which is authoritative for the resource group's metadata.
After Resource Manager authorizes the request, it forwards the operation to the resource provider that owns the resource type. The availability of the overall operation then also depends on that resource provider.
For subscription- and tenant-level operations: These operations aren't associated with a resource group location. Resource Manager routes them to the most appropriate service instance based on several factors including the caller's location, observed performance, and Resource Manager availability, within geographic boundaries.
Cross-region data replication: Some Resource Manager backing stores replicate control-plane metadata across regions. Cross-region replication behavior varies by management feature and resource type.
Behavior during a region failure
This section describes what to expect when an Azure region experiences an outage when your management operations use the global endpoint.
- Detection and response: Microsoft detects failures of Resource Manager infrastructure and automatically reroutes request processing to a backup region when one is available. You don't initiate the failover. Operations that depend on a backing store without cross-region replication might remain unavailable until the backing store or source region recovers.
- Notification: Microsoft doesn't automatically notify you when a region is down. However, you can use Azure Service Health to understand the overall health of the service, including any region failures, and you can set up Service Health alerts to notify you of problems.
Active requests: In-progress management requests might fail or time out while traffic is rerouted.
Expected data loss: Microsoft doesn't publish a uniform expected amount of control-plane metadata loss for a region outage. Resource Manager uses multiple backing stores, and their cross-region replication behavior varies. Recent metadata changes might be temporarily unavailable. If an operation is interrupted, confirm the operation status or current resource state before you retry or reconcile your management automation. State held in a backing store without cross-region replication might remain unavailable until the backing store or source region recovers.
Resource Manager regional failover doesn't protect or replicate application data stored by the resources that you manage. The potential for application data loss depends on the regional resilience and replication behavior of each target service.
Expected downtime: Resource Manager request processing might be interrupted for a few minutes while traffic is rerouted to a backup region. Individual management operations might remain unavailable for longer if they depend on a backing store that hasn't recovered.
Traffic rerouting: The global Resource Manager endpoint automatically reroutes request processing away from unavailable infrastructure. Most Azure tools and SDKs use the appropriate global endpoint for the configured Azure environment automatically. If you call the Resource Manager REST API directly, connect to the global Resource Manager endpoint for your Azure environment.
This automatic rerouting changes where Resource Manager handles the management request. It doesn't guarantee that the operation can complete, move the target resource, fail it over, or change its service-specific disaster recovery behavior.
Region recovery
Microsoft manages the recovery of Resource Manager infrastructure. You don't need to take any action to restore Resource Manager after the region recovers. Check the outcome of management operations that were interrupted during the outage, and retry or reconcile them as necessary.
Test for region failures
Microsoft manages regional failover for Resource Manager. You can't trigger Resource Manager regional failover or select a backup region. Test that your management automation handles failed, interrupted, and uncertain operation outcomes by following the guidance in Resilience to transient faults.
Backup and restore
Resource Manager doesn't provide a native backup and restore capability. It manages resource configuration, but it doesn't back up application data stored by the resources that you deploy.
To recover resource configuration after accidental deletion or an outage, define your resources and related management configuration by using infrastructure as code, such as Bicep, ARM templates, or Terraform. Store these definitions and their deployment dependencies in source control so that you can redeploy from a known source of truth. Include configuration such as policy assignments, role assignments, diagnostic settings, and deployment scripts when they're required to restore the intended environment. You can export existing resources to a Bicep file or ARM template, but don't use an exported template as a full backup or disaster recovery artifact. Export isn't guaranteed to succeed, and an exported template might omit resources, properties, or required parameters.
Infrastructure as code restores only the resource configuration that it defines. It doesn't restore application data. Some Azure services provide export, backup, or replication features, which you should use together with Resource Manager-based infrastructure definitions. Review the reliability guidance for each service in your workload.
Resilience to service maintenance
Microsoft regularly applies service updates and performs other maintenance. The Azure platform handles these activities automatically, ensuring that maintenance is seamless and transparent to you. No downtime is expected during maintenance events unless you've been advised through Azure Service Health planned maintenance.
Service-level agreement
Resource Manager doesn't have its own SLA. Review the SLA for each Azure service that your workload uses to determine whether its availability commitment covers management operations, data plane operations, or both. For more information, see SLAs for online services.