How long is too long for you to not replace computer equipment due to reliability concerns?

[RTM]

I said I needed to upgrade my 10Gbit NICs to support ESXI 7.0. This is because they are using the old ixgb driver, which is not supported natively in ESXi 7.0.

It sounds like you have researched it properly, but on the off chance that you have missed it (there has been a few misunderstandings around too), I figured I would mention this:

Unless you have the really really old (AFAIK 1st gen Intel 10g, based on the 82598 controller) cards, your Intel NIC is probably supported on ESXi 7 with the ixgben driver.
This includes the ever popular and also quite old X520 card and its like that use the same controller (82599).

 

[RTM]

This is an interesting discussion, I am no expert but I doubt there is enough data available in the public about failure rates etc. of hardware, to actually do an unbiased risk assessment, that is free of peoples own past experiences on small samples of hardware.
What's (probably) missing is something akin to Backblaze's reports on disk failure rates.

From my perspective, security is also becoming a factor besides the usual power and performance (or both) factors.
Older equipment often times means no security updates, and that may not be acceptable for what it is used for.

I for one wouldn't want to run a DC on an 18 year old server, and one of the arguments for that is security: it is an asset that should be as secure as possible (this includes BIOS/firmware updates).

 

[ddaenen1]

Supermicro stuff is super robust. I had been running 2 X8SIL-F servers with X3430 Xeon's for years and they would go on for many years but i chose to upgrade to X9SCM-F only for AES-NI as one of them is running pfsense. For the rest, i don't notice anything significant in terms of performance and stability.

 

[cesmith9999]

What's (probably) missing is something akin to Backblaze's reports on disk failure rates.

Agreed,

the big 3 OEM (HPE, Dell, Lenovo) do not care about this trending as long as they get either new server orders or extension of the service contracts. There is no reason for them to publish those numbers. it would only hurt them as the numbers for the Seagate 1.2 TB disks did for Seagate.

The issue that I had to deal with is that we had servers that were 7 years old and still serviceable. We had to migrate the data off of them to migrate to new servers as the OEM would no longer extend the warranty on those servers. Moving multiple PB of data is no small task.

Chris

 

[Sean Ho]

Plenty of IPMI vulns. IPMI shouldn't be exposed to the internet, anyway. SM has been good about BIOS updates for Spectre et al. As long as the hardware keeps on truckin', and you're ok with the power bills, I don't see why servers should have a shelf life. I, too, still have a couple 1366 boxes; will gradually replace them with "bleeding edge" (8 year old) 2011-3.

 

[Ralph_IT]

...
From my perspective, security is also becoming a factor besides the usual power and performance (or both) factors.
Older equipment often times means no security updates, and that may not be acceptable for what it is used for.

I for one wouldn't want to run a DC on an 18 year old server, and one of the arguments for that is security: it is an asset that should be as secure as possible (this includes BIOS/firmware updates).

I agree, but it is not how it is for me.
Old hardware that is never decomissioned, but almost always repurposed, even if they are long depreciated.
Things are changing one step at a time, especially since the new European Data Protection Laws, but it is a long way and we started to walk yesterday.

As for the topic: It is reliable a 18 yo server? I would say yes. I see it everyday. It is secure? Hell, no. I would trash half our DPC right now and give the other half to a museum. XD
But you know...changes don't come easy.

 

[nk215]

Agreed,

the big 3 OEM (HPE, Dell, Lenovo) do not care about this trending as long as they get either new server orders or extension of the service contracts. There is no reason for them to publish those numbers. it would only hurt them as the numbers for the Seagate 1.2 TB disks did for Seagate.

The issue that I had to deal with is that we had servers that were 7 years old and still serviceable. We had to migrate the data off of them to migrate to new servers as the OEM would no longer extend the warranty on those servers. Moving multiple PB of data is no small task.

Chris

This is where I am at. After 10 years or so, moving data may become an issue. Data need to be on the new server, get tested before the end of the weekend. My machine is way out of warranty. I work around that issue by having the spare parts on hand.