S3: add object versioning (#6945)

* add object versioning

* add missing file

* Update weed/s3api/s3api_object_versioning.go

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update weed/s3api/s3api_object_versioning.go

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update weed/s3api/s3api_object_versioning.go

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* ListObjectVersionsResult is better to show multiple version entries

* fix test

* Update weed/s3api/s3api_object_handlers_put.go

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update weed/s3api/s3api_object_versioning.go

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* multiple improvements

* move PutBucketVersioningHandler into weed/s3api/s3api_bucket_handlers.go file
* duplicated code for reading bucket config, versioningEnabled, etc. try to use functions
* opportunity to cache bucket config

* error handling if bucket is not found

* in case bucket is not found

* fix build

* add object versioning tests

* remove non-existent tests

* add tests

* add versioning tests

* skip a new test

* ensure .versions directory exists before saving info into it

* fix creating version entry

* logging on creating version directory

* Update s3api_object_versioning_test.go

* retry and wait for directory creation

* revert add more logging

* Update s3api_object_versioning.go

* more debug messages

* clean up logs, and touch directory correctly

* log the .versions creation and then parent directory listing

* use mkFile instead of touch

touch is for update

* clean up data

* add versioning test in go

* change location

* if modified, latest version is moved to .versions directory, and create a new latest version

 Core versioning functionality: WORKING
TestVersioningBasicWorkflow - PASS
TestVersioningDeleteMarkers - PASS
TestVersioningMultipleVersionsSameObject - PASS
TestVersioningDeleteAndRecreate - PASS
TestVersioningListWithPagination - PASS
 Some advanced features still failing:
ETag calculation issues (using mtime instead of proper MD5)
Specific version retrieval (EOF error)
Version deletion (internal errors)
Concurrent operations (race conditions)

* calculate multi chunk md5

Test Results - All Passing:
 TestBucketListReturnDataVersioning - PASS
 TestVersioningCreateObjectsInOrder - PASS
 TestVersioningBasicWorkflow - PASS
 TestVersioningMultipleVersionsSameObject - PASS
 TestVersioningDeleteMarkers - PASS

* dedupe

* fix TestVersioningErrorCases

* fix eof error of reading old versions

* get specific version also check current version

* enable integration tests for versioning

* trigger action to work for now

* Fix GitHub Actions S3 versioning tests workflow

- Fix syntax error (incorrect indentation)
- Update directory paths from weed/s3api/versioning_tests/ to test/s3/versioning/
- Add push trigger for add-object-versioning branch to enable CI during development
- Update artifact paths to match correct directory structure

* Improve CI robustness for S3 versioning tests

Makefile improvements:
- Increase server startup timeout from 30s to 90s for CI environments
- Add progressive timeout reporting (logs at 30s, full logs at 90s)
- Better error handling with server logs on failure
- Add server PID tracking for debugging
- Improved test failure reporting

GitHub Actions workflow improvements:
- Increase job timeouts to account for CI environment delays
- Add system information logging (memory, disk space)
- Add detailed failure reporting with server logs
- Add process and network diagnostics on failure
- Better error messaging and log collection

These changes should resolve the 'Server failed to start within 30 seconds' issue
that was causing the CI tests to fail.

* adjust testing volume size

* Update Makefile

* Update Makefile

* Update Makefile

* Update Makefile

* Update s3-versioning-tests.yml

* Update s3api_object_versioning.go

* Update Makefile

* do not clean up

* log received version id

* more logs

* printout response

* print out list version response

* use tmp files when put versioned object

* change to versions folder layout

* Delete weed-test.log

* test with mixed versioned and unversioned objects

* remove versionDirCache

* remove unused functions

* remove unused function

* remove fallback checking

* minor

---------

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
This commit is contained in:
Chris Lu
2025-07-09 01:51:45 -07:00
committed by GitHub
parent 8fa1a69f8c
commit cf5a24983a
18 changed files with 2880 additions and 86 deletions

