Categories
Mining Infrastructure

The deployment spreadsheet stops being true the first time a unit goes out for RMA. What a serial-level asset register should contain, how to reconcile it against your host, and why it matters for claims, insolvency and resale.

Ask most hosted miners to produce a current list of their machines — model, serial number, exact location, firmware version, and the date each one last ran — and you will get a spreadsheet that was accurate at deployment and has drifted ever since.

That drift is tolerable right up until the moment it is not. An RMA swap, an insurance claim, a site move, a dispute over which units were offline, or a host insolvency all turn on the same question: can you prove which specific machines are yours and where they are?

Why the deployment spreadsheet stops being true

A fleet changes constantly in ways nobody records at the time.

Units fail and go out for repair. Replacements come back with different serial numbers. Hashboards get swapped between chassis during maintenance, so the serial on the case no longer describes the hardware inside it. Machines get moved between rows, containers, or sites to balance load. Firmware is updated in batches, and the batch that failed gets left behind on an older version.

None of these events is unusual. Each one quietly invalidates part of the record, and because there is rarely a single moment where the register is obviously wrong, nobody rebuilds it.

What a register should actually contain

The useful version is more specific than most buyers maintain.

Serial number, recorded at the chassis and at the hashboard level where possible. Chassis-level alone is insufficient once boards have been swapped. If your host performs board-level repairs, ask what they record.

Precise physical location. Site, container or building, row, rack, and position. “Site A” is not a location when you need to identify forty units.

Ownership and acquisition evidence. Purchase invoice, bill of sale, and the date the unit entered the facility. This is the documentation that matters most if ownership is ever contested — see our discussion of where your miners sit legally when a host goes bankrupt.

Current firmware version and configuration profile. Including whether a unit runs stock or custom firmware, and at what power profile.

Service history. Every RMA, repair, board swap and replacement, with dates and the serial numbers going out and coming back. This is the field most often missing and the one that resolves the most disputes.

The RMA swap problem

This deserves separating out because it is where registers break most often.

A unit fails. It goes back to the vendor or to a repair facility. What returns may be the same chassis repaired, the same chassis with new boards, or an entirely different unit of the same model. Each outcome has different implications for your register, your warranty position, and — if the unit was financed or pledged — for whatever security interest attaches to it.

The practical control is simple and rarely applied: require that the outgoing and incoming serial numbers are recorded together on the same ticket, and that you receive that record rather than having to request it. Our warranty and RMA guide covers the claim process itself; the register is what makes the claim provable afterwards.

Reconciling against what the host reports

Most hosts provide a dashboard showing online units and hashrate. That is an operational view, not an audit. Two reconciliation habits are worth building.

Count reconciliation. Does the number of units the host reports match the number you believe you own at that site? A persistent discrepancy of even one or two units usually indicates an unrecorded swap or a unit that never came back from repair.

Serial-level reconciliation, periodically. Less frequent, more revealing. Request a current serial list and compare it against your register. The first time most operators do this, they find discrepancies. That is the point of doing it.

Where a host cannot produce a serial-level list on request, that is itself useful information about their internal records, and worth knowing before a dispute rather than during one.

Where the register earns its keep

Insurance claims. Equipment cover responds to identified property. A claim supported by serial numbers, acquisition evidence and location records is straightforward. A claim supported by an aggregate count is not. Our guide to mining insurance coverage sets out what these policies typically require.

Insolvency and ownership disputes. If a host enters insolvency, the question is which specific machines are yours, not how many. Contemporaneous records carry more weight than a reconstruction prepared afterwards.

Site moves. Relocating a fleet without a serial-level manifest means accepting whatever arrives.

Resale. Buyers of used equipment pay more when provenance, service history and firmware state are documented, and discount for uncertainty.

Tax and depreciation. Asset-level records support asset-level treatment. Aggregate records invite aggregate assumptions.

Chain of custody when hardware moves

Every transfer is a point where the record can break: into the facility at deployment, out and back for repair, between sites, and out at the end of the arrangement.

The discipline is the same at each point. A manifest listing serial numbers is produced before the hardware moves, both sides acknowledge it, and the receiving end confirms against the list rather than against a count. Photographs of serial plates at handover cost nothing and settle arguments that are otherwise unresolvable.

This matters most at the end of a relationship, which is also when goodwill is least reliable. A fleet that arrived with a signed manifest is far easier to recover than one that arrived as a delivery note saying two hundred units.

What to do when reconciliation finds a gap

It usually will, the first time. A discrepancy is not automatically a problem — most resolve into an unrecorded RMA swap or a unit still out for repair.

Work it in order. Confirm the host’s list is current rather than cached. Check your own register against outstanding RMA tickets, since units in transit are the most common explanation. Ask for the service history on the specific serials that do not match. Only once those are exhausted does the discrepancy become something to escalate.

Record the resolution against the affected units. A register that shows how a past discrepancy was explained is considerably more persuasive than one that simply agrees with the current count.

Keeping it current without it becoming a project

The register fails when updating it depends on someone remembering. Tie it to events that already happen: every RMA ticket updates the register as part of closing the ticket; every quarterly invoice is checked against the unit count; every site move produces a manifest before anything is loaded.

Specify it in the hosting agreement where you can. A clause requiring the host to maintain and provide serial-level records on reasonable request costs nothing to agree at signing and is difficult to obtain later. For the wider set of terms worth settling up front, see our checklist for evaluating a colocation provider.

None of this is complicated. It is simply the part of fleet management that produces no benefit at all until the day it produces a great deal.

Explore Rax Mining

Categories