Next week, in the early hours of Wednesday morning, United Airlines will be taking most of its systems offline for a controlled restart and cutover window for their SHARES reservation system. Here’s what they’re telling their own travelers internally,
On February 4, 2026, from approximately 01:55-04:00 (Central time) United will be going through a major system update that will involve an outage of our Shares GDS. During this time, it will not be possible to see flight schedules, book travel, retrieve existing bookings, perform ticketing transactions, perform cancellation and/or refund transactions, or check in for flights using myIDTravel or ID90 Travel, as well as United’s own employeeRES or United app tools.
United has already pre-canceled most flights that depart during this time period. We strongly encourage staff travelers to plan in advance for this outage if they will need to travel on United, and book early. Additionally, we recommend checking in before the outage begins (check-in is open up to 24 hours before departure using the United app), especially if travel will commence shortly after the outage ends.
Some other communications show the window as 1:30 a.m. to 5:00 a.m. rather than just 1:55 a.m. to 4. It’s framed as a “technology upgrade which will restart our systems” to maintain “better functionality and reliability.”
With the reservations system down, United.com and mobile app, call centers, kiosks, travel agency channels and partner airline access are down during the window. Departures are paused during the window, but already-airborne flights continue. There’s no:
- Schedule display
- New bookings
- Reservatoin retrieval or servicing
- Ticketing
- Cancellations/refunds
- Check-in
In terms of the different anticipated timeframes, that’s likely a function of the hard downtime for the host system (“no transactions accepted”) versus a longer period for re-opening different channels in stages.
United hasn’t said exactly what they’re doing, and their IT architecture isn’t something I’m expert in. One possiibility is moving host systems and you’d want a clean transaction stop during the cutover. If they’re upgrading database layers, messaging systems or batch schedulers or tightly coupled systems around SHARES they might schedule a restart window.
The goal they’re promoting is better functionality and reliability after the controlled restart. But United has said more about its SHARES passenger service system transformation work more broadly.
- gradually moving systems and logic into cloud
- seat maps are a starting point – legacy systems were fragmented:
“Seating at United was complex, with 30+ applications managing seat assignments.. [and seat maps/assignments are] embedded in every part of the PSS.
- moving away from old flat files, with “90% of code/business logic in assembler” and rules scattered across multiple applications and interfaces
They’re trying to move away from mainframe and green screens into cloud architecture. After the outage, though, we may have more clues abotu what United has done.
- It was mostly infrastructure/reliability. No feature or interface changes to united.com, and potentially fewer errors online or dealing with agents.
- A step towards offer-order merchandising changes in how flights and prices are shown to customers
- MileagePlus In United’s earnings call they signaled major changes to the frequent flyer program in coming weeks, so this could be preparing for how that’s handled.
A planned restart/cutover that takes SHARES reservations capabilities offline for a few hours may just be ‘operational resilience’ whether from a new system environment or better redundancy setup. But it also may be more, and we’ll start to know mid-next week. If any readers are familiar with the details, I’d love to learn more.
@gary “you want to learn more”
you’ve already nailed the key takeaway
get out of assembler = get off big iron
why: the 6-digit hexadecimal pnr limitation goes away when the last line of TPF code is sunset
what is that?: the ACP/TPF (Airline Control Program / Transaction Processing Facility) was written I believe in 1962 and 1963 by IBM and as the world of technology evolved, the functionality written on top of and within “assembler” and “flat tiles” continued to be “distorted” to accommodate the restrictions and requirements of TPF
translation: we can’t post links on your board, but the appropriate video on youtube is found with this search…
Tron defeats Sark and the Master Control Program
United is hitting on all cylinders. Best app. Strongest airline IT.
When will these boobs abandon the Prehistorically horrible Shares?? I’d was brought to United by the cheap trailer park Continental because it was owned by them after they bought/merged with Eastern. Yeah, it’s just THAT OLD. And they trashed the FAR SUPERIOR United reservation and flight management system for this piece of garbage. I recall the weeks of outrageous cancellations and delays when they first implemented this wasteful ineffective trash. You’d think they would have scrapped it long ago but NOOOOOO they keep trying to fix a terribly broken system. Cheap as the day is long. UAL under Continentals “leadership” lost a lot of its main travelers “Global Services” because Shares screwed up so many things! Continental didn’t want to pay for the far superior data management systems United had because they were leased systems!!!! But they worked far better and were living systems always being updated and improved by the contracted supplier. United is run by cheap Imbeciles to the bitter end.
DL went heavily to cloud computing which was part of the reason the CrowdStrike failure was so bad.
Any time you use a single vendor, you expose yourself to higher risks.
I doubt if DL exclusively uses CRWD and I doubt if UA will use a single vendor for cloud protection.
and UA provides lots of information but that doesn’t make it a better app.
I don’t know of any other airlines that allow anyone to see their standby lists and their seat map after departure.
The thing is that the longer you add kludges to an old system, the more expensive it gets to maintain in the long run.
Oh goodie. I’m sure this will go well. Reminds me of Anatoly Dyatlov at Chernobyl forcing the ill-advised test of the reactor. “Safety first. Always.”
I will be somewhere over the Atlantic when that happens heading into LHR. Hopefully it doesn’t trash the system. Glad I am not leaving LAX on Wed 2-4!
Ironically, I think the reason United’s app is so good is because Shares was clunky & CAL was focused on cost. They used the app to let the customer manage their own travel and be less dependent on customer service agents to save money.
