Merge pull request 'Fix #204 (full Multipart Uploads semantics)' (#553) from nlnet-task1 into next

Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/553

3 files changed, 276 insertions, 1 deletions

diff --git a/script/test-renumbering.sh b/script/test-renumbering.sh
new file mode 100644
index 00000000..0b1bd320
--- /dev/null
+++ b/script/test-renumbering.sh
@@ -0,0 +1,138 @@
+#!/usr/bin/env bash
+
+: '
+	This script tests part renumbering on an S3 remote (here configured for Minio).
+
+	On Minio:
+
+	The results confirm that if I upload parts with number 1, 4, 5 and 6,
+	they are renumbered to 1, 2, 3 and 4 after CompleteMultipartUpload.
+	Thus, specifying partNumber=4 on a GetObject/HeadObject should return
+	information on the part I originally uploaded with part number
+
+	On S3: not tested
+
+	Sample output (on Minio):
+
+		f07e1404cc527d494242824ded3a616b  part1
+		78974cd4d0f622eb3426ea7cd22f5a1c  part4
+		f9cc379f8baa61645558d9ba7e6351fa  part5
+		1bd2383eebbac1f8e7143575ba5b1f4a  part6
+		Upload ID: 6838b813-d0ca-400b-9d28-ec8b2b5cd004
+		PART 1 ETag: "f07e1404cc527d494242824ded3a616b"
+		PART 4 ETag: "78974cd4d0f622eb3426ea7cd22f5a1c"
+		PART 5 ETag: "f9cc379f8baa61645558d9ba7e6351fa"
+		PART 6 ETag: "1bd2383eebbac1f8e7143575ba5b1f4a"
+		======================================== LIST ====
+		{
+		  "Parts": [
+			{
+			  "PartNumber": 1,
+			  "LastModified": "2023-04-25T10:21:54.350000+00:00",
+			  "ETag": "\"f07e1404cc527d494242824ded3a616b\"",
+			  "Size": 20971520
+			},
+			{
+			  "PartNumber": 4,
+			  "LastModified": "2023-04-25T10:21:54.350000+00:00",
+			  "ETag": "\"78974cd4d0f622eb3426ea7cd22f5a1c\"",
+			  "Size": 20971520
+			},
+			{
+			  "PartNumber": 5,
+			  "LastModified": "2023-04-25T10:21:54.350000+00:00",
+			  "ETag": "\"f9cc379f8baa61645558d9ba7e6351fa\"",
+			  "Size": 20971520
+			},
+			{
+			  "PartNumber": 6,
+			  "LastModified": "2023-04-25T10:21:54.350000+00:00",
+			  "ETag": "\"1bd2383eebbac1f8e7143575ba5b1f4a\"",
+			  "Size": 20971520
+			}
+		  ],
+		  "ChecksumAlgorithm": "",
+		  "Initiator": {
+			"ID": "02d6176db174dc93cb1b899f7c6078f08654445fe8cf1b6ce98d8855f66bdbf4",
+			"DisplayName": "02d6176db174dc93cb1b899f7c6078f08654445fe8cf1b6ce98d8855f66bdbf4"
+		  },
+		  "Owner": {
+			"DisplayName": "02d6176db174dc93cb1b899f7c6078f08654445fe8cf1b6ce98d8855f66bdbf4",
+			"ID": "02d6176db174dc93cb1b899f7c6078f08654445fe8cf1b6ce98d8855f66bdbf4"
+		  },
+		  "StorageClass": "STANDARD"
+		}
+		======================================== COMPLETE ====
+		{
+			"Location": "http://localhost:9000/test/upload",
+			"Bucket": "test",
+			"Key": "upload",
+			"ETag": "\"8e817c8ccd442f9a79c77b58fe808c43-4\""
+		}
+		======================================== LIST ====
+
+		An error occurred (NoSuchUpload) when calling the ListParts operation: The specified multipart upload does not exist. The upload ID may be invalid, or the upload may have been aborted or completed.
