OK, i work in IT, i was called to the manufacturing floor this moring at 7am. Some PC's connected to a 3com HUB had no connectivity. I'll call this HUB1
I suspected that maybe HUB1 was being supplied by another HUB, or a switch that had went down, but there were no other people complaining of network issues and i coudn't trace the feed cable because it disapeared up into the suspended cable racks above the manufacturing floor.
So, i found another HUB (HUB2) further down the line, took a line from it and plugged it into HUB1, i stuck a laptop into a spare port and BINGO, i picked up an IP address and the PC's gained connectivity. So it looked like there was a problem with the initial cable going into HUB1 but i didn't know where it terminated so i called facilities to manually trace it for me.
Anyhoo, ten minutes later i got another call, every PC connected to HUB2 (the one i borrowed a line from) had lost connectivity, as well as HUB1 and the hosts attached to it.
Long story short, i eventually found the switch that was supplying HUB2 and the port on the switch that the HUB was plugged into was showing ERRDISABLE. After re-enabling the port, all hosts gained connectivity again.
Unfortunately we don't have a syslog server on the network, so when the switch error happened, no log was kept.
Do you think this was caused by too many collisions (running a hub from another hub)
Spanning tree violation?
Portfast?
I have no idea. What i do know is that we need to get our network sorted out. The only reason i figured this out was by methodically telnetting each switch in the factory and examining the port status on each. Luckily the switchport that supplied this HUB had been assigned with a friendly name that identified HUB2 (Bay 12 HUB) if you're interested

As it stands HUB1 is still running off HUB2, yet the port on the switch is still enabled.
Worrying..