A while back, after some months of frustration with significant levels of S3 API failures in bulk that never made it to the status page, I ended up writing a tool that continually monitors S3 via making random requests, and alerts if too many fail per unit time. The outcome of that experiment is the finding that there are a lot of significant spikes in S3 API failure rates that go entirely unannounced by Amazon, though the situation has improved considerably in the past year.
No comments yet.