> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tensor9.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Virtual Machines
> Operations, limits and service levels for Virtual Machines adapters with Azure origin APIs.
These terms describe the origin API served by the adapter, not direct calls to the target service. Management-request latency covers accepting and handling the API call; it does not promise that a cloud resource finishes provisioning within that time.
## Service level agreements
These tables show Tensor9's standard service levels and adapters whose standard terms are still being defined. Your signed agreement determines the SLAs, covered adapters and operations, limits, remedies and support terms that apply to your deployment.
Listed terms cover requests sent through this service adapter. They do not cover direct connections to the target service or replace the cloud provider's own SLA. Each origin-to-target pair has its own terms or an explicit pending status.
See the [SLA tables](/byoc/support/service-adapter-slas), [measurement rules](/byoc/support/service-levels), and [scaling conditions](/byoc/support/scaling-and-limits).
### Virtual Machines to EC2 Instances
| SLA | Service level | What is covered | Conditions |
| ------------------------------ | ----------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| Azure VM lifecycle | 30 ms monthly p99 | `azure::1.0.0::vm`: `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start` | One resource change and at most 32 KiB settings. Measures validation and durable acceptance, not completion of provisioning, restart, migration, failover or background reconciliation. |
| VM command acceptance | 20 ms monthly p99 | `azure::1.0.0::vm`: `VirtualMachines_RunCommand` | One configuration object, at most 50 entries and 32 KiB request/response data. A list is one page of at most 100 records and 256 KiB, not a paginator traversing the whole account. |
| Correct request handling | 99.9% per calendar month | `azure::1.0.0::vm`: `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start`, `VirtualMachines_RunCommand` | The request uses an operation and request shape that this adapter profile lists as available, and stays within the limits in the service page and your signed agreement. A target-service failure does not count as an adapter failure when Tensor9 correctly returns that failure to the caller. |
| Documented adapter behavior | Preserve the supported behavior documented for this adapter | `azure::1.0.0::vm`: `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start`, `VirtualMachines_RunCommand` | Retain resource identities and network associations through retries, deletion and recreation. Do not report unsupported disks, addresses or extensions as applied. A successful request acceptance is not proof that the guest command succeeded or that a workload is ready. Adapter-issued errors must identify unsupported request shapes rather than silently discarding fields. |
| Endpoint reachability | 99.9% per calendar month | Adapter endpoint reachability | The agreed production deployment has healthy target connectivity and receives a valid adapter health probe. This measures the adapter endpoint, not the target provider's service. |
| Adapter autoscaling and limits | Automatic scaling within the workload and deployment limits in your agreement | Adapter behavior across requests | Your agreement states the maximum request rate, burst size, request and response sizes, concurrency, and target calls per request. Tensor9 maintains adapter capacity inside that envelope. Target-service quotas and capacity are separate. |
**Deployment availability.** These targets describe this adapter design. Confirm that your deployment supports the listed operations; a numerical target does not establish runtime availability.
**What Tensor9 covers.** These service levels cover the adapter between the origin API and target API. They do not replace the target provider's SLA.
**How latency is counted.** Adapter work, including metadata and coordination, counts. Only separately measured permitted target waits are excluded.
**Which terms apply.** Your signed agreement names the covered operations, workload limits, remedies, and final service levels for your deployment.
#### What this adapter does
For `aws::1.0.0::ec2::instance`, the adapter handles Azure VM resource requests and coordinates their network attachments before translating to target compute. Azure resource identity and native machine identity remain linked across retries. RunCommand has an acceptance budget only: guest execution duration is separate. Metadata and managed-identity behavior remains the supported profile's contract, not an extra guarantee inferred from lifecycle timing. These targets describe this adapter design; confirm that your deployment supports the listed operations. A numerical target does not establish runtime availability.
#### How the latency targets were chosen
**Azure VM lifecycle: 30 ms monthly p99.** `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start`. Validate the origin request, retain its resource identity and requested configuration, and construct the target change. Adapter-owned metadata and coordination waits remain included.
**VM command acceptance: 20 ms monthly p99.** `VirtualMachines_RunCommand`. Bounded configuration validation, durable adapter metadata and response encoding; native provisioning completion and propagation are separate.
Your signed agreement sets the terms for your deployment. A target does not add an operation or option that the compatibility tables mark unavailable.
#### What counts toward latency
For a request-response row, measure from the agreed ingress boundary to dispatch of the complete response. A row that explicitly names a first response chunk ends at that chunk; its number does not cover the rest of the stream. A long-poll row names the intentional wait and when adapter delay starts. Include parsing, authorization, admission, translation, serialization, adapter-owned storage and coordination, retries and response handling. Subtract only separately measured target-workload waits and external network segments allowed by the measurement rules. The adapter's own response handling and dispatch remain covered. A database used for adapter metadata is still adapter work, even if a cloud provider hosts it.
