External Temporal | NVIDIA Switch Infrastructure

NVIDIA Config Manager can run its worker, Workflow API, scheduler, and archive services against a Temporal service managed outside the deployment. Set an endpoint under external_services.temporal.address; the installer then omits the Temporal server, UI, database-schema setup, namespace creation, and search-attribute bootstrap jobs.

1 external_services:
2   temporal:
3     address: temporal.example.com:7233
4     namespace: network-automation
5     auth_method: mtls
6     tls_secret_name: temporal-client-tls
7     tls_server_name: temporal.example.com

For auth_method: mtls, create temporal-client-tls in the deployment namespace before installing. It must use the normal Kubernetes TLS keys: ca.crt, tls.crt, and tls.key. The installer mounts this Secret only into the NVIDIA Config Manager workloads that use the Temporal SDK.

The external server administrator must create the namespace and the following search attributes before starting the deployment:

Name	Type
`User`	`Keyword`
`DeviceID`	`Keyword`
`DeviceRole`	`Keyword`
`DeviceName`	`Text`
`DevicePlatform`	`Keyword`
`Site`	`Text`
`ReadRoles`	`KeywordList`
`ExecuteRoles`	`KeywordList`
`PendingApproval`	`Bool`
`FailedStage`	`Bool`
`IssueKey`	`Keyword`

For example, provision one attribute with the Temporal CLI:

$ temporal operator search-attribute create \
>   --address temporal.example.com:7233 \
>   --name DeviceID --type Keyword

When NVIDIA Config Manager manages Temporal itself (the default), the project-owned distroless bootstrap image creates the default namespace and these attributes idempotently during deployment. External Temporal deployments never run this image and retain full control of their own schema, namespace, and search attributes.