+		======================================== GET PART 4 ====
+		{
+			"AcceptRanges": "bytes",
+			"LastModified": "2023-04-25T10:21:59+00:00",
+			"ContentLength": 20971520,
+			"ETag": "\"8e817c8ccd442f9a79c77b58fe808c43-4\"",
+			"ContentRange": "bytes 62914560-83886079/83886080",
+			"ContentType": "binary/octet-stream",
+			"Metadata": {},
+			"PartsCount": 4
+		}
+		1bd2383eebbac1f8e7143575ba5b1f4a  get-part4
+
+	
+	Conclusions:
+
+	- Parts are indeed renumbered with consecutive numbers
+	- ListParts only applies to multipart uploads in progress,
+	  it cannot be used once the multipart upload has been completed
+'
+
+export AWS_ACCESS_KEY_ID=1D8Pk2k4oQSoh1BU
+export AWS_SECRET_ACCESS_KEY=4B46SR8U7FUgY0raB8Zuxg1NLyLTvbNV
+
+function aws { command aws --endpoint-url http://localhost:9000 $@ ; }
+
+aws --version
+
+aws s3 mb s3://test
+
+for NUM in 1 4 5 6; do
+	dd if=/dev/urandom of=part$NUM bs=1M count=10
+done
+md5sum part*
+
+UPLOAD=$(aws s3api create-multipart-upload --bucket test --key 'upload' | jq -r ".UploadId")
+echo "Upload ID: $UPLOAD"
+
+PARTS=""
+
+for NUM in 1 4 5 6; do
+	ETAG=$(aws s3api upload-part --bucket test --key 'upload' --part-number $NUM \
+		--body "part$NUM" --upload-id "$UPLOAD" | jq -r ".ETag")
+	echo "PART $NUM ETag: $ETAG"
+	if [ -n "$PARTS" ]; then
+		PARTS="$PARTS,"
+	fi
+	PARTS="$PARTS {\"ETag\":$ETAG,\"PartNumber\":$NUM}"
+done
+
+echo "======================================== LIST ===="
+aws s3api list-parts --bucket test --key upload --upload-id "$UPLOAD" | jq
+
+echo "======================================== COMPLETE ===="
+echo "{\"Parts\":[$PARTS]}" > mpu
+aws s3api complete-multipart-upload --multipart-upload file://mpu \
+		--bucket test --key 'upload' --upload-id "$UPLOAD"
+
+echo "======================================== LIST ===="
+aws s3api list-parts --bucket test --key upload --upload-id "$UPLOAD" | jq
+
+echo "======================================== GET PART 4 ===="
+aws s3api get-object --bucket test --key upload --part-number 4 get-part4
+md5sum get-part4
diff --git a/script/test-skip-part.sh b/script/test-skip-part.sh
new file mode 100644
index 00000000..20ae017d
--- /dev/null
+++ b/script/test-skip-part.sh
@@ -0,0 +1,103 @@
+#!/usr/bin/env bash
+
+: '
+	This script tests whether uploaded parts can be skipped in a
+	CompleteMultipartUpoad
+
+	On Minio: yes, parts can be skipped
+
+	On S3: not tested
+
+	Sample output (on Minio):
+
+		f23911bcd1230f5ebe8887cbf5bc396e  part1
+		a2657143167eaf647c40473e78a091dc  part4
+		72f72c02c5163bc81024b28ac818c5e0  part5
+		e29cf500d20498218904b8df8806caa2  part6
+		Upload ID: e8fe7b83-9800-46fb-ae90-9d7ccd42fe76
+		PART 1 ETag: "f23911bcd1230f5ebe8887cbf5bc396e"
+		PART 4 ETag: "a2657143167eaf647c40473e78a091dc"
+		PART 5 ETag: "72f72c02c5163bc81024b28ac818c5e0"
+		PART 6 ETag: "e29cf500d20498218904b8df8806caa2"
+		======================================== COMPLETE ====
+		{
