More efficient copy object (#6665)

* it compiles

* refactored

* reduce to 4 concurrent chunk upload

* CopyObjectPartHandler

* copy a range of the chunk data, fix offset size in copied chunks

* Update s3api_object_handlers_copy.go

What the PR Accomplishes:
CopyObjectHandler - Now copies entire objects by copying chunks individually instead of downloading/uploading the entire file
CopyObjectPartHandler - Handles copying parts of objects for multipart uploads by copying only the relevant chunk portions
Efficient Chunk Copying - Uses direct chunk-to-chunk copying with proper volume assignment and concurrent processing (limited to 4 concurrent operations)
Range Support - Properly handles range-based copying for partial object copies

* fix compilation

* fix part destination

* handling small objects

* use mkFile

* copy to existing file or part

* add testing tools

* adjust tests

* fix chunk lookup

* refactoring

* fix TestObjectCopyRetainingMetadata

* ensure bucket name not conflicting

* fix conditional copying tests

* remove debug messages

* add custom s3 copy tests
This commit is contained in:
Chris Lu
2025-07-11 18:51:32 -07:00
committed by GitHub
parent 4fcbdc1f61
commit d892538d32
11 changed files with 2343 additions and 43 deletions

View File

@@ -18,7 +18,7 @@ full_install: admin-generate
cd weed; go install -tags "elastic gocdk sqlite ydb tarantool tikv rclone"
server: install
weed -v 0 server -s3 -filer -filer.maxMB=64 -volume.max=0 -master.volumeSizeLimitMB=1024 -volume.preStopSeconds=1 -s3.port=8000 -s3.allowEmptyFolder=false -s3.allowDeleteBucketNotEmpty=true -s3.config=./docker/compose/s3.json -metricsPort=9324
weed -v 0 server -s3 -filer -filer.maxMB=64 -volume.max=0 -master.volumeSizeLimitMB=100 -volume.preStopSeconds=1 -s3.port=8000 -s3.allowEmptyFolder=false -s3.allowDeleteBucketNotEmpty=true -s3.config=./docker/compose/s3.json -metricsPort=9324
benchmark: install warp_install
pkill weed || true