Appearance
Operations: Azure Local Foundry
Scope
This is the operations page for Azure Local Foundry, one of the three targets in ADR-0011. Compare all three targets on the Deployment targets hub.
This target has not been built yet
Nothing on this target has been deployed from this repository, and no automation exists for it yet. Everything below is drawn from accepted decisions and first-party research. Treat it as a design, not an as-built record.
What is decided today
- Drift detection is split three ways, because
what-ifsees only the middle of the three layers. - More than sixty standard Azure Local infrastructure metrics are available at no extra cost.
- Managed Prometheus is deliberately deferred until a Prometheus-capable workload exists.
- Upgrade ordering carries the same load-bearing constraint as the install, and only the install half is documented. An AKS Arc upgrade is a rolling node replacement that necessarily restarts
istiod, which is the exact event Microsoft's own install warning calls flaky (SPIKE-29). - No Azure Monitor or Prometheus metric is documented for a
ModelDeployment. The infrastructure is richly instrumented and the model serving on it is not: no request count, no latency, no token count. Neither on-premises target can alert on a model, and ADR-0016 cannot be met here (SPIKE-27 Q6). - No preview-to-GA migration path is published and in-place migration is not promised. A breaking change has already landed inside the preview: three deprecated
ModelDeploymentendpoint fields plus an nginx-to-Gateway-API annotation migration table. - Teardown leaves residue, some of it by design. Certificate material and disconnected-operations artifacts are the parts that are not simply rebuildable.
The research and decisions behind this page
These were open when this page was first written. They are now written and linked below. Where a finding contradicted an earlier claim on this page, the page has been corrected and says so inline.