A misconfigured rate limit on an internal transfer queue slowed replication for objects larger than 1 GiB. Reads and writes were not affected. The limit was corrected and the backlog drained within 47 minutes.
Cross-region replication · duration 47 min · resolved
A regional shield node was drained for hardware maintenance before its replacement finished warming. Cache misses in sin1 took up to 180 ms longer for 18 minutes. We now gate drains on a warm-cache check.
Edge delivery · duration 18 min · resolved
A schema migration held a lock longer than expected. Creating and deleting buckets failed for 12 minutes; existing buckets and the S3 API were unaffected. Migrations now run in online mode with a lock timeout.
Management API · duration 12 min · resolved
The metrics pipeline fell behind after a traffic spike, so usage charts lagged by up to 40 minutes. Billing data was complete. We doubled pipeline partitions and added lag alerting.
Console · duration 26 min · resolved
A rolling deploy of the metadata tier briefly reduced write capacity. 0.4% of PUT requests in fra1 returned 503 SlowDown for 9 minutes; SDK retries succeeded. Deploy batch size was reduced.
S3 API · duration 9 min · resolved
Incidents older than twelve months are archived and available from support on request.