For concurrent calls, exclude the union of permitted wait intervals, not the sum of overlapping spans. Calculate each request's adapter duration first, then the monthly p99. Do not subtract one service's p99 from another's. Known adapter timeouts are over-budget samples; failed or incomplete requests cannot disappear to improve the percentile. Missing measurements do not become zero latency. An SDK call span alone does not prove how much of its duration can be excluded.
#### Availability and failures
The 99.9% request target measures correct adapter handling, not the percentage of application calls that return success. Correctly forwarding a target quota or permission error is different from producing that error because the adapter sent the wrong request. Adapter-caused failures count even when the target is healthy.
With 1,000,000 eligible calls in a month, a 99.9% target permits at most 1,000 adapter-attributable failures. Endpoint probes have their own denominator. Correctness defects remain actionable even when the monthly availability percentage is met.
#### Scaling and target-service capacity
Scale API handlers and background workers separately. The agreement bounds resource count, concurrent changes, status polling, configuration size and burst growth. More replicas do not remove a shared metadata-store bottleneck or a target API quota. Tensor9 is responsible for adapter capacity within that envelope; the customer supplies target capacity and permissions. Native resource readiness is monitored separately from request acceptance. Pair compute and network capacity because a VM request can depend on a network interface and addresses. Rollouts must budget target instance limits and adapter resource-record contention, while guest CPU and boot time are separate.
Tell Tensor9 the expected steady rate, bursts, concurrency, payload sizes and operation mix. Tensor9 sizes and scales the adapter for the agreed load; you choose and monitor the target service's capacity with Tensor9's help. A latency budget is not a requests-per-second rating. Larger requests and higher rates need explicit terms, not silent inheritance of a small-request SLA.
#### Data and behavior guarantees
Retain resource identities and network associations through retries, deletion and recreation. Do not report unsupported disks, addresses or extensions as applied. A successful request acceptance is not proof that the guest command succeeded or that a workload is ready. Adapter-issued errors must identify unsupported request shapes rather than silently discarding fields.
#### Example and diagnosis
Create `batch-worker`, wait for its status, then deallocate it. The lifecycle target covers durable acceptance and origin response handling. Check target machine readiness and guest logs separately. A retry must reference the same intended VM and network attachment, not create an unrelated second machine.
Use `tensor9 explain` and the documented explain headers to understand the selected adapter and its behavior. Correlate available request diagnostics with the target provider's latency, throttling and capacity metrics. An explain report helps diagnose a request; it is not by itself a qualified SLA timing measurement. Share the operation, request shape, timestamps and request identifiers with support, with credentials and customer payloads removed.
### Virtual Machines to Compute Engine
| SLA | Service level | What is covered | Conditions |
| ------------------------------ | ----------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| Azure VM lifecycle | 30 ms monthly p99 | `azure::1.0.0::vm`: `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start` | One resource change and at most 32 KiB settings. Measures validation and durable acceptance, not completion of provisioning, restart, migration, failover or background reconciliation. |
| VM command acceptance | 20 ms monthly p99 | `azure::1.0.0::vm`: `VirtualMachines_RunCommand` | One configuration object, at most 50 entries and 32 KiB request/response data. A list is one page of at most 100 records and 256 KiB, not a paginator traversing the whole account. |
| Correct request handling | 99.9% per calendar month | `azure::1.0.0::vm`: `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start`, `VirtualMachines_RunCommand` | The request uses an operation and request shape that this adapter profile lists as available, and stays within the limits in the service page and your signed agreement. A target-service failure does not count as an adapter failure when Tensor9 correctly returns that failure to the caller. |
| Documented adapter behavior | Preserve the supported behavior documented for this adapter | `azure::1.0.0::vm`: `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start`, `VirtualMachines_RunCommand` | Retain resource identities and network associations through retries, deletion and recreation. Do not report unsupported disks, addresses or extensions as applied. A successful request acceptance is not proof that the guest command succeeded or that a workload is ready. Adapter-issued errors must identify unsupported request shapes rather than silently discarding fields. |
| Endpoint reachability | 99.9% per calendar month | Adapter endpoint reachability | The agreed production deployment has healthy target connectivity and receives a valid adapter health probe. This measures the adapter endpoint, not the target provider's service. |
| Adapter autoscaling and limits | Automatic scaling within the workload and deployment limits in your agreement | Adapter behavior across requests | Your agreement states the maximum request rate, burst size, request and response sizes, concurrency, and target calls per request. Tensor9 maintains adapter capacity inside that envelope. Target-service quotas and capacity are separate. |
**Deployment availability.** These targets describe this adapter design. Confirm that your deployment supports the listed operations; a numerical target does not establish runtime availability.
**What Tensor9 covers.** These service levels cover the adapter between the origin API and target API. They do not replace the target provider's SLA.
