Staged delivery helps when part of the infrastructure can launch earlier. Without an asset register and readiness criteria it becomes unclear which systems were accepted, which have defects and when billing began.
The protocol does not need to be a complicated legal document. A shared table, named owners and an agreed initial test window are enough to create a reliable record.
Create the register before the first batch
Give every server a permanent internal code. Record the serial or asset ID, configuration, facility or rack reference, management address, allocated IPs and handover time.
Do not store passwords in the shared table. Link to a secure vault or record that credentials were transferred through a separate channel.
- server code and serial number;
- CPU, RAM, drives and network adapters;
- public and management addresses;
- batch and handover time;
- status: testing, accepted, defect or replacement.
Define every batch
Before credentials are sent, the supplier states the number of machines and expected readiness time. The customer confirms that the batch can enter acceptance testing.
If one machine is late, the rest should not remain without status. Record whether an incomplete batch can be accepted and whether the volume discount remains unchanged.
NextUse one hardware test set→Use one hardware test set
Collect CPU model, memory layout, drive list, SMART data, network interfaces and controller versions. Run a short load, memory and storage check within the agreed window.
The goal is not a benchmark score. It is to find a wrong build, degraded drive, overheating or instability before the live-workload migration.
- inventory matches the order;
- logs contain no repeating hardware errors;
- SMART and drive condition meet agreed criteria;
- all memory is visible and stable;
- IPMI/KVM and managed reboot work.
Accept network and addresses separately
For each server test the port, agreed bandwidth, loss, MTU and routes from several external networks. For a subnet, check the next hop, PTR process, filters and address conflicts.
BGP or BYOIP adds a network criterion: the session is established, the prefix is externally visible and traffic passes in both directions.
NextSeparate a minor issue from rejection→Separate a minor issue from rejection
A minor issue may not block launch, while a critical defect makes the server unusable. Agree criteria in advance. A wrong CPU, missing RAM, failed drive, unavailable IPMI or missing route will usually block acceptance.
Every defect needs a correction deadline, permitted substitute and repeat test. A verbal promise to replace later should not mark the server accepted.
NextConnect billing to actual readiness→Connect billing to actual readiness
With staged delivery, each accepted server can have its own start date. The volume discount remains tied to the ordered quantity, while one late batch does not shorten another server’s paid term.
A single start after the whole fleet is ready is also possible, but it must be written before payment. Acceptance dates should match billing and the renewal invoice.
- credential handover time;
- test completion time;
- defects and correction deadline;
- acceptance and rental start;
- new date after replacement or re-delivery.