Merge pull request #187 from coreweave/eta/signature-v4

fix(stream_io): Default to signature version 4 for most object storage endpoints

Author: wbrown

.github/workflows/benchmark.yml

@@ -10,7 +10,7 @@ jobs:
     name: Build & push docker image
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v2
+      - uses: actions/checkout@v4
 
       - name: 'Login to GitHub Container Registry'
         uses: docker/login-action@v1

.github/workflows/build.yml

@@ -7,13 +7,13 @@ on:
 jobs:
   build_dist:
     name: Build distribution on Ubuntu
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
 
-      - uses: actions/setup-python@v3
+      - uses: actions/setup-python@v5
         with:
-          python-version: '3.8'
+          python-version: '3.10'
 
       - name: Install build dependencies
         run: python -m pip install --no-cache-dir -U setuptools build
@@ -28,7 +28,7 @@ jobs:
   upload_dist:
     needs: [ build_dist ]
     name: Upload distribution to PyPI
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     if: github.event_name == 'release' && github.event.action == 'published'
     steps:
       - uses: actions/download-artifact@v3

.github/workflows/test.yml

@@ -5,13 +5,13 @@ on: [push, pull_request]
 jobs:
   run_tests:
     name: Run tests
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
 
-      - uses: actions/setup-python@v3
+      - uses: actions/setup-python@v5
         with:
-          python-version: '3.8'
+          python-version: '3.10'
           cache: 'pip'
 
       - name: Install dependencies

CHANGELOG.md

@@ -5,6 +5,14 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+[2.9.3] - 2025-05-09
+
+### Changed
+
+- `stream_io.open_stream()` now defaults to authenticating with signature
+  version 4 rather than signature version 2 when no `s3_signature_version`
+  is specified for reads from most object storage endpoints
+
 [2.9.2] - 2025-02-20
 
 ### Fixed
@@ -416,6 +424,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   - `get_gpu_name`
   - `no_init_or_tensor`
 
+[2.9.3]: https://github.com/coreweave/tensorizer/compare/v2.9.2...v2.9.3
 [2.9.2]: https://github.com/coreweave/tensorizer/compare/v2.9.1...v2.9.2
 [2.9.1]: https://github.com/coreweave/tensorizer/compare/v2.9.0...v2.9.1
 [2.9.0]: https://github.com/coreweave/tensorizer/compare/v2.8.1...v2.9.0

tensorizer/_version.py

@@ -1 +1 @@
-__version__ = "2.9.2"
+__version__ = "2.9.3"

tensorizer/stream_io.py

@@ -978,8 +978,9 @@ def _s3_download_url(
     s3_signature_version: Optional[str] = None,
 ) -> str:
     bucket, key = _parse_s3_uri(path_uri)
-    # v2 signature is important to easily align the presigned URL expiry
-    # times. This allows multiple clients to generate the exact same
+    # v2 signature is important with accel-object
+    # to easily align the presigned URL expiry times.
+    # This allows multiple clients to generate the exact same
     # presigned URL, and get hits on a HTTP caching proxy.
     #
     # why v2 signature?
@@ -989,7 +990,8 @@ def _s3_download_url(
     # See upstream bug https://github.com/boto/botocore/issues/2230
     #
     if not s3_signature_version:
-        s3_signature_version = "s3"
+        is_accel: bool = "accel-object" in s3_endpoint.lower().split(".")
+        s3_signature_version = "s3" if is_accel else "v4"
     client = _new_s3_client(
         s3_access_key_id,
         s3_secret_access_key,