I had a curious episode unfold over a course of a week involving a server running F18. I'm just trying to solve a little puzzle in the aftermath, as a mental excersize.
I went through three bad UPSes. The manufacturer had a bad batch of UPSes.
From all appearances they worked when plugged into mains power. The server
was up and running normally plugged into the UPS, which had mains power, but the bad UPSes wre draining their internal battery, even though they had main power. Each UPS lasted about a day before its battery went completely dead, it started beeping at me, with the server continuing to run on mains power.
Here's the mystery. While I was dealing with that headache, the server started to drop off the network, maybe about every 5-6 hours, and stop responding to pings. The server itself, from all appearances, was running fine. I could log in on the console, everything seems to be working, except that it wasn't on the network. ifdown/ifup didn't help, ifup was getting stuck getting its IP address from DHCP. The switch had the server port lit, and blinking, as it was trying to contact the DHCP server, but the server wasn't getting its IP address.
About 20% of the time, unplugging the cable and plugging it back in worked. A reboot always worked. The server would reboot, come up on the network, have no problem talking to the DHCP server, and run fine for a few hours before it stops responding to pings again.
Anyway, while I was dealing with the UPSes, I also thought that I had this server's onboard NIC was going bad, and started getting quotes for a replacement server. But, after finally putting the server on a good, working UPS, it stopped having any kind of problems whatsoever with the network. It's perfectly fine.
So, that leaves me wondering how a bad UPS would affect a server's network port, and only its network port. As I understand it, with main power, the UPS is pretty much out of the loop, and it kicks in only if mains power is lost.
Allegedly, on or about 27 April 2013, Sam Varshavchik sent:
So, that leaves me wondering how a bad UPS would affect a server's network port, and only its network port.
If it has a noisy power supply, it could generate enough hash to cause interference with certain things. I have two or three external hard drive enclosures that do that. Whenever they're connected to a computer, my ADSL2+ modem continually drops out every few seconds. What they're connected to doesn't even have to be turned on (e.g. the computer is plugged into a switched off wall socket), it's just providing a point where the HDD supply can radiate crap through wiring. Some switchmode power supplies are just awful.
Bad earthing in the equipment or the building could cause earth loops, which can create peculiar problems. Particularly if some equipment is sensitive to that, but others are not. It gets even more difficult to manage when you network between buildings. And unless you buy the really expensive network switches or routers, they don't have transformer isolated inputs on each ethernet port.
As I understand it, with main power, the UPS is pretty much out of the loop, and it kicks in only if mains power is lost.
Well, a truly uninterruptible power supply would work like this:
The mains input charges a battery and provides DC to the inverter. The inverter provides AC to its output (going to your PC), either from the mains supplied DC or the battery. This way, there's an instant, unbroken, changeover. And it can cope with the mains voltage going up and down, by either the regulator coping with over voltage, or using the battery when there's undervoltage.
A cheap system might work as you suggest, and suffer glitches at the times that the mains becomes bad enough that it has to go over to the battery.
It could well be that you have had a series of bad UPSs, or it could be that your mains supply is so bad that the UPS spends most of its time supplying the computer, rather than charging its batteries. Or you could be putting too much of a load on it.
Tim writes:
It could well be that you have had a series of bad UPSs, or it could be that your mains supply is so bad that the UPS spends most of its time supplying the computer, rather than charging its batteries. Or you could be putting too much of a load on it.
Nah, it was the UPSes :-) Have three others nearby, some of them a few years old, chugging away.
Did some Googling, found a few discussion threads in the last month, talking about the exact problem I had, the exact same brand slowly discharging its battery, even though it had mains power. Looks like this manufacturer had some bad luck recently, with their quality control.
-
Sam Varshavchik -
Tim