Repair, Upgrade or Replace: Making the Right Call

Something has failed or is struggling, and there are three options: fix what is there, improve what is there, or replace it. The wrong choice is usually made by defaulting rather than deciding, in both directions.

First, confirm what is actually wrong

The decision is worthless if it is made about the wrong component.

Failures are misattributed with remarkable consistency. A server that will not post is blamed on the system board when power supply, memory, processor seating or an expansion card are all more common and all produce the same symptom. A drive that is not detected is usually a carrier, interface, firmware or controller-state issue rather than a failure. A printhead is blamed when the print line is contaminated.

Our claims guide covers the checks worth doing first, and the management controller log frequently identifies the failing component directly.

Spend the ten minutes. Deciding to replace a server because of a failed power supply is an expensive way to be wrong.

Repair: when it is right

Replacing a failed component in otherwise healthy equipment.

Clearly right when: the failed part is cheap relative to the whole, the rest of the equipment has service life left, the platform is still supported by your operating system, and the part is obtainable.

Drives, power supplies, fans and printheads are all in this category. They are designed to be replaced, and the surrounding equipment frequently has years remaining.

Less clearly right when: the component is a large fraction of a replacement unit’s cost, or the failure is one of several on the same machine. A second failure on ageing equipment is information — components of the same age fail at similar times, as our lifespan guide covers.

The system board is the clearest case where repair frequently is not right. On an older server the board can approach the cost of a complete working unit on the secondary market, before counting the labour and the configuration work. Our board guide covers this.

Upgrade: when it is right

Adding capability to existing equipment rather than replacing it.

Clearly right when a specific constraint is binding and the platform can address it: memory when the host is memory-bound, drives when storage is the limit, a second processor to unlock slots and lanes.

Upgrading keeps the configuration, the firmware level and the staff familiarity, which is a real saving beyond the hardware.

Check three things first. That the platform actually supports it — a processor that fits the socket may not be on the supported list, and firmware may need updating before it is fitted. That the constraint you are addressing is the real one, since a bottleneck frequently moves rather than disappears. And that the upgrade does not exceed a chassis limit, such as processor TDP or power supply capacity.

The trap: upgrading a platform close to the end of its supported life. Money spent on a server that will be replaced in a year for operating system reasons buys very little.

Replace: when it is right

When the platform is leaving support. Operating system compatibility is the most common forcing factor, and it arrives regardless of hardware condition. Our compatibility guide covers checking for the target version rather than the installed one.

When parts are becoming unobtainable. A platform whose market is thinning has a shorter practical life than its condition suggests, and being forced into replacement by an unsourceable failure removes every option. Our spares guide covers watching for this.

When the constraint cannot be addressed. Memory ceiling reached, no free bays, insufficient PCIe lanes, chassis thermal limit.

When failures are accumulating. Several components failing on one machine over a short period usually means the whole thing is at the same point in its life.

When repair cost approaches replacement. Particularly true once labour and configuration time are counted honestly.

The costs people leave out

Four, and they change the answer often enough to matter.

Labour and configuration. A cheap part that takes a day to fit and configure is not cheap.

Risk during the work. Powering down ageing equipment is when drives fail, as our relocation guide covers. That risk is real and belongs in the repair column.

Remaining service life. Divide the cost by the years you will get from it. A repair buying eighteen months is not comparable to a replacement buying five years.

What happens at the next failure. If you repair now and the platform is unmaintainable in a year, you have spent money and still face the replacement.

A workable order

Confirm what has actually failed. Check whether the platform is still supported by the operating system you need. Check whether parts remain obtainable. Count the honest cost including labour, and divide it by the remaining service life. Then ask what happens at the next failure.

Those five questions usually produce an obvious answer, and they frequently produce a different one from the instinct.

One position worth noting: replacing does not mean discarding. If similar equipment is still running, the retired unit is the cheapest known-compatible spares source available — which is covered in our decommissioning guide.

Common questions

Send us the symptoms and the platform and we will help establish what has actually failed before you decide.