Failed to get object with error storage: object doesn't exist @ 1.37.0

[chunming-c]

Hi, thank you for the work. However, when upgrade to v1.37.0, object was failed to be retrieved by gcs client. The tricky part is the object does exist and accessible by curl.

curl -I --insecure https://0.0.0.0:4443/storage/v1/b/foo/o/bar.txt | jq .

{
  "kind": "storage#object",
  ...
}

func setup() (gcsClient *storage.Client) {
	transCfg := &http.Transport{
		TLSClientConfig: &tls.Config{InsecureSkipVerify: true},
	}
	httpClient := &http.Client{Transport: transCfg}
	gcsClient, _ = storage.NewClient(
		context.TODO(),
		option.WithEndpoint("https://localhost:4443/storage/v1/"),
		option.WithHTTPClient(httpClient),
	)
	return
}

func main() {
	gcsClient := setup()
	obj := gcsClient.Bucket("foo").Object("bar.txt").ReadCompressed(true)

	_, err := obj.NewReader(context.Background())
	if err != nil {
		log.Fatal(err.Error()) // storage: object doesn't exist
	}
}

Version

• cloud.google.com/go/storage: v1.16.0
• fake-gcs-server: 1.37.0

[fsouza]

Hi @chunming-c, can you share how you started the server? Like which arguments you provided to it.

[chunming-c]

Yeah sure. The server is started by docker compose with config below

version: '3.9'
services:
    fake-gcs-server:
        container_name: fake-gcs-server
        image: fsouza/fake-gcs-server:1.37.0
        ports:
            - 4443:4443
        volumes:
            - ../../test/gcs_test_bucket:/data/foo
        entrypoint: ['/bin/fake-gcs-server', '-data', '/data', '-public-host', 'localhost']

[fsouza]

Can you try setting -public-host to localhost:4443?

[nabeelz7]

We were also seeing similar issue as @chunming-c. Then changed our entrypoint -public-host value to the following per @fsouza tip and now it works again in v1.37 (we use 80/http instead of 4443/https):

entrypoint: ["fake-gcs-server", "-backend=filesystem", "-data=/data", "-scheme=http", "-port=80", "-public-host=cloudstorage:80"]

Is this worth an update to README.md @fsouza ?

[chunming-c]

Yup it works after change, thanks! But what change caused the issue?

[miklosboros]

We had the same issue and I have tracked down the possible change that introduced it. Please, refer to my note at this pull request: PR comment