Root cause

Pinterest had a platform-wide outage that took down both their API and their website. Since we publish to Pinterest through their API, posts going out during that window failed. This was on Pinterest's side - nothing in Buffer changed.

Customer impact

For roughly two hours on the evening of May 14 (US time), publishing to Pinterest failed for anyone posting during the outage, and connecting new Pinterest channels didn't work either. Other channels and the rest of Buffer were unaffected. We didn't see any customer reports come in while it was happening.

Steps to resolution

We spotted the spike quickly - nearly all the errors in our publishing workers were coming straight back from Pinterest's servers with their own error message. We confirmed it was Pinterest (their site was down too, and outage trackers were lighting up), put a notice on our status page covering both publishing and new connections, and monitored until they recovered. Once Pinterest came back, we published a test pin to confirm and closed it out.

Key learnings

One thing we keep noticing across platform outages like this: a failed post currently shows up immediately as a failure rather than being retried, so we're looking at whether smarter retry handling on our side would make short outages like this one less visible next time.