2

The title is telling, I guess. In particular, I am wondering how risky is it not to change the fan of a 4 years old Supermicro server?

Improve this question
3
  • 5
    Seriously? Is there a group of systems administrators going around pre-emptively replacing non-failed parts in their servers? How risky is it not to change a light bulb that hasn't burned out yet? Not to be rude but I see so many questions here that are just so much tilting at windmills. Is not replacing a non-failed server fan really that worrysome to you?
    – joeqwerty
    Aug 23, 2012 at 19:29
  • 4
    I can't imagine replacing a fan that hasn't failed or showed any signs of an issue on the reasoning that newer fan is more reliable.
    – David Schwartz
    Aug 23, 2012 at 19:31
  • 4
    i feel like this is one of those "my boss heard you should replace fans every 4 years" scenarios :)
    – au_stan
    Aug 23, 2012 at 19:43

3 Answers 3

Reset to default
2

This is an anecdote, not evidence. I've had (SuperMicro) server fans in service for 7+ years of continuous spinning without a problem.

I've also had one utterly catastrophic failure of one after five years. It became discolored and ultimately one of the blades shattered. We believe that it was the root cause of a multiple-drive failure in that machine. I keep it on my desk as a reminder to always make sure that hardware monitoring is set up, active, and properly configured to send alerts.

You can ask the server itself for information on how fast the fans are spinning. This information is available from the on-board sensors. Under Linux, look into lm_sensors. If your servers have IPMI, you should be able to get that information out of any IPMI tool as well. A fan that is acting up or that has failed will unexpectedly have a low or zero RPM.

Improve this answer
3

This will vary considerably: ball bearings vs sleeve bearings, made in Japan or China, ambient temperature, number of on/off cycles, etc.

But this is kinda like asking if you should be changing out hard drives periodically as they age. You don't do that because there's no reliable way to predict failure. So the proper way to do this is to gracefully handle failures when they do occur. This would be a RAID array for hard drives, or using redundant fans for cooling.

Improve this answer
2

IMHO, it's impossible to know. However, the best thing you can do for this is to maintain a proper computer room / closet / datacenter. A good clean environment with operating temps and humidity within ASHRAE (a quick blog post here) or manufacture specs is about the best you can do. Also try to keep it as dust free as possible.

You can always look up The Green Grid and Datacenter Dynamics for whitepapers.

Improve this answer
2
  • +1, ambient temperatures are a huge factor in computer component lifetime. If the fans designed to push cold air over warm components aren't getting cold air, you're going to be in for a world of pain.
    – Charles
    Aug 23, 2012 at 19:26
  • very true. i've been doing a lot of work in the space of raising the setpoints on my CRACS (also working with VFDs). One thing we have noticed is our servers fans going from say "medium" speed to "full" speed with an increase in our setpoits. basically 40% to 100%. I've heard stats/lectures that fan (mechanical for that matter) are some of the biggest power consumers. several thousand servers increasing fan speed could be a huge power draw that may not offset the price of say increasing our cold water temps. increase in fan usage may wear them down faster. let to be seen though.
    – au_stan
    Aug 23, 2012 at 19:33

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .