Gregory's ServerI work with servers that slot into massive 12V busbars (thousands of amps). These servers have massive capacitors on their power distribution boards, and I can't even tell you the number of times I've had to tell folks on the factory floor "No, I need you to /actually/ power cycle it. How long did you wait?" Usually the answer is "I pulled it out and immediately reseated it." That... doesn't work.
|
6 comments
@dragonarchitect @SwiftOnSecurity I've done FPGA development, so I can tell you from experience (understanding a bit about how chips work internally) that yeah, sometimes you've got to power cycle to properly reset state. There's a lot of complicated shit inside there and sometimes there's bugs, either in silicon, firmware, or software. Gotta do what you gotta do!  @zetasyanthis @SwiftOnSecurity Soft errors can and do happen with enough frequency that a full power cycle is pretty much always the standard first step. That and "Did you try unplugging it and plugging it back in again?" Legit I fixed comms to a power distribution cabinet in a data center on Friday by just unplugging the network cable in the top and plugging it back in. Bad electrical contact prevented the WHOLE thing showing up on EPMS. @dragonarchitect @SwiftOnSecurity Yep. There's a reason it's a joke as old at time (though made famous by IT Crowd). :P @dragonarchitect @SwiftOnSecurity Basically, everything is a state machine, and you just have to reset state sometimes. (Especially when your computer or other electronics are made of, let's say, hundreds of thousands of them all tied together.) @dragonarchitect @SwiftOnSecurity See also: The need to sleep, which does a lot of things, but also this for (parts of) your brain. :P |
@zetasyanthis @SwiftOnSecurity I actually reverse-tech-supported a customer support agent over the phone once, when they commiserated to me about some of their tech woes and how they're asked to just turn it off and back on again.
I told them that sometimes a device just gets itself into a jam for no good reason and that "turning it off and on again" is a way of resetting the device back to a known working first-start state.
Now they know to try a 30 second poweroff as the first Tx step. :3