On 7 September, 2024 at 05:10 UTC customers using Object Storage in the Singapore datacenter began seeing HTTP 503 errors. Investigation revealed that the hosts providing the service were running out of memory, which caused the Object Storage service to become unavailable. This was mitigated by adding additional swap space to the hosts. Further investigation revealed that some servers had an incorrect service configuration, which caused the Object Storage cluster to be unable to recover after the memory issues were rectified.
To address this issue, we have corrected the configuration and optimized our Object Storage service for better memory usage.