I'm currently making use of a node.js plugin called s3-upload-stream to stream very large files to Amazon S3. It uses the multipart API and for the most part it works very well.
However, this module is showing its age and I've already had to make modifications to it (the author has deprecated it as well). Today I ran into another issue with Amazon, and I would really like to take the author's recommendation and start using the official aws-sdk to accomplish my uploads.
BUT.
The official SDK does not seem to support piping to s3.upload(). The nature of s3.upload is that you have to pass the readable stream as an argument to the S3 constructor.
I have roughly 120+ user code modules that do various file processing, and they are agnostic to the final destination of their output. The engine hands them a pipeable writeable output stream, and they pipe to it. I cannot hand them an AWS.S3 object and ask them to call upload() on it without adding code to all the modules. The reason I used s3-upload-stream was because it supported piping.
Is there a way to make aws-sdk s3.upload() something I can pipe the stream to?
Wrap the S3 upload() function with the node.js stream.PassThrough() stream.
Here's an example:
inputStream
.pipe(uploadFromStream(s3));
function uploadFromStream(s3) {
var pass = new stream.PassThrough();
var params = {Bucket: BUCKET, Key: KEY, Body: pass};
s3.upload(params, function(err, data) {
console.log(err, data);
});
return pass;
}
In the accepted answer, the function ends before the upload is complete, and thus, it's incorrect. The code below pipes correctly from a readable stream.
async function uploadReadableStream(stream) {
const params = {Bucket: bucket, Key: key, Body: stream};
return s3.upload(params).promise();
}
async function upload() {
const readable = getSomeReadableStream();
const results = await uploadReadableStream(readable);
console.log('upload complete', results);
}
You can also go a step further and output progress info using ManagedUpload as such:
const manager = s3.upload(params);
manager.on('httpUploadProgress', (progress) => {
console.log('progress', progress) // { loaded: 4915, total: 192915, part: 1, key: 'foo.jpg' }
});
A bit late answer, it might help someone else hopefully. You can return both writeable stream and the promise, so you can get response data when upload finishes.
const uploadStream = ({ Bucket, Key }) => {
const s3 = new AWS.S3();
const pass = new stream.PassThrough();
return {
writeStream: pass,
promise: s3.upload({ Bucket, Key, Body: pass }).promise(),
};
}
And you can use the function as follows:
const { writeStream, promise } = uploadStream({Bucket: 'yourbucket', Key: 'yourfile.mp4'});
const readStream = fs.createReadStream('/path/to/yourfile.mp4');
readStream.pipe(writeStream);
promise.then(console.log);
Type Script solution:
This example uses:
import * as AWS from "aws-sdk";
import * as fsExtra from "fs-extra";
import * as zlib from "zlib";
import * as stream from "stream";
And async function:
public async saveFile(filePath: string, s3Bucket: AWS.S3, key: string, bucketName: string): Promise<boolean> {
const uploadStream = (S3: AWS.S3, Bucket: string, Key: string) => {
const passT = new stream.PassThrough();
return {
writeStream: passT,
promise: S3.upload({ Bucket, Key, Body: passT }).promise(),
};
};
const { writeStream, promise } = uploadStream(s3Bucket, bucketName, key);
fsExtra.createReadStream(filePath).pipe(writeStream); // NOTE: Addition You can compress to zip by .pipe(zlib.createGzip()).pipe(writeStream)
let output = true;
await promise.catch((reason)=> { output = false; console.log(reason);});
return output;
}
Call this method somewhere like:
let result = await saveFileToS3(testFilePath, someS3Bucket, someKey, someBucketName);
For those complaining that the when they use the s3 api upload function and a zero byte file ends up on s3 (@Radar155 and @gabo) - I also had this problem.
Create a second PassThrough stream and just pipe all data from the first to the second and pass the reference to that second to s3. You can do this in a couple of different ways - possibly a dirty way is to listen for the "data" event on the first stream and then write that same data to the second stream - the similarly for the "end" event - just call the end function on the second stream. I've no idea whether this is a bug in the aws api, the version of node or some other issue - but it worked around the issue for me.
Here is how it might look:
var PassThroughStream = require('stream').PassThrough;
var srcStream = new PassThroughStream();
var rstream = fs.createReadStream('Learning/stocktest.json');
var sameStream = rstream.pipe(srcStream);
// interesting note: (srcStream == sameStream) at this point
var destStream = new PassThroughStream();
// call your s3.upload function here - passing in the destStream as the Body parameter
srcStream.on('data', function (chunk) {
destStream.write(chunk);
});
srcStream.on('end', function () {
dataStream.end();
});
If it helps anyone I was able to stream from the client to s3 successfully:
https://gist.github.com/mattlockyer/532291b6194f6d9ca40cb82564db9d2a
The serverside code assumes req is a stream object, in my case it was sent from the client with file info set in the headers.
const fileUploadStream = (req, res) => {
//get "body" args from header
const { id, fn } = JSON.parse(req.get('body'));
const Key = id + '/' + fn; //upload to s3 folder "id" with filename === fn
const params = {
Key,
Bucket: bucketName, //set somewhere
Body: req, //req is a stream
};
s3.upload(params, (err, data) => {
if (err) {
res.send('Error Uploading Data: ' + JSON.stringify(err) + '\n' + JSON.stringify(err.stack));
} else {
res.send(Key);
}
});
};
Yes it breaks convention but if you look at the gist it's much cleaner than anything else I found using multer, busboy etc...
+1 for pragmatism and thanks to @SalehenRahman for his help.
I'm using KnexJS and had a problem using their streaming API. I finally fixed it, hopefully the following will help someone.
const knexStream = knex.select('*').from('my_table').stream();
const passThroughStream = new stream.PassThrough();
knexStream.on('data', (chunk) => passThroughStream.write(JSON.stringify(chunk) + '\n'));
knexStream.on('end', () => passThroughStream.end());
const uploadResult = await s3
.upload({
Bucket: 'my-bucket',
Key: 'stream-test.txt',
Body: passThroughStream
})
.promise();
None of the answers worked for me because I wanted to:
- Pipe into
s3.upload() - Pipe the result of
s3.upload()into another stream
The accepted answer doesn't do the latter. The others rely on the promise api, which is cumbersome to work when working with stream pipes.
This is my modification of the accepted answer.
const s3 = new S3();
function writeToS3({Key, Bucket}) {
const Body = new stream.PassThrough();
s3.upload({
Body,
Key,
Bucket: process.env.adpBucket
})
.on('httpUploadProgress', progress => {
console.log('progress', progress);
})
.send((err, data) => {
if (err) {
Body.destroy(err);
} else {
console.log(`File uploaded and available at ${data.Location}`);
Body.destroy();
}
});
return Body;
}
const pipeline = myReadableStream.pipe(writeToS3({Key, Bucket});
pipeline.on('close', () => {
// upload finished, do something else
})
pipeline.on('error', () => {
// upload wasn't successful. Handle it
})If you know the size of the stream you can use minio-js to upload the stream like this:
s3Client.putObject('my-bucketname', 'my-objectname.ogg', stream, size, 'audio/ogg', function(e) {
if (e) {
return console.log(e)
}
console.log("Successfully uploaded the stream")
})
来源：https://stackoverflow.com/questions/37336050/pipe-a-stream-to-s3-upload