Live data from Hacker News

Level 3 technician's misstep causes largest telephone outage ever reported

fiercetelecom.com

1–10 of 85 posts

Re: Level 3 technician's misstep causes largest telephone outage ever reported

#4
that is an impressive flaw. why would you ever assume a wild card across all parts of a phone number? I get a lot of spam calls where there area code + prefix matches my number and I just ignore them.

I have a hard time justifying even wildcard values for the last four digits in whole, partial wild card yes to get a PBX or such

Re: Level 3 technician's misstep causes largest telephone outage ever reported

#6
Oh wow.

> The technician left empty a field that would normally contain a target telephone number. The network management software interpreted the empty field as a 'wildcard,' ...

Exactly the same type of technical error happened nine years ago at Google!

> We maintain a list of [malicious] sites through both manual and automated methods. We periodically update that list and released one such update to the site this morning. Unfortunately (and here's the human error), the URL of '/' was mistakenly checked in as a value to the file and '/' expands to all URLs.

What it looked like: https://i.imgur.com/W5ICyVq.png, https://i.imgur.com/LrtLceN.png

Source: https://googleblog.blogspot.com.au/2009/01/this-site-may-har...

Re: Level 3 technician's misstep causes largest telephone outage ever reported

#7
I was an employee of Level 3 for a very long time (in IT, various roles -- most of my career was there). A small disclaimer: I left before this incident happened (about a year prior) and have spoken with nobody about it, so I have no insider knowledge specific to this. I was also an employee of Global Crossing that was acquired by Level 3 and this incident appears to have happened on the Level 3 side of the network (though it's not immediately clear; GC operated a SONET network and it's entirely possible it was from there).

The article and at least one other commenter mentioned this being a UI problem, and all I can say is "bingo". They didn't identify the vendor, but the article called out Cisco. I am a little skeptical of that, personally[0]. I lean toward it being something else, mainly because of the statement that followed "no one at Level 3 was aware of the consequences of leaving that field empty". We had a lot of very knowledgeable Cisco folks there. There are a few folks that I knew personally that were probably among the top folks on administering that equipment outside of Cisco. In addition, if a problem like this arose, they're accessible and helpful.

It was almost certainly one of the many, many, outdated software applications that make up the vast array of management interfaces into the equipment. I worked inside one of the phone switches (technically, one of the test switches). About a year prior to my leaving, the room next to my desk was filled floor to ceiling with cards that had been there on the day I started. I was told these cards ran along the lines of $30,000 a piece. If you wanted to upgrade the software, you had to replace the card. So we had machines in our operations center that were on isolated networks running Windows 98 in order to run the executable required to configure the switches[1]. We had devices made by Pirelli[3] that had similarly awful software. I knew of about 5 different devices with 5 different problem management tools but there were several more. And there were the handful of devices that were well known as ones "you don't even breathe on" (and I've never heard those words said about the Cisco products ... not that people spoke terribly highly of them, but never quite that negatively).

Telecom went through a really bad time in the early 2000s. At Global Crossing, every 6 months about 10% of the staff was laid off ... this happened for 10 years. Hardware didn't get upgraded, and therefore software didn't get upgraded. Sofware support contracts and maintenance were allowed to lapse. The quality of the software -- since its audience was a small handful of companies, many of which in the same financial state as Global Crossing -- was awful. After that many layoffs, the one or two guys who knew the myriad of corner cases involved in operating some of the management interfaces were on to other jobs at other companies, or retired. That this sort of thing didn't happen somewhat regularly is still surprising to me to this day.

At the time that I left, Level 3 was doing better from a financial standpoint and money was being invested in modernising a lot of these problems, but it's a huge network with many switch sites[4] in any major city that Level 3 terminated wires. Each of those sites had an array of equipment, some of it common, some of it from companies that went under shortly after the dot-com bust.

[0] It could easily have been a Cisco product, but there are so many -- far worse -- products out there that nobody outside of telecom has ever dealt with. I lean more toward that.

[1] Before someone says "Why didn't you virtualize them?". That was done for some machines with both running inside the isolated network, but it became more of a hassle than it was worth since these machines had to occasionally connect to the devices via serial port. The very buggy software included a very temperamental driver that only worked with a few models of older PCs and IBM (not Lenovo) Thinkpad laptop serial ports ... and only then if one of the wires was cut on the cable.

[2] Earlier than the style recently featured in a post about enthusiasts for the brand upgrading the motherboards.

[3] https://www.pirelli.com/global/en-ww/homepage ... Yes, that Pirelli. There's crossover between the rubber tires and fiber optics, apparently ... that was news to me. And they had some of the worst software I've ever encountered.

[4] We had, I think, 4 in Detroit. Some are not in terribly convenient locations, either. One in the area required passing an expensive OSHA certification, wearing steel toed boots and a hard hat if you wished to access it due to its proximity to a rail yard. I'd been there once and couldn't imagine the need for the restrictions.

Re: Level 3 technician's misstep causes largest telephone outage ever reported

#9
post #6

Oh wow. > The technician left empty a field that would normally contain a target telephone number. The network management software interpreted the empty field as a 'wildcard,' ... Exactly the same type of technical error happened nine years ago at Google! > We maintain a list of [malicious] sites through both manual and automated methods. We periodically update that list and released one such update to the site this…

This is a painful reminder for those of us who routinely design form-based interfaces to ensure ambiguous fields are explicitly understood by the user before entry or called out during verification of the submission.

Re: Level 3 technician's misstep causes largest telephone outage ever reported

#10
post #3

This sounds more like a failed ui in the managing software than the technicians fault if noone there knew what that empty field would do

I commented more extensively in the root of the post, but you can't even begin to imagine.

Think about every script you've ever written for "some thing at home" and how you only cared that it worked for the very narrow, specific, circumstances you were looking for. Maybe you left out error handling and just let it crash when you failed to put in the right parameter. Who cares? It's just a script for your one, lonely, workstation/server.

That's about the quality we're talking about. The companies that make these switches sell them to, maybe, five customers[0]. Software upgrades? Sure, if you replace that $30,000 card with the new version. Having trouble with the software? A support contract can be purchased for a similarly high fee[1]. A company producing this equipment doesn't put a lot of money into QA. In security, there was a general fear about these programs. It was so concerning that the management interfaces to the equipment was on "as close to air-gapped as you can be without being air-gapped" networks with the kinds of logging, auditing and the likes that you'd expect for a network holding government classified information[2].

[0] So few customers, in fact, that you can call them up with a serial number and find out who the purchaser is. I know this first hand due to someone propping the door to the switch site open resulting in, I think, 5 of what I was told were $30,000 a-piece cards being stolen. I was told they were effectively worthless to the thief, though, because nobody would buy them second hand in that manner and the moment they were offered for sale, if someone realized what they were, the thief would be caught.

[1] To be fair, I know of one specific circumstance where the company only offered paid support but that was mainly because I didn't work on that team; I'd speculate that all of them functioned this way.

[2] Well, maybe what you'd expect in an ideal world, anyway.

Post reply on HN