GCS multipart upload behaviour

[larshagencognite]

I have some questions regarding the multipart upload behaviour for GCS.

I am working on getting FoundationDB to write backups in GCS through the S3Proxy. The backup has previously gone through the minio gateway, which is now deprecated, so I looking for a replacement.

When S3Proxy gets a multi-part PUT request, the debug log shows

[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.gaul.s3proxy.S3ProxyHandler:301 |::] request: Request(PUT http://<s3proxy-endpoint>/<bucket>/<path>?partNumber=5&uploadId=<uploadId>)@26638b8f
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.gaul.s3proxy.S3ProxyHandler:326 |::] header: Authorization: AWS <identity:credential>
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.gaul.s3proxy.S3ProxyHandler:326 |::] header: Accept: application/xml
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.gaul.s3proxy.S3ProxyHandler:326 |::] header: Host: <s3proxy-endpoint>
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.gaul.s3proxy.S3ProxyHandler:326 |::] header: Content-Length: 5242880
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.gaul.s3proxy.S3ProxyHandler:326 |::] header: Date: 20230818T110456Z
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.gaul.s3proxy.S3ProxyHandler:326 |::] header: Content-MD5: <md5>
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.j.r.i.InvokeHttpMethod:56 |::] >> invoking Object:get
[s3proxy] D 08-18 11:04:56.100 S3Proxy-Jetty-48 o.j.h.i.JavaUrlHttpCommandExecutorService:56 |::] Sending request 753038757: GET https://storage.googleapis.com/storage/v1/b/<bucket>/o/<uploadId>_00000001 HTTP/1.1
[s3proxy] D 08-18 11:04:56.101 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> GET https://storage.googleapis.com/storage/v1/b/<bucket>/o/<uploadId>_00000001 HTTP/1.1
[s3proxy] D 08-18 11:04:56.101 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> Accept: application/json
[s3proxy] D 08-18 11:04:56.118 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> Authorization: Bearer <token>
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 o.j.h.i.JavaUrlHttpCommandExecutorService:56 |::] Receiving response 753038757: HTTP/1.1 404 Not Found
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << HTTP/1.1 404 Not Found
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Server: UploadServer
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << X-GUploader-UploadID: <uploadId>
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Vary: X-Origin
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Vary: Origin
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Pragma: no-cache
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Date: Fri, 18 Aug 2023 11:04:56 GMT
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Cache-Control: no-cache, no-store, max-age=0, must-revalidate
[s3proxy] D 08-18 11:04:56.198 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Content-Type: application/json; charset=UTF-8
[s3proxy] D 08-18 11:04:56.199 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Content-Length: 383
[s3proxy] D 08-18 11:04:56.199 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Expires: Mon Jan 01 00:00:00 UTC 1990
[s3proxy] D 08-18 11:04:56.219 S3Proxy-Jetty-48 o.j.r.i.InvokeHttpMethod:56 |::] >> invoking Object:simpleUpload
[s3proxy] D 08-18 11:04:56.219 S3Proxy-Jetty-48 o.j.h.i.JavaUrlHttpCommandExecutorService:56 |::] Sending request 1133015402: POST https://storage.googleapis.com/upload/storage/v1/b/<bucket>/o?uploadType=media&name=<uploadId>_00000001_00000005 HTTP/1.1
[s3proxy] D 08-18 11:04:56.231 S3Proxy-Jetty-48 o.j.http.internal.HttpWire:56 |::] over limit 5242880/262144: wrote temp file
[s3proxy] D 08-18 11:05:56.021 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> POST https://storage.googleapis.com/upload/storage/v1/b/<bucket>/o?uploadType=media&name=<uploadId>_00000001_00000005 HTTP/1.1
[s3proxy] D 08-18 11:05:56.021 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> Accept: application/json
[s3proxy] D 08-18 11:05:56.024 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> Authorization: Bearer <token>
[s3proxy] D 08-18 11:05:56.030 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> Content-Type: application/unknown
[s3proxy] D 08-18 11:05:56.032 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> Content-Length: 5242880
[s3proxy] D 08-18 11:05:56.036 S3Proxy-Jetty-48 jclouds.headers:56 |::] >> Content-MD5: lB8VF7jGtzaAuI0Rp/yPMg==
[s3proxy] D 08-18 11:05:56.492 S3Proxy-Jetty-48 o.j.h.i.JavaUrlHttpCommandExecutorService:56 |::] Receiving response -56785471: HTTP/1.1 200 OK
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << HTTP/1.1 200 OK
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Server: UploadServer
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << ETag: <etag>
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << X-GUploader-UploadID: <uploadId>
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Vary: X-Origin
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Vary: Origin
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Pragma: no-cache
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Date: Fri, 18 Aug 2023 11:05:56 GMT
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Cache-Control: no-cache, no-store, max-age=0, must-revalidate
[s3proxy] D 08-18 11:05:56.493 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Content-Type: application/json; charset=UTF-8
[s3proxy] D 08-18 11:05:56.500 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Content-Length: 1045
[s3proxy] D 08-18 11:05:56.503 S3Proxy-Jetty-48 jclouds.headers:56 |::] << Expires: Mon Jan 01 00:00:00 UTC 1990

I don't get the point of the GET call, as it seems to be calling it on <uploadId>_00000001, and after it gets a 404 on the GET it calls POST on <uploadId>_00000001_00000005, presumably creating a new object for the part.

Are these parts later combined so that one can read the whole object?

The backup is currently failing to make progress, and I am trying to find out if it is related to this.