Categories
Hosting, Mining Infrastructure

Intake is the one period where responsibility for your hardware is genuinely ambiguous. Advance manifests, pallet inspection before signing, serial verification, the DOA clock that starts at delivery, and the rack map to insist on.

A pallet of miners arriving at a hosting site looks like the end of a process. It is actually the start of the one period where responsibility for your hardware is genuinely ambiguous, and where a problem that goes unrecorded becomes nobody’s fault and therefore yours.

Most hosting relationships are good ones. This still matters, because intake is the moment when three separate parties could each plausibly have caused a fault: the seller who shipped the unit, the carrier who moved it, and the facility that unpacked and installed it. Once a machine has been racked and running for a week, telling those apart is impossible.

Here is what should happen between the truck and the first hashrate, and what to insist on if it does not.

Tell the site what is coming, before it comes

A hosting site that receives an unannounced pallet has no way to check it against anything. Send an advance manifest: model, quantity, serial numbers, the carrier, the tracking reference, and the expected delivery window. If units came from different sellers or different batches, say so, because batches fail in correlated ways and that is useful information later.

Advance notice is also a scheduling courtesy with real consequences. Receiving several hundred units takes labor and floor space, and a site that knows what day the truck is coming can staff for it. A site that does not will put the pallets in a corner until someone has time, and hardware sitting unprocessed in a corner is hardware nobody is accountable for.

The manifest you send should be the same serial list you intend to maintain for the life of the deployment. Starting a serial-level asset register at intake rather than reconstructing one later is the difference between an afternoon of admin now and a genuinely unresolvable problem at the end of the agreement.

The receiving inspection: pallets before boxes

Inspection happens in two passes, and the first one occurs while the driver is still there.

Before signing the delivery receipt, somebody should look at the pallets. Is the shrink wrap intact? Are cartons crushed, punctured, water-stained, or visibly restacked? Is the pallet count what the paperwork says? Damage that is visible at this moment can be noted on the delivery receipt, and a notation on the receipt is the single strongest piece of evidence in any later freight claim.

Signing a clean receipt for a visibly damaged pallet forfeits most of that leverage. Carriers are entitled to assume that a clean signature means the goods arrived in good order, and unwinding that later requires proving something you had the opportunity to document and did not.

Agree in advance, in writing, who performs this inspection. If the site receives on your behalf, they are acting as your agent at the single most consequential moment, and they should know that is the expectation rather than discovering it during a dispute.

Unboxing: count, match, and photograph

The second pass happens at unboxing, and it is simple but tedious. Every unit gets checked against the manifest by serial. Not by quantity, by serial. A shipment of two hundred units where the count is right and four serials are wrong is a common and entirely survivable problem if it is caught on day one, and a nightmare if it surfaces eighteen months later during an exit.

Each unit should be looked at before it goes anywhere near a rack: bent fan guards, cracked shrouds, dented chassis corners, loose or rattling internals, missing power cords or brackets. Photograph anything that is not right, with the serial visible in the frame. Photograph the carton too if the damage pattern suggests transit rather than manufacture, because a crushed corner on a box with a dented unit inside it tells a coherent story that a photograph of the unit alone does not.

This mirrors the condition record that should be taken at the other end of the relationship, when the machines leave. The two together are what let anyone say with confidence what condition hardware was in when it entered and left a facility, which is exactly the question that gets argued about when a hosting agreement ends and the fleet moves out.

Dead on arrival, and the clock you did not know was running

Some proportion of any shipment will not power up. That is normal, and it is why the DOA window exists.

The important and frequently missed point is that the clock starts at delivery, not at installation. Manufacturer and reseller DOA windows are often short, measured in days to a few weeks from receipt. A pallet that sits unopened for three weeks because the site was busy can consume the entire window before anyone has pressed a power button, and a unit that would have been replaced for free becomes a repair you pay for.

So power-on testing should be part of intake, not a later step. Each unit comes up, reports its model and serial, and hashes briefly. Units that do not are set aside immediately and reported the same day, with photographs and whatever error the controller gave. Make sure the hosting agreement says who is responsible for performing that test and within how many days of delivery, because if it is silent, both parties will assume the other was doing it.

Installation, and the records that come out of it

Once units pass intake they get racked, and this is where you want one more piece of data: where each serial physically sits. Rack and position, recorded against the serial.

That mapping is not bureaucracy. It is what makes every subsequent support request specific and therefore actionable, as anyone who has tried to describe a failing machine without it quickly discovers. It is what lets you spot that your failures are concentrated in one aisle rather than scattered across the fleet, which is an environmental finding rather than a hardware one. It is the basis for a useful conversation about remote hands requests, because a technician can be sent to a location rather than asked to hunt.

Ask for the completed map at the end of installation. A site that cannot produce one is telling you something about how it will handle your fleet for the rest of the term.

First power and the burn-in window

The first seventy-two hours are the most informative period your deployment will have. Infant mortality is real, and units that are going to fail early mostly do so in that window under sustained load.

Watch three things rather than just hashrate. Hashrate against the expected figure for that model at the configured settings. Temperatures, both absolute and relative between units, because a single machine running consistently hotter than its neighbors in the same row is usually a mechanical issue rather than an environmental one. And error and reject rates, which often degrade before a unit fails outright.

Agree with the site what happens to units that fail burn-in, and specifically whether a replacement comes from your spares, from site stock, or waits for a shipment. That answer determines how much of your fleet is earning in month one.

The intake checklist worth keeping

Send an advance manifest with serials, carrier and expected date. Agree in writing who inspects at delivery. Inspect pallets before signing, and note damage on the delivery receipt. Verify every unit by serial, not by count. Photograph damage with serials visible, and the cartons too. Power-test within the DOA window, and know when that window closes. Report DOA units the same day with evidence. Record rack and position against every serial. Watch hashrate, temperatures and error rates through burn-in. Get the completed serial-to-position map before you consider intake finished.

It is an afternoon of discipline at the start of a relationship that will run for years. Every item on it exists because somebody skipped it and discovered the cost much later, when the evidence that would have settled the question no longer existed.

Explore Rax Mining

Categories