Duration: October 2, 23:46 UTC to October 3, 02:54 UTC (3 hours 8 minutes)

Affected: SDK API, server-side API and web API in the US region. The Adapty Dashboard and the EU region were not affected.

We're sorry for the disruption this caused you and your users. We know how much you rely on Adapty, and we let you down.

What happened

At 23:46 UTC, a network switch at our hosting provider's data center had a disruption lasting about one second. Because of it, one of the servers in our US profile database briefly lost connection to the rest of the cluster and then reconnected. When a server rejoins like this, the database redistributes and re-syncs data across the whole cluster to make sure every copy is up to date.

While that re-sync was running, one type of lookup failed: finding a profile by customer user ID or device. The SDK relies on this lookup to create new profiles and identify users. Until the re-sync finished at 02:54 UTC, about 30% of SDK requests returned errors, and many of the rest responded more slowly than usual. During the same window, about 90% of server-side API requests failed. Errors stopped once the re-sync completed.

How it affected apps

Profile creation failed when a customer user ID was used. New users opening an app often couldn't do it.
Apps that wait for profile creation before showing a paywall were affected the most. For these apps, the paywall often didn't load, so some users couldn't make a purchase.
Apps that let purchases go ahead without waiting for profile creation were much less affected. Purchases went through, but some profile data, such as attribution, may be missing for users affected during the incident.

What we're doing next

The lookup that failed should never have been able to take down critical SDK requests like these. We're working on a permanent fix that removes this dependency, so the main SDK paths, including profile creation and paywalls, keep working normally even while the database is re-syncing. Once again, we're truly sorry this happened.