Skip to content

Releases: awslabs/mountpoint-s3

mountpoint-s3 v1.7.0

06 Jun 09:09
42007f7
Compare
Choose a tag to compare

New features

  • Metadata caching can now be configured independently of data caching. When passing the --metadata-ttl <seconds> argument without also specifying --cache <directory>, Mountpoint will cache file metadata in memory for up to the given TTL, but will not cache object data. The --metadata-ttl argument also accepts two special values: minimal to enable only the minimal necessary caching, and indefinite to cache indefinitely. These modes can help accelerate workloads that touch many files but do not need to cache object data for re-use (for example, listing a directory and then reading each file within it once). (#855)

Breaking changes

  • The --metadata-ttl 0 setting is no longer supported and will be removed in a future release. The new --metadata-ttl minimal has a similar effect, but behaves better when latency for S3 requests is high. (#855)
  • When using the --cache flag, the default metadata TTL is now set to 60 seconds (--metadata-ttl 60) instead of 1 second. (#855)
  • Mountpoint now uses STS regionalized endpoints when assuming IAM roles configured in a CLI profile. If you specify a region in your CLI profile and want Mountpoint to use a role from that profile you have to make sure Mountpoint can access STS regionalized endpoints. This allows Mountpoint to assume an IAM role in regions outside of the aws partition. (#877)

Other changes

  • The checksum algorithm to use for uploads to S3 can now be chosen with the --upload-checksums <ALGORITHM> command-line argument. The only supported values in this release are crc32c (the default, and the existing behavior) and off, which disables including checksums in uploads. The off value allows uploads to S3 implementations that do not support additional checksums. This option defaults to off when the bucket name is an S3 on Outposts bucket access point (either an ARN or a bucket alias). (#849)
  • Fixed an issue where Mountpoint did not send the Content-Length header when creating multi-part uploads. (#875)
  • Fixed an issue where Mountpoint could not assume an IAM role specified in a CLI profile with EcsContainer as a credential source. (#875)
  • Added support for AWS_ENDPOINT_URL environment variable. (#895)

Binaries

x86_64

arm64

mountpoint-s3 v1.6.0

11 Apr 17:12
ad0f530
Compare
Choose a tag to compare

New features

  • Mountpoint for Amazon S3 now supports specifying an AWS Key Management Service (AWS KMS) key for server-side encryption with KMS (SSE-KMS) when mounting an S3 bucket or prefix. (#839)

Breaking changes

  • No breaking changes.

Other changes

  • Mountpoint now retries S3 requests up to a total of 10 attempts (up from 4), which should make file operations more robust to transient failures or throttling. The maximum number of attempts can be overridden by setting the AWS_MAX_ATTEMPTS environment variable. (#830)
  • Fix an issue where Mountpoint could become unresponsive after opening too many files in write mode. (#832)
  • Add support for rewinddir by restarting readdir if offset is zero. (#825)

Binaries

x86_64

arm64

mountpoint-s3 v1.5.0

07 Mar 17:05
56591e7
Compare
Choose a tag to compare

New features

  • When caching is enabled, Mountpoint also remembers when objects do not exist, in order to reduce repeated lookups. (#696)

Other changes

  • Cancel S3 requests when dropped. Addresses an issue where the prefetcher could keep streaming up to 2GB of data that would never be used. (#794)
  • Improve read throughput in more non-sequential access patterns by better accounting for the progress of in-flight prefetch requests. (#797)
  • Stop limiting the number of connections based on the number of known IPs when connecting to S3. Improves maximum throughput on S3 Express. (#796)

Binaries

x86_64

arm64

mountpoint-s3 v1.4.1

16 Feb 16:28
1c00fd3
Compare
Choose a tag to compare

Other changes

  • Fix an issue where read file handles could be closed too early, leading to bad file descriptor errors on subsequent reads. As a consequence of this fix, opening an existing file to overwrite it immediately after closing a read file handle may occasionally fail with an "Operation not permitted" error. In such cases, Mountpoint logs will also report that the file is "not writable while being read". (#751)
  • File handles are no longer initialized lazily. Lazy initialization was introduced in version 1.4.0 but is reverted in this change. If upgrading from 1.4.0, you may see errors that were previously deferred until read/write now raised at open time. (#751)

Binaries

x86_64

arm64

[RECALLED] mountpoint-s3 v1.4.0

26 Jan 16:49
e52d68e
Compare
Choose a tag to compare

Caution

This release has been recalled due to an issue with file reads. Please use v1.4.1 or later instead.

New features

  • Allow file overwrites when mounting with --allow-overwrite option. The upload will start as soon as Mountpoint receives write request and cannot be aborted. Once it is started the file is guaranteed to be overwritten. (#487)

Breaking changes

  • No breaking changes.

Other changes

  • Update default network throughput values for newer EC2 instance types. (#702)
  • Improve error logging for various unsupported operations. (#699)
  • Fix a race condition where calling mknod and forget under the same directory could cause Mountpoint to hang indefinitely. (#711)

Binaries

x86_64

arm64

mountpoint-s3 v1.3.2

11 Jan 15:15
7dcaee0
Compare
Choose a tag to compare

Breaking changes

  • No breaking changes.

Other changes

  • Log messages now include file names and S3 keys more consistently. (#665)
  • Successful mount message is now output to stdout for both foreground and background mode. (#668)
  • Added new metric tracking contiguous reads. This new metric may be used to help understand how much data is being read successfully using prefetching before needing to discard prefetched progress when seeking around the file. (#629)
  • Fix a race condition where FUSE read operations may have completed and subsequently sent back to the Kernel while locks were still being held on a file handle. If a FUSE release operation was executed while the file handle was still held by read, this would result in the file handle never being deallocated. (#691)

Binaries

x86_64

arm64

mountpoint-s3 v1.3.1

30 Nov 17:38
7530462
Compare
Choose a tag to compare

mountpoint-s3 v1.3.0

28 Nov 17:37
99d5277
Compare
Choose a tag to compare

New features

  • Mountpoint now supports resolving S3 Express One Zone endpoints and the new SigV4-Express signing algorithm will be used for S3 Express One Zone buckets. Note that readdir results on these buckets will not be ordered because ListObjectsV2 is unordered on S3 Express. (#642)

Breaking changes

  • No breaking changes.

Other changes

  • New Mountpoint cache directories will be created with owner access only permission. Additionally, the cache directory will be removed entirely at mount time rather than just the contents. (#637)

Binaries

x86_64

arm64

mountpoint-s3 v1.2.0

22 Nov 13:37
211503b
Compare
Choose a tag to compare

mountpoint-s3 v1.1.1

14 Nov 15:43
efc4c0e
Compare
Choose a tag to compare

Breaking changes

  • No breaking changes.

Other changes

  • Some applications that read directory entries out of order (for example, PHP) will now work correctly. (#581)
  • Fixed a bug that caused file creation to fail if a file with the same name had previously been created with Mountpoint and then deleted remotely. (#584)
  • Fixed an issue where Mountpoint could time out or hang on launch if IMDS was not available. (#601)

Binaries

x86_64

arm64