warning to everyone out there!
The London Internet Exchange (LINX) suffered an hour-long outage yesterday evening, after an unnamed external network caused an ethernet loop and protective measures failed to work. LINX is currently trying to diagnose what went wrong. Reports on Twitter suggested that the exchange went titsup after Juniper's PTX Series packet …
Are you trying to say that the drive to push women who may or may not be able to do the job into a job that they may or may not even want caused this, you maybe right.
Comes around I guess. At least on poorly configured networks anyway.
What's up with the hipster lookin' PTX photo?
See "recursion" See "recursion"
Frack, why did the network stop working.
Yeah, I got the hour outage but my IDNet FTTC connection was dicky for the rest of the evening. It got better but even by 10pm I found I couldn't get to The Regster (oh noes!) or the BBC news site on some occasions. It was a bit odd really. When the connection failed I'd get the BT Wholesale gateway page. I can't quite see how that can be triggered by specific sites. The page seems to be saying that traffic is not getting from BTw to my ISP but surely the impact of that would be site agnostic.
At one point earlier in the evening I could reliably get to Thinkbroadband's main site but couldn't get to the forums. How does that equate to this web page?
it depends on how you isp is routing traffic to those sites.
According to my copy of Tannenbaum, ethernets are a logical circular topology. i.e. they are *supposed* to be loops even though they don't look like them from the spaghetti wiring in the switch closet.
Readers! Don't be confused by Netadmin Babblespeek! Demand the *real* reason for outages.
Er... are you confusing Ethernet with Token Ring?
If you could just leave your full name, current employment details and CV on the desk over there we'll make sure you get a job appropriate to your knowledge level.
In the meantime you might find it helpful to learn these handy phrases:
Do you want fries with that?
Would you like that large?
We stop serving breakfast at 10:30
You'll find a good explanation of ethernet topology here.
The short version is : it's a STAR topology.
I'm not the only one who's done this... dark wiring closets stuffed full of blue cables, all look the same.
reminds me of when a users network cable came out of the back of his pc and he kindly plugged it back into the hub under his desk. taking the other 50 users down with him.
Corrupt config on smart switches can also do this, seen it on Procurves.
Some twat plugging in an RJ crimped as a loopback can also create havoc. Allegedly.
Well you often on switch ports connected to servers or desktop/laptops disable spanning tree as the learning process can slow down connecting to the network significantly. Someone then messes up the patching and bang a switching loop. Surprised however that it could happen on a LINX segment that could affect so much though!
Or ethernet fabric are supposed to prevent this kind of bozogity from taking down LANs.
Anyone who thinks their network is well enough managed to never have it happen are either master of a tiny domain or supremely overconfident.
(yes, I have this argument regularly. Enduser ports are the most likely place for a loopback to occur, so switching off the protection to gain 30 seconds is an exercise in foot shooting.)
Spanning tree makes 802.1x even more complicated to implement.
I'm still waiting for SecureFAST v3.
In other words: "have you tried turning it off and back on again?"
Actually, the request made sense, as LINX allows one MAC address seen per port. Dropping carrier empties the list of seen MAC addresses.
There was quite a big gap between the install and the crash... Still, seems a little suspicious.
Biting the hand that feeds IT © 1998–2018