This document explains how to resolve errors that you might encounter when using
the [report faulty host API](https://docs.cloud.google.com/ai-hypercomputer/docs/manage/report-faulty-host.md.txt).

## Error messages

When you use the report faulty host API, you might encounter one of the following
errors.

### Compute instance deleted during a running operation

This error occurs when you delete a compute instance while a faulty host report
operation is still in progress.

**Error messages**:

    INTERNAL_ERROR

**Resolution** : Before you delete a compute instance, verify that the
[`reportHostAsFaulty` operation has completed](https://docs.cloud.google.com/ai-hypercomputer/docs/manage/report-faulty-host.md.txt#review-operations).
Otherwise, the report faulty host operation fails.

### Compute instance not part of a reservation

This error occurs when you report a compute instance that was created without
using a reservation.

**Error message**:

    INSTANCES_WITHOUT_RESERVATION_NOT_SUPPORTED

**Resolution** : Verify that you created the compute instance by using the
[reservation-bound provisioning model](https://docs.cloud.google.com/ai-hypercomputer/docs/consumption-models.md.txt).
To report a faulty host for an A5X, A4X Max, A4X, A4, A3 Ultra, A3 Mega, and A3 High (8 GPUs) instance that uses the
flex-start or spot provisioning model, contact your account team.

### Duplicate faulty host report operations

This error occurs when you send a request to the repair faulty host API while
another request is already in progress for the same compute instance.
Additionally, your compute instance may remain in the `REPAIRING` state.

**Error message**:

    FAILED_WITH_UNEXPECTED_STATUS

**Resolution** : Wait until the
[first report faulty host operation completes](https://docs.cloud.google.com/ai-hypercomputer/docs/manage/report-faulty-host.md.txt#review-operations)
before you send another request to the API. If the compute instance is
unresponsive, then restart the compute instance before you send another request
to the API.

### Fault reason is missing

This error occurs when you don't specify the `faultReasons` field in a report
faulty host request.

**Error message**:

    FAULT_REASONS_EMPTY_SHOULD_BE_SPECIFIED

**Resolution** : The `faultReasons` field is required. To resolve this issue,
specify the
[`faultReasons` field](https://docs.cloud.google.com/ai-hypercomputer/docs/manage/report-faulty-host.md.txt#report-a-faulty-host)
in your request.

### Incorrect machine type

This error occurs when the compute instance uses an unsupported machine type.

**Error message**:

    MACHINE_TYPE_NOT_SUPPORTED

**Resolution**: The report faulty host API supports only A5X, A4X Max, A4X, A4, A3 Ultra, A3 Mega, and A3 High (8 GPUs)
machine types.

### Rate limit exceeded

**Error message**:

    RATE_LIMIT_EXCEEDED

**Reason** : You exceeded the request limit for the
`instances.reportHostAsFaulty` method, or Google lacks sufficient capacity to
fulfill your request.

### Unsupported compute instance state

This error occurs when the compute instance is in an unsupported state.

**Error message**:

    INSTANCE_SHOULD_BE_RUNNING

**Resolution** : Before you send a request to the report faulty host API, verify
that the compute instance is in the
[`RUNNING` state](https://docs.cloud.google.com/compute/docs/instances/instance-life-cycle).