@Tim
The reason Delta handled Crowdstrike had very little to do with cloud computing. They had to reboot and apply patches manually to physical servers. In fact, cloud computing might have made things easier.
I can see how it might be hard for you to acknowledge that delta is not perfect — teens who realize their crushes aren’t perfect paragons doubtless feel the same way!
Jon,
they said they had to manually reboot something like 40,000 servers and CRWD did not provide an automated solution until days into the whole deal.
It isn’t about blaming or accepting. It is that far too many people have measured the success or failure of an airline’s recovery on the number of flights they cancelled.
If you know even the slightest thing about automation, then the answer as to how quickly a company restarts is how reliant a company is on the failed system
DL was heavily reliant on CRWD – they said about 60% or more of their servers used it.
And they had moved heavily to cloud computing.
I hope the lesson is to for anyone – not just DL – to be so heavily reliant on one vendor.
and obviously whatever UA is doing is a full scale shutdown and system conversion – which carries risks that everything will not restart as intended.
We fly Friday afternoon. We’re glad to be getting out of SFO this weekend – really – non-sports fan so there’s no reason to be in this city this weekend. Just hope it’s enough time for the ‘new’ system to have settled in by then.
I guess we’ll find out later on Wednesday how robust the testing of all the changes were. After a career in software development and deployment, I’ll be shocked if there aren’t any glitches.
And proof that fancy “EDR” tools are no substitute for application whitelisting and fine-grained network segmentation. But that takes work (and understanding your environment), so it’s easier to just buy something off-the-shelf that had a funny Super Bowl ad.
yes, you are right, refugee
Just like AS and UA should have checked the door bolts on its MAX9s so there would have been no door blow out or the potential that it could be repeated on another plane.
If UA had only thought ahead….
you either trust your vendors and supervise their work or you pay the consequences.
tim dunn writing about information technology and systems architecture is comedy gold
Brace! This will most likely blow up in IT’s faces. It usually does- then they have to rollback.
Trust me, I know. I’ve worked too many midnight shifts and seen these “outages” go every different way. It is a shock and a relief when everything goes right and everyone can get back to work.
hagbard,
people talking about DL’s failure regarding CRWD to cover up their “preferred” companies have had dealing with some of Boeing’s products over the past, say, 15 years
It isn’t hard to see which has been more impactful but they would never want to hear it.
Anyone want to start a pool on how long after United’s announced time for the system coming back does the system really come back? Speaking as a QA person wholesale system changeovers for something like this have to go over perfectly. If United was smart, they set up a duplicate system somewhere and already did a dry run migration (if that was actually feasible) to work the kinks out. If United wasn’t smart, they’re just hoping it all goes well and everyone needing to use the system Wednesday is going to be a human guinea pig and this could rival Southwest, Delta’s, and American’s system meltdowns for the chaos it will cause. Glad I’m not flying UA metal for another two weeks…
Agree with Adam. Hopefully it goes well but we have all seen IT implosions when systems are upgraded. Glad my family is not flying next week.
Give UA credit for notifying everyone in advance so we can plan for it. That definitely beats DL and WN meltdowns even if that comes to pass.
i assume this is part of the holdup for completing the Bluesky partnership.
I would love to get my Mosaic status when I am forced to take United flights.
For transcon, B6 is untouchable in my book. I’m 6’1 and short for my weight, JBU economy space is great.
Let’s get it going.
@hagbard celine, “tim dunn writing about information technology and systems architecture is comedy gold”
Amazing, isn’t it? Thanks for explaining it for we, who are IT illiterates.
@Tim Dunn “and UA provides lots of information but that doesn’t make it a better app”
What, exactly, does this mean? When you’re a consumer, when is less information not better?
I wrote exactly what I mean.
Anyone can see UA’s standby and upgrade lists off of their website – not just if you are on the flight. How many airlines do that?
UA allows anyone on their website to see seat maps after departure. Again, how many airlines to that?
More information does not necessarily make UA more technologically advanced.
DL has had its own version of connection saver for years and yet UA has touted its tool as if it was the greatest thing since sliced bread.
and DL moved heavily to cloud computing and storage after its massive IT meltdown that was caused by a power failure in its Atlanta IT facilities.
CRWD was so bad because DL was excessively dependent on a single IT vendor – which is not much different from AS, WN and UA’s excessive dependence on Boeing which created far larger and longer lasting impact to those companies that all of DL’s IT meltdowns combined.
Did anyone understand what TD just said there? Too funny.
UA is moving their Shares system that is housed in Charlotte to its new location in Chicago. The system will have to go offline for that to take place. Everything it was running on in Charlotte was old and outdated and needed replaced. They took this opportunity to move it to headquarters. This has been in planning for a long time. Was supposed to happen in the fall but they pushed it back to February.
I see Tim Dunn is still the biggest DL Cheerleader and making up crap to sound informed…
This is simple and a known activity for anyone who asked. UA is moving its SHARES instance to new hardware at a data center it owns and operates near ORD. It had been in DXC’s CLT data center, which has increasingly been impacted by hurricanes and other power issues with Duke Energy.
UA seems to be one of the few airlines left who own and manage their own data center vs. paying DXC or someone else to manage it.
@uait @eric
after this migration takes place, will UA have the capability to use IATA NDC XML to bypass GDS EDIFACT/ANSIX12 ??
Well done United!