Amazon S3's Preliminary Report:
7/21 7:19 PM PDT We wanted to share a brief note about what we observed during yesterday's event and where we are at this stage. As a distributed system, the different components of Amazon S3 need to be aware of the state of each other. For example, this awareness makes it possible for the system to decide to which redundant physical storage server to route a request. In order to share this state information across the system, we use a gossip protocol. Yesterday, we experienced a problem related to gossiping our internal state information, leaving the system components unable to interact properly and causing customers' requests to Amazon S3 to fail. After exploring several alternatives, we determined that we had to temporarily take the service offline so that we could clear all gossipped state and restart gossip to rebuild the state.
These are sophisticated systems and it generally takes a while to get to root cause in such a situation. We're working very hard to do this and will be providing more information here when we've fully investigated the incident. We also wanted to let you know that for this particular event, we'll be waiving our standard SLA process and applying the appropriate service credit to all affected customers for the July billing period. Customers will not need to send us an e-mail to request their credits, as these will be automatically applied. This transaction will be reflected in our customers' August billing statements.