Facebook Inc. (FB.O) said a shift in the setup of its computer servers caused a widespread outage described as the largest ever recorded to hit the social-media network and its other services.

“As a result of a server configuration change, many people had trouble accessing our apps and services,” Facebook said in a tweet. “We’ve now resolved the issues and our systems are recovering.”

Beginning about 11:15 a.m. New York time Wednesday, global users encountered only partially loaded pages or no content at all on Facebook’s main social network and its related services, including photo-sharing site Instagram and messaging tools Messenger and WhatsApp. The outage continued into Thursday. It was the longest time Facebook’s properties have been offline since 2012 when Downdetector began tracking the company’s service level.

Several brand marketers tweeted that Facebook’s ad-buying system was down as well. The company said it was considering refunds for advertisers. Ad sales are the company’s central revenue source.

“As people become more reliant on digital services like Facebook, Instagram and YouTube, we are witnessing an ever-growing number of problems during outages,” Tom Sanders, co-founder of Downdetector, said in a statement. “The tolerance for downtime decreases and people are increasingly expecting services to operate flawlessly 365 days per year.”

Instagram service resumed shortly after midnight, the app tweeted from its official Twitter account.

Facebook shares declined 1.6 per cent at 2:29 p.m. in New York, and have jumped 32 per cent this year through Wednesday’s close.