Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
var ErrInstanceLimitReached = errors.New("instance limit reached")
ErrInstanceLimitReached is returned when a deployment fails due to an instance-count limit (e.g. Azure PublicIPCountLimitReached). The job queue uses this to learn the limit and stop further attempts.
var ErrUnsupportedFlavor = errors.New("unsupported flavor")
ErrUnsupportedFlavor is returned when a deployment fails because the selected flavor/VM-size is permanently incompatible with the provider configuration (e.g. Azure confidential-compute VMs requiring a specific securityType that scitq does not set). The job queue uses this to blacklist the flavor for that provider/region so the recruiter skips it.
Functions ¶
This section is empty.
Types ¶
type Provider ¶
type Provider interface {
Create(workerName, flavor, location string, hasGPU bool, image, gpuImage string, jobId int32) (string, error)
List(location string) (map[string]string, error)
Restart(workerName, location string) error
Delete(workerName, location string) error
}
Provider defines the methods that a cloud provider must implement.
Image selection on Create has three sources, in precedence order:
- `gpuImage` parameter (when hasGPU=true) — per-recruiter override from the workflow's worker_pool.gpu_image. Empty string falls through.
- `image` parameter (regardless of hasGPU) — per-recruiter override from worker_pool.image. Empty string falls through.
- scitq.yaml provider config — AzureConfig.Image / AzureConfig.GPUImage / OpenstackConfig.ImageID / OpenstackConfig.GPUImageID. The provider picks GPUImage when hasGPU=true, else Image.
Format of the image strings is provider-specific (Azure: "publisher/offer/sku/version"; OpenStack: image name or UUID). Each provider parses its own value and errors at Create time on malformed input — the recruiter doesn't validate format.
Source Files
Directories