* Add azure-storage-blob-go dependency
* Implement Azure BlobStorage store
* Add AzureStore Mock test
* Refactor Blob interfaces to use uppercase fields
* Refactor and remove the Create function
When getting the offset, and we get the status code BlobNotFound, we can say the offset is 0, and start from the beginning
* Update the mock
* Refactor error checking of GetOffset to actually check the service code
* Begin testing azurestore
* Write more tests
* New feature allows to set access type on new containers and blob access tier
* Write more docs
* Upgrade azure-storage-blob-go to v0.13.0
* Remove AzError, not needed
* Update link to container access type information
* Remove ?toc from link in comments
* Remove trailing spaces from workflow
* Run tests with go1.15 and 1.16
* Don't fail fast
This lets all other tests complete, and makes it easier to see if it's just a one-off fail, or on different OSes and versions
* Remove darwin 386 from `build_all.sh` script
Removed in go1.15 https://github.com/golang/go/issues/37610
* Update go version in `Dockerfile`
* Compile for Apple Silicone (darwin arm64)
Only go1.16 supports it
* Add first draft of parallel upload queue
* s3store: Use queue for parallel uploads
* Revert "Add first draft of parallel upload queue"
This reverts commit 86a329cef2.
* Revert "s3store: Use queue for parallel uploads"
This reverts commit 29b59a2c90.
* s3store: Cache results from listing parts and checking incomplete object
* s3store: Remove debugging output`
* s3store: Make requests for fetching info concurrently
* s3store: Make parallel uploads work and tests pass
* s3store: Add semaphore package
* s3store: Add comments to semaphore package
* s3store: Encapsulate more logic into s3PartProducer
* s3store: Refactor WriteChunk
* s3store: Remove TODO
* s3store: Acquire lock before uploading
* cli: Add flag for setting concurrency limit
* s3store: One more comment
* Add TLS support to tusd
* Adds `-tls-certificate`, `-tls-key`, and `-tls-mode` flags
* Alter printed URL to reflect protocol in use
* For non-TLS, do an eary-exit if http.Serve() returns
* Configure TLS for the following modes
* TLSv1.3-only
* (default) TLSv1.3+TLSv1.2, with recommended ciphersuites from Mozilla + matching RSA key transport modes
* TLSv1.2 only, with only 256-bit AES ciphersuites
* All modes disable HTTP/2, given that it’s not supported in non-TLS mode
* Update documentation
* * Remove RSA-based key transport ciphersuites as they don’t support forward secrecy
* Improve the TLS/HTTPS example in the usage documentation
Signed-off-by: Joey Coleman <joey.coleman@kirasystems.com>
* Update docs further to a) record that the key file must be unencrypted, and b) clean up the RSA-based ciphersuite comments.
* Enable S3 transfer acceleration in SDK
* Format better
* Setup for flag changes
* Place feature behind new flag
* Fix a docs issue
Co-authored-by: Cloud User <centos@ip-10-0-0-184.us-west-2.compute.internal>
* Allow empty metadata values
* Make tests less fragile by allowing loose call ordering
* Add s3ChunkProducer
* Integrate s3ChunkProducer to support chunk buffering
* Remove completed chunk files inline to reduce disk space usage
* Add tests for chunk producer
* docs: Use value from Host header to forward to tusd
* Use int64 for MaxBufferedParts field
* Default to 20 buffered parts
* Rename s3ChunkProducer -> s3PartProducer
* Document s3PartProducer struct
* Clarify misleading comment
* Revert "Remove completed chunk files inline to reduce disk space usage"
This reverts commit b72a4d43d6.
* Remove redundant seek
This is already being done in s3PartProducer.
* Clean up any remaining files in the channel when we return
* Make putPart* functions responsible for cleaning up temp files
* handler: Add tests for empty metadata pairs
* Factor out cleanUpTempFile func
* Add test to ensure that temporary files get cleaned up
Co-authored-by: Jens Steinhauser <jens.steinhauser@gmail.com>
Co-authored-by: Marius <marius@transloadit.com>