The Facebook OpenGraph Search API is experiencing an outage that began last night somewhere between 7 PM EST and 8 PM EST, as best we can tell. After receiving a tip from a Facebook developer who noticed that his social search engine built on top of the Facebook platform, Booshaka, stopped updating with fresh content, we began investigating. This morning, we performed several searches against the API including queries for popular words like “graduation,” and found that although the results displayed show a recent “updated” timestamp, the “created” timestamp is now pushing 15 hours old. Surely, someone on Facebook has mentioned “graduation” in a Facebook update after 7:18 PM EST last night?
Tracking the Outage
The OpenGraph Search API is a part of Facebook’s newly launched developer API, an interface to the social network which allows developers to access (with permission) users’ profiles, friends, status updates, etc. for use within Facebook applications and external services built on top of Facebook’s platform.
The search API in particular allows developers the ability to search all public objects available in the social graph, including posts, people, pages, events, groups and even an individual user’s New Feed as documented here on the Facebook developer site.
In Facebook’s provided example query, the search term used is “watermelon,” which may or may not be a very popular topic, so we used keywords that we were sure had been posted publicly somewhere, at some point, by at least one of Facebook’s 400 million users.
For instance, in a query for the word “graduation,” the most recent result result was a status update from 23:20 GMT (19:20 EST military time or 7:20 PM EST). The “updated_time” of that post was this morning, but that only refers to when a user comments on a post according to Facebook’s developer documentation.
In other words, around 7:20 PM EST last night, the OpenGraph API stopped pulling in new public status updates.
According to the developer who reported the issue, both his forum posting and bug report have, so far, been unanswered. We have also reached out to Facebook’s PR team and have yet to receive a response. (We will update this post if we do).
Facebook Outage Raises Questions
The larger issue here isn’t just that a portion of Facebook’s platform has gone down – numerous web services have issues from time to time, including everything from Gmail to Twitter. An outage of this length, however, with no official communication from the company itself is disturbing.
With Facebook’s new “Open Graph” initiative, the network has launched an incredibly ambitious plan to socialize the entire web by way of publisher plugins, APIs, “like” buttons, instantly personalized websites like Pandora and Yelp, “login with Facebook” boxes, live streams (like those found on CNN during live events) and more.
The social network is weaving itself into the very fabric of the web which means publishers, developers and users now depend on it to stay up-and-running at all times. If it’s not able to do so, then it essentially becomes a single point of failure for a number of web sites and services.
Today, only Facebook applications and services requiring access to public streams have been affected by the outage. Tomorrow, who knows? You can no longer “like” something or login to your favorite website? How will Facebook communicate its outages going forward? What will this mean for developers whose apps rely on Facebook’s platform?These questions remain unanswered. We hope that Facebook is thinking about these things too as it struggles to fix its broken API this morning. We hope it will have an answer soon.
Update from Facebook, 2:45 PM EST: “Our engineering team investigated questions about specific search functionality of the API and confirmed that at no point did search or other aspects of the Graph API go down. It appears that indexing of new stream content was delayed for 12 hours, but this has been fixed.”
Also, they would like us to know it’s called the Graph API, not OpenGraph API.