+			"Location": "http://localhost:9000/test/upload",
+			"Bucket": "test",
+			"Key": "upload",
+			"ETag": "\"48246e44d4b38bdc2f3c10ee25b1af17-3\""
+		}
+		======================================== GET FULL ====
+		{
+			"AcceptRanges": "bytes",
+			"LastModified": "2023-04-25T10:54:35+00:00",
+			"ContentLength": 31457280,
+			"ETag": "\"48246e44d4b38bdc2f3c10ee25b1af17-3\"",
+			"ContentType": "binary/octet-stream",
+			"Metadata": {}
+		}
+		97fb904da7ad310699a6afab0eb6e061  get-full
+		97fb904da7ad310699a6afab0eb6e061  -
+		======================================== GET PART 3 ====
+		{
+			"AcceptRanges": "bytes",
+			"LastModified": "2023-04-25T10:54:35+00:00",
+			"ContentLength": 10485760,
+			"ETag": "\"48246e44d4b38bdc2f3c10ee25b1af17-3\"",
+			"ContentRange": "bytes 20971520-31457279/31457280",
+			"ContentType": "binary/octet-stream",
+			"Metadata": {},
+			"PartsCount": 3
+		}
+		e29cf500d20498218904b8df8806caa2  get-part3
+
+	Conclusions:
+
+	- Skipping a part in a CompleteMultipartUpoad call is OK
+	- The part is simply not included in the stored object
+	- Sequential part renumbering counts only non-skipped parts
+'
+
+export AWS_ACCESS_KEY_ID=1D8Pk2k4oQSoh1BU
+export AWS_SECRET_ACCESS_KEY=4B46SR8U7FUgY0raB8Zuxg1NLyLTvbNV
+
+function aws { command aws --endpoint-url http://localhost:9000 $@ ; }
+
+aws --version
+
+aws s3 mb s3://test
+
+for NUM in 1 4 5 6; do
+	dd if=/dev/urandom of=part$NUM bs=1M count=10
+done
+md5sum part*
+
+UPLOAD=$(aws s3api create-multipart-upload --bucket test --key 'upload' | jq -r ".UploadId")
+echo "Upload ID: $UPLOAD"
+
+PARTS=""
+
+for NUM in 1 4 5 6; do
+	ETAG=$(aws s3api upload-part --bucket test --key 'upload' --part-number $NUM \
+		--body "part$NUM" --upload-id "$UPLOAD" | jq -r ".ETag")
+	echo "PART $NUM ETag: $ETAG"
+	if [ "$NUM" != "5" ]; then
+		if [ -n "$PARTS" ]; then
+			PARTS="$PARTS,"
+		fi
+		PARTS="$PARTS {\"ETag\":$ETAG,\"PartNumber\":$NUM}"
+	fi
+done
+
+echo "======================================== COMPLETE ===="
+echo "{\"Parts\":[$PARTS]}" > mpu
+aws s3api complete-multipart-upload --multipart-upload file://mpu \
+		--bucket test --key 'upload' --upload-id "$UPLOAD"
+
+echo "======================================== GET FULL ===="
+aws s3api get-object --bucket test --key upload get-full
+md5sum get-full
+cat part1 part4 part6 | md5sum
+
+echo "======================================== GET PART 3 ===="
+aws s3api get-object --bucket test --key upload --part-number 3 get-part3
+md5sum get-part3
diff --git a/script/test-smoke.sh b/script/test-smoke.sh
index eababf38..6965c0f3 100755
--- a/script/test-smoke.sh
+++ b/script/test-smoke.sh
@@ -31,6 +31,11 @@ dd if=/dev/urandom of=/tmp/garage.1.rnd bs=1k count=2 # No multipart, inline sto
 dd if=/dev/urandom of=/tmp/garage.2.rnd bs=1M count=5 # No multipart but file will be chunked
 dd if=/dev/urandom of=/tmp/garage.3.rnd bs=1M count=10 # by default, AWS starts using multipart at 8MB
 
