Skip to content

Question about cross-shipping RMA when you're your own datacenter

Dedicated Servers by wendy 24 replies 3.9K views
#11

This. The courier angle is underrated. In Phoenix I have had Micro Center same-day courier to the datacenter for consumer parts that were close enough to enterprise spec. NVMe is NVMe, the boot drive does not need to be the exact model if the firmware is compatible.

mediaaustin said:
Your real problem is not the RMA workflow, it is that you have no cold spare on site

With respect, this is easy to say when you are in Austin with same-day everything. Some of us are running lean on purpose.

seedbox, NAS, tape, and three offsite
#12

Update: I went and checked my notes from the KnownHost advance replacement. It was actually 2023, not 2024, and the faxed form was for a motherboard RMA on a Supermicro X10. They required the original invoice, which I had from the secondhand seller, and the hold was 350 USD. Released in nine days after they received the failed board.

So KnownHost has done it for Supermicro, at least historically. Whether they still do in 2026 is a phone call you would need to make.

#13

The KnownHost discord is mostly dead since their migration to the new platform but I can ask around if anyone has done this recently. The AI inference people are cheap and resourceful, if there is a loophole they have found it.

That said, I am with mediaaustin on the cold spare thing. Even a single 1TB consumer NVMe in a static bag at the cage is worth more than the 80 EUR it costs when your alternative is a 72-hour round trip. I learned this the hard way on a Sunday night when Hetzner had no stock in Hillsboro.

CUDA cores are my love language
#14

Following this thread because the infrastructure reliability angle is real. I had a client whose entire mail pipeline went down for 14 hours because of a failed boot drive in colo. Their SPF was perfect, their DKIM was rotated properly, none of it mattered.

hankels said:
Some of us are running lean on purpose

Lean is fine until it is not. The cost of one emergency courier or one advance replacement hold is usually less than one hour of your actual downtime cost. Calculate it properly.

SPF, DKIM, DMARC — holy trinity ✉️
#15

I am in São Paulo and the parts situation here is worse than you can imagine. For my colo in São Paulo I keep a full spare chassis because shipping anything from the US or Europe takes a week minimum and the import taxes make it absurd.

For wendy in Manchester, have you looked at local e-waste recyclers? There is a broker in London who pulls enterprise NVMe from decommissioned finance servers. I bought four Intel P4510s this way, all with 90% health, for less than one new unit. Not same-day but next-day courier within UK.

#16

In Kyiv we have learned to stock everything because supply chains are unpredictable. I run three Hetzner dedicated boxes in Falkenstein and I keep a spreadsheet of which parts are interchangeable.

wendy said:
The chassis is X11 generation

X11 is recent enough that many parts are still in channel. Try contacting Broadberry in the UK, they stock Supermicro spares and will take a card hold for advance ship. I have not used them for RMA but I bought a backplane from them in 2024 and they were responsive.

indentation is not optional
#17

Broadberry is a good lead, thank you YuriDavid. I will call them Monday.

flyuno said:
Local e-waste recyclers

I had not considered this. The London angle especially. I know there are decommission brokers in Slough and Reading who sell by the pallet but I assumed they only moved volume.

To everyone saying cold spare: noted. I will add a 1TB NVMe to my next parts order. The 80 EUR is less than my train ticket to the datacenter anyway.

Have you tried restarting it?
#18

I learning server very much and I have question. Why not use cloud like AWS or Azure? Then no need drive. I use Hetzner cloud in Nuremberg and it is very cheap.

2 #19

This was covered in the opening post: wendy has a specific colo footprint with Hyper-V and existing investment. Migrating to cloud is a different thread entirely.

minhdock said:
Why not use cloud like AWS or Azure?

Please search before asking, this is a 10-year-old forum and most beginner questions have been answered. I will not lock the thread but let us stay on topic.

To wendy: if Broadberry works out, please update the wiki page I mentioned. I will create a stub for UK-specific parts brokers this weekend.

sudo make me a sandwich
#20

I used to run colo in London and Manchester both. For same-day to a datacenter address, try calling the distributor Entatech if they still exist, or CMS Computers in Chester. They used to do emergency parts for MSPs and might take a card hold over the phone.

This was 2018-2020 though. The market has consolidated since then.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft