Autodesk 模型衍生翻译作业导致“TranslationWorker-InternalFailure”响应？

Autodesk model derivative translation job results in `TranslationWorker-InternalFailure` response?

我正在尝试使用 Autodesk 的 Model Derivative API 将 .zip 文件转换为 .svf 清单。我能够成功创建一个存储桶，使用分块上传将 zip 放入存储桶中，并使用相应的端点（所有端点 return 200）开始翻译作业。当我来检查作业的翻译进度时，它停在“0% 完成”，最终 return 显示 TranslationWorker-InternalFailure 的失败消息。

这一切都是在 Node.js 和超级代理堆栈上以编程方式完成的，以发出 HTTP 请求。我可以通过使用 Paw 手动翻译来确认文件本身是有效的；并使用我们正在更换的旧工具对其进行翻译。

请参阅下文了解我的代码的具体部分。

Autodesk 翻译作业 (Endpoint)

translate: async function(accessToken, obj) {
    return await request.post('https://developer.api.autodesk.com/modelderivative/v2/designdata/job')
        .set("Authorization", "Bearer " + accessToken)
        .set("Content-Type", "application/json")
        .set('x-ads-force', true)
        .send({
            input: {
                urn: obj.base64Urn,
                compressedUrn: true,
                rootFilename: obj.parentFile
            },
            output: {
                formats: [
                    {
                        type: "thumbnail"
                    },
                    {
                        type: "svf",
                        views: ["2d", "3d"]
                    }
                ]
            }
        });
}

在上面的代码中，变量具有以下值：

Autodesk 翻译进度 (Endpoint)

getTranslationProgressAndDerivatives: function(accessToken, obj) {
    return request.get('https://developer.api.autodesk.com/modelderivative/v2/designdata/' + obj.base64Urn + '/manifest')
        .set('Authorization', "Bearer " + accessToken);
}

当然，obj 有时必须放在 Autodesk 的服务器上。这是通过这段代码完成的：

Autodesk 将对象放入桶中 (Endpoint)

class HttpAutodeskPutObjectWriteStream extends stream.Writable {

    /**
     * Constructor sets all the properties for the instance.
     *
     * @param {string} accessToken - The OAuth2 access token needed for server authentication to engage with Autodesk's API.
     * @param {Obj} obj
     */
    constructor(accessToken, bucket) {
        super();
        this.accessToken = accessToken;
        this.obj = obj;
        this._bytesTransferred = 0;
        this._putHttpUrl = 'https://developer.api.autodesk.com/oss/v2/buckets/' + this.obj.name + '/objects/' + this.obj.name + '/resumable';
    }

    /**
     * Return a bytes transferred statement.
     *
     * @private
     *
     * @param chunk - The chunk currently being transferred.
     *
     * @returns {string}
     */
    _bytesTransferredStatement(chunk) {
        return `bytes ${this._bytesTransferred}-${this._bytesTransferred+chunk.byteLength-1}/${this.obj.zipFileSize}`;
    };

    /**
     * Writes data to the stream. Note the use of the serialize method in the request.
     *
     * @private
     *
     * @param chunk - The chunk currently being transferred.
     * @param encoding - The encoding of the chunk data.
     * @param callback - The function to be called on chunk completion (success or otherwise).
     *
     * @returns {Promise.<void>}
     */
    async _write(chunk, encoding, callback) {
        try {
            let stmt = this._bytesTransferredStatement(chunk);
            this._bytesTransferred += chunk.byteLength;

            let response = await request.put(this._putHttpUrl)
                .set("Authorization", "Bearer " + this.accessToken)
                .set("Session-Id", this.bucket.key)
                .set("Content-Length", chunk.byteLength)
                .set("Content-Range", stmt)
                .serialize(function(d) { return d; })
                .send(chunk.toString());

            if (response.status === 200) {
                this.urn = response.body.objectId;
            }
            callback(null);

        } catch (e) {
            callback(e);
        }
    };
}

然后在此处调用：

put: function(accessToken, bucketEntity) {
    return new Promise(async (resolve, reject) => {
        const maximalChunkedTransferSize = 2*1024*1024; // 2MB (Minimum amount, Autodesk recommends 5MB).

        const objStorageLocation = (await config.get())[process.env.NODE_ENV].objStorageLocation;
        const pathToFile = path.join(__dirname, "../../", objStorageLocation, obj.name + ".zip");

        let readFileStream = fs.createReadStream(pathToFile, { highWaterMark: maximalChunkedTransferSize });

        let ws = new HttpAutodeskPutObjectWriteStream(accessToken, obj);
        readFileStream.pipe(ws);

        ws.on("finish", () => resolve(ws.urn));
        ws.on("error", () => reject());
    });
}

当作业失败时...

一旦失败，这是我从翻译进度端点收到的响应：

{
    "type": "manifest",
    "hasThumbnail": "false",
    "status": "failed",
    "progress": "complete",
    "region": "US",
    "urn": "<redacted>",
    "version": "1.0",
    "derivatives": [
        {
            "name": "LMV Bubble",
            "hasThumbnail": "false",
            "status": "failed",
            "progress": "complete",
            "messages": [
                {
                    "type": "error",
                    "message": "Translation failure",
                    "code": "TranslationWorker-InternalFailure"
                }
            ],
            "outputType": "svf"
        }
    ]
}

我在这里做错了什么会导致尝试翻译文件时 100% 的失败率吗？

当然，在 Whosebug 上发帖会让您从不同的角度审视您的代码。问题出在我上传文件时，特别是最后一个链接方法：

let response = await request.put(this._putHttpUrl)
    .set("Authorization", "Bearer " + this.accessToken)
    .set("Session-Id", this.bucket.key)
    .set("Content-Length", chunk.byteLength)
    .set("Content-Range", stmt)
    .serialize(function(d) { return d; })
    .send(chunk.toString()); // HERE

在块上调用 .toString() 是不正确的。