krazykanuck

I was in air when this happened, so when we landed at IAH there were no available gates per the captain as they were filled with planes that couldn't take off. Finally pulled up to Terminal C over an hour after landing.

I guess I should be happy we were in the air before the system went down otherwise I'd still be on a plane and not at home with a drink.

EmailKid

Going a little OT, but I'll take XP over Vista any day

But glad for all flying that planes are taking off again ^

WineCountryUA

CNN - United Airlines resumes flights after temporary ground order

buckeyefanflyer

How about a backup system to prevent this.

mjg59

Having a backup system means there's two things to go wrong.

That seems kind of glib, but it's not really. What would the backup system actually look like? What would be the conditions to trigger using it rather than the primary system? Would it use the same data? If so, how would you prevent data errors causing the same problems that might cause the primary system to stop working?

Designing complex IT solutions to keep working when there's a failure of an important system is a very hard problem. I'd be somewhat astonished if United didn't have a backup - I'd also be pretty surprised if United didn't end up failing over to the backup system moderately frequently without us knowing, and it wouldn't surprise me if this case were due to a failure of both the primary and backup systems.

United's certainly big enough that they should have sufficient IT to consider this kind of failure and plan for it, but the reality is that computer systems are very complicated and almost nobody is good at making sure they work all the time. Google's entire business is based on having working computers, but even they've had outages.

Is this bad? Yes. Lessons should be learned and I'd hope this specific problem (and similar problems!) never reoccur. But it's very hard to draw general conclusions about IT competence from occurrences like this.

televisor

To be completely fair: it looks like things got moving quite fast. Delta's outage last year was a completely different order of magnitude.

Still bad, but could be much worse.

amejr999

Just glad I wasn't on UA279 MCO-EWR:

Super80Fan

Yikes.


But as others are saying, this could've turned out a lot worse...

DenverBrian

Commodore VIC-20.

Evident

"Flight changes: The change fee and any difference in fare will be waived for new flights departing on or before January 25, 2017, as long as travel is rescheduled in the same cabin (any fare class) and between the same cities as originally ticketed.​​​"

Not giving an unconditional full refund? What if someone's travel plans got totally hosed and they don't need to travel anymore? Just going to get a credit then?

Rowyourboat

Tech employee here. Backup systems are well understood. You failover to them when the main system is down, and there's various protocols to make sure that the backup date is kept up to date and can reliably be switched over to when the main system fails. This is not a hard problem in 2017.

Google guarantees 99.9% uptime for their cloud services [1]. That means that in one year their services will be down for no more than 9 hours [2]. Now they can break their guarantee and in that case you get service credit, but in 2014 Amazon services had 2.5 hours of downtime and Google services had 4.5 hours of downtime for the whole year.

It is totally possible for United to dramatically reduce their issues if they invest in the IT. The problem is that the cost of dealing with outages is probably less than the cost of upgrading all their infrastructure, so they don't do anything. There's also the issue that SHARES/GDS is pretty old infrastructure that can't just be swapped out, but they can certainly improve a lot internally.


[1] https://support.google.com/work/answer/6056635?hl=en
[2] https://uptime.is/99.9
[3] http://www.networkworld.com/article/...last-year.html

rufflesinc

It's nice that large airports in the US don't routinely use remote bus gates, but wouldn't they be easily available for situations like this?

notquiteaff

random google shows a local outage for Sydney AWS last year.

http://www.zdnet.com/article/amazon-...uffers-outage/

How will this be counted against the uptime of the entire cloud? These services have many separate services, and it's not as simple as saying "Amazon Cloud is up" or "Amazon cloud is down".

(tech employee, here, too)

lincolnjkc

My thoughts as well -- it can also be said that United was not entirely down -- a significant portion of UAs dispatch ability was impaired, but the entire airline didn't go "down". For all we know backup systems were available and what allowed UA to prioritize dispatch of INTL/Hawaii/Alaska and for various considerations [capacity?] they deferred domestic dispatch as part of their recovery -- but domestic in the air stayed in the air.

If this was an ACARS issue as has been reported upthread how much of that infrastructure is in UA's direct control vs. ARINC/SITA [though if an ACARS issue I wonder why the effects were concentrated on UA mainline and didn't seem to spread to UAX/AA*/DL*/B6/F9/etc.

Ultimately it sounds like there was a relatively small failure that was magnified by lack of communication/expectation management/resources on the front line -- i.e. available gates/staff to move aircraft off gates for inbound aircraft to deplane, etc.

clubord

I totally agree with you. However it's not as simple as it seems.

Liability exposure is the main reason remote gates aren't prevalent in the US.