0

I recently submitted a suggested edit to this answer on HP DL380p Fans run 100% as soon as OS Loads but even though it highlights a potential fire risk involved with following the solution given in the answer, the edit was rejected.

Luckily my most recent comment has not yet been deleted (my previous comments were).

The answer suggests using a 1W resistor to slow down fan speeds, but when run at 100% PWM, as in the question, the resistor has to sink over 3W (from the specs and verified on the bench).

I tried out this solution with 2W resistor packs (made of 2x1W resistors) in my server and even though my server doesn't even run the fans at 100% except for at start-up, one of my resistor packs is already started showing signs of overheating. It had burnt through the heat shrink used to protect the leads and part of the outer covering has starting to flake away.

I originally questioned this answer (in comments which have been deleted) because I found a video which showed resistors getting to 150°C when using a similar solution to this, and I asked a question on electronics to make sure my calculations were correct.

If someone follows this advice and isn't careful about making sure the resistors don't touch anything, and where the resistors aren't in a the fan air flow (rather than hidden behind the fans housing, as in the pictures in the answer), this could potentially pose a significant risk of getting hot enough to ignite non fire resistant plastics in the case.

So, given the potential consequences, should I rely on my comment highlighting the risk not getting deleted, or should my edit have been approved?

| |
4

I can see how the reviewers went with "attempt to reply" as their rejection reason. Edits are supposed to maintain the "feel" of a coherent answer, rather than making the answer look like a Wikipedia talk page. So including your warnings about safety considerations in the answer -- or even just changing the recommended resistors directly -- would make a better approach.

However, I'm taking a rather more drastic view of the answer -- specifically, that it shouldn't exist at all. As you noted in your first comment, it's nothing more than a worse version of Phil McGuinness' answer. So I've nuked that answer entirely.

Further, I'm not sure either answer should exist -- modifying hardware that is in-warranty is an astonishingly terrible idea, and if the hardware is out of warranty/support it isn't on-topic on SF. I've not nuked Phil's answer for now, but I'm not entirely sure I don't want to.

| |
  • Thanks womble, that's a little more harsh than I'd apply on Robotics, but I can see why you made this decision. I agree with with the point about modifying in warranty servers, but I think the context here is for home lab use, with servers bought after they come out of warranty. Even so, my own solution was to make up new cables and add the resistors in there rather than modify the fans, which meant that I was able to remove the hack when I'd confirmed it didn't work well enough for my liking. – Mark Booth May 12 at 8:03
  • Incidentally, I was going to write up my experiences and post an answer here once I'd completed my Arduino PWM fan controller solution, would you consider this inappropriate for server fault, even if I included the corollary that this is not recommended for use in production? – Mark Booth May 12 at 8:04
  • Home labs aren't on topic, since "Questions on Server Fault must be about managing information technology systems in a business environment". Anything that isn't recommended in a business environment isn't suitable as an answer. SF is a bit of a special case in the SE world overall, because we're about "professional" system administration, which significantly changes the dynamic from every other SE site. – womble May 13 at 4:07
  • Yes, I suspect this would have been better off on Super User, but it's a little too late to migrate now. *8') – Mark Booth May 14 at 22:05

You must log in to answer this question.

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