Understanding the present, shaping the future.

Search
04:33 PM UTC · SATURDAY, MAY 2, 2026 XIANDAI · Xiandai
May 2, 2026 · Updated 04:33 PM UTC
Technology

AWS Launches New S3 File System Feature; Stress Tests Reveal Metadata Sync Delays and Visibility Issues

Amazon AWS's new S3 Files feature allows S3 buckets to be mounted as NFS shares, but recent stress tests have revealed that objects with certain special names may become invisible once mounted.

Alex Chen

2 min read

AWS Launches New S3 File System Feature; Stress Tests Reveal Metadata Sync Delays and Visibility Issues
New AWS S3 Files feature allows mounting S3 buckets as NFS shares.

Amazon AWS has officially rolled out the S3 Files feature, which enables users to mount S3 buckets directly as NFS shares. Built on EFS infrastructure, the feature is designed to provide a file-system-like experience for object storage.

During stress testing of the new feature, the core product demonstrated excellent stability. By performing simultaneous write operations via both the NFS mount point and the S3 API, testers found that the system handled conflicts robustly, resolving all inconsistencies within two seconds without any instances of "split-brain" behavior.

Performance Metrics and Potential Technical Flaws

Test data revealed three distinct synchronization patterns. Write operations within the file system are aggregated and committed to S3 within a fixed 60-second window; new files created via the S3 API become visible at the NFS mount point after approximately 30 seconds; and updates to already cached files propagate the fastest, taking as little as 1.8 seconds.

However, the tests also uncovered flaws when handling edge cases. When objects are created using "creative" key names containing special characters—such as trailing slashes, double slashes, or emojis—some objects become invisible after being mounted via NFS. While AWS tracks these failures through the `ImportFailures` metric in CloudWatch, the client (such as the `ls` command) does not trigger an error, meaning users must proactively monitor specific metrics to detect data loss.

Additionally, there is a delay in the propagation of deletion operations. Tests showed that files deleted via S3 remain readable at the NFS mount point for anywhere between 6 and 18 seconds. AWS engineer Warfield noted that this may be due to delays in S3's internal deletion notifications.

Regarding pricing, S3 Files employs a tiered billing strategy. Users pay a premium for "hot data" actively used within the file system (at a storage rate of $0.30/GB), while the bulk of the remaining data is billed at standard S3 rates ($0.023/GB). For read requests exceeding 128 KB, the system bypasses the file system to stream directly from S3, incurring no additional S3 Files fees.

Comments