View File

@@ -3,14 +3,15 @@ package s3api
import (
"bytes"
"fmt"
"github.com/seaweedfs/seaweedfs/weed/filer"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"io"
"net/http"
"net/url"
"strings"
"time"
"github.com/seaweedfs/seaweedfs/weed/filer"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3_constants"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3err"
"github.com/seaweedfs/seaweedfs/weed/util/mem"
@@ -120,7 +121,73 @@ func (s3a *S3ApiServer) GetObjectHandler(w http.ResponseWriter, r *http.Request)
return
}
destUrl := s3a.toFilerUrl(bucket, object)
// Check for specific version ID in query parameters
versionId := r.URL.Query().Get("versionId")
// Check if versioning is enabled for the bucket
versioningEnabled, err := s3a.isVersioningEnabled(bucket)
if err != nil {
if err == filer_pb.ErrNotFound {
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchBucket)
return
}
glog.Errorf("Error checking versioning status for bucket %s: %v", bucket, err)
s3err.WriteErrorResponse(w, r, s3err.ErrInternalError)
return
}
var destUrl string
if versioningEnabled {
// Handle versioned GET - all versions are stored in .versions directory
var targetVersionId string
var entry *filer_pb.Entry
if versionId != "" {
// Request for specific version
glog.V(2).Infof("GetObject: requesting specific version %s for %s/%s", versionId, bucket, object)
entry, err = s3a.getSpecificObjectVersion(bucket, object, versionId)
if err != nil {
glog.Errorf("Failed to get specific version %s: %v", versionId, err)
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchKey)
return
}
targetVersionId = versionId
} else {
// Request for latest version
glog.V(2).Infof("GetObject: requesting latest version for %s/%s", bucket, object)
entry, err = s3a.getLatestObjectVersion(bucket, object)
if err != nil {
glog.Errorf("Failed to get latest version: %v", err)
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchKey)
return
}
if entry.Extended != nil {
if versionIdBytes, exists := entry.Extended[s3_constants.ExtVersionIdKey]; exists {
targetVersionId = string(versionIdBytes)
}
}
}
// Check if this is a delete marker
if entry.Extended != nil {
if deleteMarker, exists := entry.Extended[s3_constants.ExtDeleteMarkerKey]; exists && string(deleteMarker) == "true" {
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchKey)
return
}
}
// All versions are stored in .versions directory
versionObjectPath := object + ".versions/" + s3a.getVersionFileName(targetVersionId)
destUrl = s3a.toFilerUrl(bucket, versionObjectPath)
glog.V(2).Infof("GetObject: version %s URL: %s", targetVersionId, destUrl)
// Set version ID in response header
w.Header().Set("x-amz-version-id", targetVersionId)
} else {
// Handle regular GET (non-versioned)
destUrl = s3a.toFilerUrl(bucket, object)
}
s3a.proxyToFiler(w, r, destUrl, false, passThroughResponse)
}
@@ -130,7 +197,73 @@ func (s3a *S3ApiServer) HeadObjectHandler(w http.ResponseWriter, r *http.Request
bucket, object := s3_constants.GetBucketAndObject(r)
glog.V(3).Infof("HeadObjectHandler %s %s", bucket, object)
destUrl := s3a.toFilerUrl(bucket, object)
// Check for specific version ID in query parameters
versionId := r.URL.Query().Get("versionId")
// Check if versioning is enabled for the bucket
versioningEnabled, err := s3a.isVersioningEnabled(bucket)
if err != nil {
if err == filer_pb.ErrNotFound {
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchBucket)
return
}
glog.Errorf("Error checking versioning status for bucket %s: %v", bucket, err)
s3err.WriteErrorResponse(w, r, s3err.ErrInternalError)
return
}
var destUrl string
if versioningEnabled {
// Handle versioned HEAD - all versions are stored in .versions directory
var targetVersionId string
var entry *filer_pb.Entry
if versionId != "" {
// Request for specific version
glog.V(2).Infof("HeadObject: requesting specific version %s for %s/%s", versionId, bucket, object)
entry, err = s3a.getSpecificObjectVersion(bucket, object, versionId)
if err != nil {
glog.Errorf("Failed to get specific version %s: %v", versionId, err)
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchKey)
return
}
targetVersionId = versionId
} else {
// Request for latest version
glog.V(2).Infof("HeadObject: requesting latest version for %s/%s", bucket, object)
entry, err = s3a.getLatestObjectVersion(bucket, object)
if err != nil {
glog.Errorf("Failed to get latest version: %v", err)
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchKey)
return
}
if entry.Extended != nil {
if versionIdBytes, exists := entry.Extended[s3_constants.ExtVersionIdKey]; exists {
targetVersionId = string(versionIdBytes)
}
}
}
// Check if this is a delete marker
if entry.Extended != nil {
if deleteMarker, exists := entry.Extended[s3_constants.ExtDeleteMarkerKey]; exists && string(deleteMarker) == "true" {
s3err.WriteErrorResponse(w, r, s3err.ErrNoSuchKey)
return
}
}
// All versions are stored in .versions directory
versionObjectPath := object + ".versions/" + s3a.getVersionFileName(targetVersionId)
destUrl = s3a.toFilerUrl(bucket, versionObjectPath)
glog.V(2).Infof("HeadObject: version %s URL: %s", targetVersionId, destUrl)
// Set version ID in response header
w.Header().Set("x-amz-version-id", targetVersionId)
} else {
// Handle regular HEAD (non-versioned)
destUrl = s3a.toFilerUrl(bucket, object)
}
s3a.proxyToFiler(w, r, destUrl, false, passThroughResponse)
}