**How latency is counted.** Adapter work, including metadata and coordination, counts. Only separately measured permitted target waits are excluded.
**Which terms apply.** Your signed agreement names the covered operations, workload limits, remedies, and final service levels for your deployment.
#### What this adapter does
For `google::1.0.0::compute::instance`, the adapter handles Azure VM resource requests and coordinates their network attachments before translating to target compute. Azure resource identity and native machine identity remain linked across retries. RunCommand has an acceptance budget only: guest execution duration is separate. Metadata and managed-identity behavior remains the supported profile's contract, not an extra guarantee inferred from lifecycle timing. These targets describe this adapter design; confirm that your deployment supports the listed operations. A numerical target does not establish runtime availability.
#### How the latency targets were chosen
**Azure VM lifecycle: 30 ms monthly p99.** `VirtualMachines_CreateOrUpdate`, `VirtualMachines_Deallocate`, `VirtualMachines_Delete`, `VirtualMachines_Start`. Validate the origin request, retain its resource identity and requested configuration, and construct the target change. Adapter-owned metadata and coordination waits remain included.
**VM command acceptance: 20 ms monthly p99.** `VirtualMachines_RunCommand`. Bounded configuration validation, durable adapter metadata and response encoding; native provisioning completion and propagation are separate.
Your signed agreement sets the terms for your deployment. A target does not add an operation or option that the compatibility tables mark unavailable.
#### What counts toward latency
For a request-response row, measure from the agreed ingress boundary to dispatch of the complete response. A row that explicitly names a first response chunk ends at that chunk; its number does not cover the rest of the stream. A long-poll row names the intentional wait and when adapter delay starts. Include parsing, authorization, admission, translation, serialization, adapter-owned storage and coordination, retries and response handling. Subtract only separately measured target-workload waits and external network segments allowed by the measurement rules. The adapter's own response handling and dispatch remain covered. A database used for adapter metadata is still adapter work, even if a cloud provider hosts it.
For concurrent calls, exclude the union of permitted wait intervals, not the sum of overlapping spans. Calculate each request's adapter duration first, then the monthly p99. Do not subtract one service's p99 from another's. Known adapter timeouts are over-budget samples; failed or incomplete requests cannot disappear to improve the percentile. Missing measurements do not become zero latency. An SDK call span alone does not prove how much of its duration can be excluded.
#### Availability and failures
The 99.9% request target measures correct adapter handling, not the percentage of application calls that return success. Correctly forwarding a target quota or permission error is different from producing that error because the adapter sent the wrong request. Adapter-caused failures count even when the target is healthy.
With 1,000,000 eligible calls in a month, a 99.9% target permits at most 1,000 adapter-attributable failures. Endpoint probes have their own denominator. Correctness defects remain actionable even when the monthly availability percentage is met.
#### Scaling and target-service capacity
Scale API handlers and background workers separately. The agreement bounds resource count, concurrent changes, status polling, configuration size and burst growth. More replicas do not remove a shared metadata-store bottleneck or a target API quota. Tensor9 is responsible for adapter capacity within that envelope; the customer supplies target capacity and permissions. Native resource readiness is monitored separately from request acceptance. Pair compute and network capacity because a VM request can depend on a network interface and addresses. Rollouts must budget target instance limits and adapter resource-record contention, while guest CPU and boot time are separate.
Tell Tensor9 the expected steady rate, bursts, concurrency, payload sizes and operation mix. Tensor9 sizes and scales the adapter for the agreed load; you choose and monitor the target service's capacity with Tensor9's help. A latency budget is not a requests-per-second rating. Larger requests and higher rates need explicit terms, not silent inheritance of a small-request SLA.
#### Data and behavior guarantees
Retain resource identities and network associations through retries, deletion and recreation. Do not report unsupported disks, addresses or extensions as applied. A successful request acceptance is not proof that the guest command succeeded or that a workload is ready. Adapter-issued errors must identify unsupported request shapes rather than silently discarding fields.
#### Example and diagnosis
Create `batch-worker`, wait for its status, then deallocate it. The lifecycle target covers durable acceptance and origin response handling. Check target machine readiness and guest logs separately. A retry must reference the same intended VM and network attachment, not create an unrelated second machine.
Use `tensor9 explain` and the documented explain headers to understand the selected adapter and its behavior. Correlate available request diagnostics with the target provider's latency, throttling and capacity metrics. An explain report helps diagnose a request; it is not by itself a qualified SLA timing measurement. Share the operation, request shape, timestamps and request identifiers with support, with credentials and customer payloads removed.
## Debug this service
For requests through a service adapter, use Explain and request diagnostics to investigate the selected mapping.
Follow the [BYOC service-adapter debugging runbook](/byoc/service-adapters/debugging/overview) to capture and interpret the diagnostic evidence.
[Service Catalog](/byoc/service-adapters/catalog).