+dd if=/dev/urandom of=/tmp/garage.part1.rnd bs=1M count=5
+dd if=/dev/urandom of=/tmp/garage.part2.rnd bs=1M count=5
+dd if=/dev/urandom of=/tmp/garage.part3.rnd bs=1M count=5
+dd if=/dev/urandom of=/tmp/garage.part4.rnd bs=1M count=5
+
 # data of lower entropy, to test compression
 dd if=/dev/urandom bs=1k count=2  | base64 -w0 > /tmp/garage.1.b64
 dd if=/dev/urandom bs=1M count=5  | base64 -w0 > /tmp/garage.2.b64
@@ -40,7 +45,7 @@ echo "🧪 S3 API testing..."
 
 # AWS
 if [ -z "$SKIP_AWS" ]; then
-  echo "🛠️ Testing with awscli"
+  echo "🛠️ Testing with awscli (aws s3)"
   source ${SCRIPT_FOLDER}/dev-env-aws.sh
   aws s3 ls
   for idx in {1..3}.{rnd,b64}; do
@@ -51,8 +56,36 @@ if [ -z "$SKIP_AWS" ]; then
     rm /tmp/garage.$idx.dl
     aws s3 rm "s3://eprouvette/&+-é\"/garage.$idx.aws"
   done
+
+  echo "🛠️ Testing multipart uploads with awscli (aws s3api)"
+  UPLOAD=$(aws s3api create-multipart-upload --bucket eprouvette --key 'upload' | jq -r ".UploadId")
+  echo "Upload ID: $UPLOAD"
+  ETAG3=$(aws s3api upload-part --bucket eprouvette --key 'upload' \
+      --part-number 3 --body "/tmp/garage.part1.rnd" --upload-id "$UPLOAD" \
+      | jq -r ".ETag")
+  ETAG2=$(aws s3api upload-part --bucket eprouvette --key 'upload' \
+      --part-number 2 --body "/tmp/garage.part2.rnd" --upload-id "$UPLOAD" \
+      | jq -r ".ETag")
+  ETAG3=$(aws s3api upload-part --bucket eprouvette --key 'upload' \
+      --part-number 3 --body "/tmp/garage.part3.rnd" --upload-id "$UPLOAD" \
+      | jq -r ".ETag")
+  ETAG6=$(aws s3api upload-part --bucket eprouvette --key 'upload' \
+      --part-number 6 --body "/tmp/garage.part4.rnd" --upload-id "$UPLOAD" \
+      | jq -r ".ETag")
+  MPU="{\"Parts\":[{\"PartNumber\":2,\"ETag\":$ETAG2}, {\"PartNumber\":3,\"ETag\":$ETAG3}, {\"PartNumber\":6,\"ETag\":$ETAG6}]}"
+  echo $MPU > /tmp/garage.mpu.json
+  aws s3api complete-multipart-upload --multipart-upload file:///tmp/garage.mpu.json \
+      --bucket eprouvette --key 'upload' --upload-id "$UPLOAD"
+  aws s3api get-object --bucket eprouvette --key upload /tmp/garage.mpu.get
+  if [ "$(md5sum /tmp/garage.mpu.get | cut -d ' ' -f 1)" != "$(cat /tmp/garage.part{2,3,4}.rnd | md5sum | cut -d ' ' -f 1)" ]; then
+    echo "Invalid multipart upload"
+    exit 1
+  fi
 fi
 
+echo "OK!!"
+exit 0
+
 # S3CMD
 if [ -z "$SKIP_S3CMD" ]; then
   echo "🛠️ Testing with s3cmd"
@@ -141,6 +174,7 @@ rm eprouvette/winscp
 EOF
 fi
 
+rm /tmp/garage.part{1..4}.rnd
 rm /tmp/garage.{1..3}.{rnd,b64}
 
 echo "🏁 Teardown"