Skip to content
CSA Loom — the Microsoft Fabric experience for Azure tenants where Fabric isn't yet available: lakehouses, warehouses, notebooks, semantic models, Activator rules, Data Agents, across Commercial, GCC, GCC-High, and DoD IL5

LHDS brownfield: existing Azure estate

Not executed, and mostly not implemented. deploy-integrity.md R5 requires discover, offer, validate and accept for every backing service. LHDS code does not do that today. This page states what an operator can do with the merged code and lists the gaps exactly. No brownfield subscription has been used.

What the code does today

Service R5 step Reality in merged code
Loom hub registry / Container Apps environment discover, accept deploy-health.yml discovers them and passes them as controlPlaneConfig. The only automated discovery
Existing LHDS stamp refuse duplicates Workflow refuses if an rg-csa-loom-health-* group exists unless allow_existing_hub=true
ACR accept supplied healthConfig.acrLoginServer; controlPlaneConfig.acrName/acrResourceGroup/acrLoginServer
DNS / networking accept supplied healthConfig.extraDnsLinkVnetIds, vnetAddressPrefix; the stamp creates its own VNet
TLS certificate accept supplied healthConfig.tlsCertificateSecretUri (version-less Key Vault secret URI of a CA certificate; the gateway identity needs Key Vault Secrets User on that vault). Empty = the control plane issues a self-signed certificate in the stamp vault; see greenfield TLS
APIM accept supplied modules/health/ingress.bicep takes apimServiceName, apimResourceGroupName, apimSubscriptionId and publishes a policy fragment to an existing service. Not reachable from stamp.bicep / health.bicep (gap 3)
AKS, PostgreSQL, Blob, Key Vault, Event Hubs, Log Analytics discover / offer / validate / accept None. stamp.bicep always creates its own. The existing references in tenant.bicep and engines.bicep point at the stamp's own resources, not customer ones

Not yet implemented (exact gaps)

  1. No multi-subscription analysis of existing AKS, PostgreSQL, storage, Key Vault, Event Hubs, APIM or Log Analytics, and no per-service use-existing-or-deploy-new choice.
  2. No existing-resource parameters in stamp.bicep for those services (none of its params accepts a resource id for them), so there is nothing to validate SKU, region, network reach or RBAC against.
  3. APIM params not wired: ingress.bicep accepts the apim* params but stamp.bicep and health.bicep do not pass them through, so the policy fragment can only be published by invoking the module directly.
  4. APIM identity Reader role is not granted by any bicep.
  5. OpenID configuration discovery: there is no tenant-independent discovery URL for the LHDS authorization server, which the APIM fragment requires.
  6. Per-stamp PDP data and bundle publishing pipeline (see greenfield).
  7. HealthCP.Reader / HealthCP.Admin app roles, the federated credential and the caller allowlist are done by the post-deploy bootstrap step bootstrap-health-cp-entra.mjs (#5163; merged, not deployed, never run against a tenant). It needs the Graph permissions named in its failure output.

Interim operator path (unverified)

Until the above exists, a brownfield subscription can only supply registry/ACR values, extra DNS link VNets and a TLS secret URI, and let LHDS deploy its own stamp resource group alongside the existing estate. That is "deploy new alongside", which R5 forbids as the default. It is documented as the current limitation, not as a supported brownfield path.