Single node and rack aware clusters
Surveyor models hyperconverged Azure Local: 1 to 16 machines (8 when rack aware). Disaggregated deployments on an external SAN scale to 64 machines and are out of scope; the hardware page says so. Sources are in the in-app References page (facts registry, src/engine/facts.ts), checked 2026-10-15 against Azure Local 12.2609.
Single node
| Topic | What Surveyor enforces |
|---|---|
| Resiliency | Two-way mirror only (copies across drives). Three-way, parity and nested are unavailable. |
| Cache | Not supported. Cache media is forced to none; cache drives are an error. |
| Drives | One flash type (NVMe or SSD), at least 2 capacity drives. HDD is an error. |
| Reserve | min(1, 4) x drive size: one drive. |
| Availability | Warning: no live-migration target, no N+1, and updates that restart the host interrupt every workload. N+1 or N+2 leaves no workload node and is reported on the Fit page. |
Worked example: 1 node, 4 x 3.84 TB. Raw 15.36 TB; pool 15.36 x 0.99 = 15.2064; reserve 3.84; infrastructure volume 0.25 / 0.5 = 0.5 footprint; available 15.2064 - 3.84 - 0.5 = 10.8664 TB; usable at 50% = 5.4332 TB.
Rack aware clusters
Turn on Rack aware cluster on the hardware page. The node count is split into two equal zones.
| Zones | Nodes | Volumes | Efficiency | Fault tolerance (Learn) |
|---|---|---|---|---|
| 1+1 | 2 | Two-way mirror | 50% | One fault: drive, node or rack |
| 2+2, 3+3, 4+4 | 4, 6, 8 | Four-way mirror (rack level nested mirror) | 25% | Three drive or node faults; after a rack failure the survivor tolerates one more |
Rules enforced: node count 2, 4, 6 or 8; all-flash data drives; no three-way mirror, parity or nested volumes. Prerequisites shown as an information note: new deployments only (no conversion), Active Directory identity only, a cloud witness, a dedicated storage network, round-trip latency of 1 ms or less, and a 1+1 cluster cannot be expanded in this release.
Worked example (2+2): 4 nodes x 8 x 3.84 TB. Raw 122.88; pool 121.6512; reserve min(4, 4) x 3.84 = 15.36; infrastructure 0.25 / 0.25 = 1; available 105.2912 TB; usable at 25% = 26.3228 TB.
Assumption to know: Learn does not publish a rack-aware-specific reserve, so Surveyor applies the standard rule (one drive per machine up to 4). The Fit page also checks whether one zone alone has the CPU and memory for the full demand.
Dual parity efficiency
Learn publishes two tables and Surveyor uses the one that matches the capacity media.
| Nodes | All-flash | Hybrid (HDD capacity) |
|---|---|---|
| 4 to 6 | 50% | 50% |
| 7 to 8 | 66.7% | 66.7% |
| 9 to 11 | 75% | 66.7% |
| 12 to 15 | 75% | 72.7% |
| 16 | 80% | 72.7% |
Drive and cache rules
- At least 2 capacity drives per node (each 500 GB or more).
- A cache tier needs at least 2 cache drives and 4 capacity drives per node.
- Hybrid (HDD): cache of 32 GB or more per device, cache at least 15% of capacity, capacity drives a whole multiple of cache drives. All-HDD without cache is not supported.
- All-flash with a cache tier warns below about 5% cache (guidance, not a requirement).