Repair, also when the vendor is done
A switch with two dead ports is not a broken switch, and a server that drops out at random is usually not a broken server. On equipment that is out of support, replacement is the first answer from everyone who earns something on it. This page says when repair is the cheaper route and when we tell you to replace it instead.
Five kinds of equipment, and the faults that keep coming back
The pattern is the same across brands: the part that moves or gets hot goes first.
- Switches and routersDead ports, PoE dropping out across one block, a power supply giving up, and the memory the configuration lives on. On older models that last one is the most common fault and at the same time the cheapest to fix.
- Firewalls and access pointsPower supplies, fans and storage. A firewall that does not come back after a restart more often has a worn storage module than a failed mainboard, and that is a factor in the cost.
- ServersPower supplies, fans, memory modules, RAID controllers and backplanes. A machine that drops out at random far more often has a power supply than a system board, and that is the first place to look rather than the last.
- StorageControllers, cache batteries, backplanes and trays. Here the diagnosis weighs heavier than the part: a wrongly replaced controller costs data, and that is a different kind of damage than a day of downtime.
- Workplace and peripheralsLaptops, scanners, printers and point-of-sale equipment. In this group the answer hangs on the purchase price more often than on the technology, and we say so.
From fault description to tested device
Four steps. The second takes the most time and drives most of the cost.
You report it with type number and symptoms
What it is, what it does and when it started. Symptoms with a timestamp save a round trip on almost every fault.
Diagnosis
The fault is reproduced and narrowed down to one component. If that turns out differently from what was reported, you hear it before anything is ordered.
Repair and test
The part out, the replacement in, and then load it and measure. A device that boots is not the same as a device that holds up, and that difference surfaces in a test rather than at the customer.
Back with a report
What had failed, what was replaced and what was measured. On equipment that is out of support, that report is the only record still in existence.
Three situations where we advise replacement
Repair is not always the cheapest answer and we say so even when it costs us the job.
- The device no longer gets software updatesA firewall or switch that no longer receives security updates is not saved by a new power supply. After the repair it still sits open in your network, and that is a bigger problem than the fault.
- The fault comes from wear across the whole deviceWhen the third fan fails it is no longer an incident but a pattern. Repair then buys months, not years, and that sum belongs on the table before you decide.
- A replacement is available second-hand for lessOn a lot of network equipment a refurbished unit costs less than the repair. Then we supply that, and keep the failed device for parts.
What comes onto our bench
Multi-vendor here means one point of contact for equipment from different brands, including where the manufacturer no longer accepts the model.
- NetworkingCisco, Juniper, HPE Aruba, Extreme Networks, Brocade, Fortinet and NETGEAR. This is the group with the most repairs, and at the same time the group where it is most often a dead component rather than a dead device.
- ServersDell, HPE, IBM, Lenovo, Fujitsu and Supermicro. Power supplies, fans, memory and RAID controllers are the bulk of the work here.
- StorageNetApp, Dell, HPE, IBM, Hitachi Vantara, Quantum and Synology. With this group it always starts with the diagnosis, because a wrongly replaced controller costs data.
- What goes inManufacturer parts where they are available, and otherwise refurbished on part number. Memory and drives from Micron, Kingston, Seagate and Western Digital, processors from Intel.
What goes with this
A repair is an incident; the contract around it is the rest.
Questions we get about this
The ones that come up most, answered briefly.
Do you repair equipment that was not bought from you?
Yes. That is the bulk of this work: equipment that is out of support, or whose model the manufacturer no longer accepts. What we need is the type number and a description of what the device does.
What happens if the repair does not succeed?
Then you hear what was found and what the alternative is, and that is sometimes a refurbished unit that costs less than the repair. The diagnosis always produces a report, including when the device does not come back.
How long does a repair take?
That nearly always hangs on the part and not on the work. If the part is in the market it is a matter of days; if it is scarce, that is the first thing you hear rather than the last.
First find out what is broken
Give the type number and describe what the device does. You get back what the likely cause is and whether repair is the cheaper route in this case.
Practical IT knowledge in your inbox
New guides on management, security and the workplace, written by the people doing the work. No sales talk, and you can unsubscribe in one click.