How old is your oldest server? (At a host)

Mine is 4 years. Webair has been great to me

6+ years

Changing out hardware just to change it is insane.

The question I posed was a little different to the other thread, in this case I'm talking purely about preventative maintenance. I agree that quality electronic parts are going to last a fair while, but mechanical parts typically have a much shorter lifespan.

Hard drives failing are a given sooner or later, but what happens if a chassis fan fails, or even worse a CPU fan?

Then you change it out.

I believe modern CPUs throttle back to prevent thermal overload, but if a chassis or PSU fan fails the results may be less spectacular... the customer may not be aware of anything being wrong until the parts in the case have been slow cooked and a capacitor eventually explodes. I can't see or hear my server, and the ability to reliably monitor its internals remotely is dependent on the hardware and my OS being compatible.

Anyway, I'm not being overly paranoid, I'm just curious whether any hosts actually do preventative maint...

If your host has your servers in a dusty basement, then it's your fault taking that host.

A decent datacenter tends to have air filters and closed server rooms, so there is almost no dust if any. you know, air filter companies sell these air filter products that come very handy when making a datacenter.
Also seems you have not been in a server room with 200 full racks deployed and the air blowing at full. If there is any dust around you have the risk of dust tornados.

For your home servers I suggest you to get a rack with air filters, so you will not need to clean the fans every 6 months.

i work with a site hosted with realitychecknetwork, dunno how old the server itself is but the uptime is over 2 years running on freebsd 4.10

not a server in the place is older than 5 months

come back when your company six months old

Hearing protection would also be recommended.. Stay in a datacenter server room any longer than 10 minutes, and your ears will be ringing for hours.

It generally doesn't make much sense to swap out known-good equipment, with possibly flawed equipment.

You use hard drives for example. The failure curve on those goes like a bathtub - relatively lots of failures the first 6 months, much less for the next 4-6 years (depending on whose stats you choose to use), then starts going back up slowly peaking at around 9 years.

If you're in the 3-4 year range, why take the chance? Same thing goes for power supplies.

CPU fans? What servers actually use CPU fans? Any decent server should be utilizing chassis fans blowing over passive cooling, for the exact reasons you cite. Modern servers make monitoring these fans rather trivial (and have all sorts of external indicators that you could hire a security guard to walk through the datacenter marking down machines w/ the blinky red light on if you cannot/will not automate it).

Your datacenter/host should already be doing things like cleaning air filters/etc. periodically.

Now.. retiring old hardware preemptively DOES make sense! But for differing reasons.

1. You start to see a pattern of a certain model line w/ increasing failure rates. Probably time to start migrating folks off of it. Semi-rare, but does happen.

2. You no longer can affordably, reliably, or otherwise stock a proper spare parts inventory for a given server type. Usually this is a financial reason, as it's much cheaper to have 3-4 models to stock parts for, than 3 dozen. Your host is stocking cold spares identical to your system right?

3. For a managed host - support reasons. Older hardware doesn't support newer management techniques (IPMI, onboard gige, old RAID adapters, whatever) and thus is much harder to standardize a management platform across. Probably makes sense to expire older stuff at an accelerated rate as it will save you money in software development or admin time. This can be counter-acted some with careful hardware selection at the time of purchase.


Now, these life cycles are measured in multiple years. 4 years in my experience seems to be about the longest you can expect to keep a given machine in revenue production. Some go a bit longer (rare), and some get pulled earlier for a variety of reasons. Also, you would be surprised about how many customers simply are not interested in a "free upgrade" to newer hardware - their current stuff is working just fine, why tempt fate?

Hope the info helps!

-Phil

2 years and going strong...

Got one from 2002 in a datacenter going strong... a power supply went on it and was replaced. Hard drive failed requiring the manual fsck but is still alive for now. Working on moving the last domains out to retire it for good. Yes that's right, 2002.

i got a few p3's still up

In the early 2000's I paid about $1.5k for a 1RU Celeron 733 server in a rack at he.net with free hosting by a "friend" ... he was caught nabbing BGP ASNs or unused IPs and his IP range was derouted, so any attempts to send packets to that network would fail. Outbound still worked though: I could still see the server trying to do DNS zone file fetches on my home office server for at least 2-3 years after that. I have no idea what happened to it, maybe it's still up... I never really chased it up since I didn't have to pay for the hosting...

almost 9 years.

What dust!?!?!?
And that's all quality and tested hardware with long lifetime
not some crappy china hardware....

all hardware comes from china lol

I just retired one at theplanet.com that had a 1300 day uptime. She was a good one

company maybe new, but i have been in hosting and ISP since 95, longer than pretty much anyone here, 2 of my techs, have been with me for a total of 21 years. It might be a good idea to go with experienced people and new equipment. maybe thats just me.