SvelteKit S3 Multipart Upload: Video Cloud Storage # SvelteKit S3 Multipart Upload: Video Cloud Storage #

⚙️ SvelteKit S3 Multipart Upload: Getting Started #

Instead of building everything from scratch, we will use the previous tutorial on SvelteKit S3 Compatible Storage Uploads as a starting point. You can start here and check out the other tutorial another day, although multipart S3 uploads might make more sense if you start with the other tutorial. If you did work through the pre-signed URL upload tutorial, you can create a new branch in your repo and carry on from your existing code. Otherwise, clone the following repo to get going:

    
git clone https://github.com/rodneylab/sveltekit-s3-compatible-storage.git sveltekit-s3-multipart-upload
cd sveltekit-s3-multipart-upload
pnpm install

  

We won't need to add any extra packages beyond the ones we used last time.

🔨 Utility Functions #

With multipart uploads, the pre-signed URL part works much as it did for a single upload. The workflow is a little different though. We will still keep the single file upload code and only use this when the file is small. With a multipart upload, we need to create a signed URL for each part we need to upload. Another difference is that once we have uploaded all the parts to their respective URLs, we then we need to tell the provider we are done. This is so that they can combine the pieces at their end. For this to work, we need to add a few more utility functions to our src/lib/utilities.js file. On top, we will be restructuring our app slightly, so need to export some of the existing functions.

To get going, let us import a few extra functions from the S3 SDK. Remember, although we are using the S3 SDK, we can expect our code to work with any S3 compatible provider (recalling only the initial authorization step will vary from provider to provider).

    
1 import {
2   CompleteMultipartUploadCommand,
3   CreateMultipartUploadCommand,
4   GetObjectCommand,
5   PutObjectCommand,
6   S3,
7   UploadPartCommand,
8 } from '@aws-sdk/client-s3';

  

Continuing, in line 18, export the authoriseAccount function because we will want to access it from our SvelteKit endpoint:

    
18 export async function authoriseAccount() {

  

Multipart Upload Functions #

Next, we have to create the function which tells the provider we are done uploading. Add this code to the same file:

    
62 export async function completeMultipartUpload({ parts, client, key, uploadId }) {
63   try {
64     const { VersionId: id } = await client.send(
65       new CompleteMultipartUploadCommand({
66         Key: key,
67         Bucket: S3_COMPATIBLE_BUCKET_NAME,
68         MultipartUpload: { Parts: parts },
69         UploadId: uploadId,
70       }),
71     );
72     if (id) {
73       return { successful: true, id };
74     }
75   } catch (error) {
76     console.error('Error in completing multipart upload: ', error);
77   }
78   return { successful: false };
79 }

  

As with authoriseAccount, we will need to export getS3Client:

    
85 export function getS3Client({ s3ApiUrl }) {

  

Next, we want a function to generate pre-signed URLs. This works just like the function we had for single file upload pre-signed URLs:

    
100 export async function generatePresignedPartUrls({ client, key, uploadId, partCount }) {
101   const signer = new S3RequestPresigner({ ...client.config });
102   const createRequestPromises = [];
103  
104   for (let index = 0; index < partCount; index += 1) {
105     createRequestPromises.push(
106       createRequest(
107         client,
108         new UploadPartCommand({
109           Key: key,
110           Bucket: S3_COMPATIBLE_BUCKET_NAME,
111           UploadId: uploadId,
112           PartNumber: index + 1,
113         }),
114       ),
115     );
116   }
117  
118   const uploadPartRequestResults = await Promise.all(createRequestPromises);
119  
120   const presignPromises = [];
121   uploadPartRequestResults.forEach((element) => presignPromises.push(signer.presign(element)));
122   const presignPromiseResults = await Promise.all(presignPromises);
123   return presignPromiseResults.map((element) => formatUrl(element));
124 }

  

Talking of the single upload, the generatePresignedUrls function needs exporting too:

    
126 export async function generatePresignedUrls({ key, s3ApiUrl }) {</CodeFragment>

  

Lastly, we will create a function to initiate a multipart upload using the S3 SDK:

    
139 export const initiateMultipartUpload = async ({ client, key }) => {
140   const { UploadId: uploadId } = await client.send(
141     new CreateMultipartUploadCommand({ Key: key, Bucket: S3_COMPATIBLE_BUCKET_NAME }),
142   );
143   return uploadId;
144 };

  

That was a lot of pasting! Do not worry if it is not 100% clear what we are doing yet, We will start to pull everything together in the next section where we call these functions from our endpoint.

📹 Multipart Pre‑signed Upload Endpoint #

You might remember, from our SvelteKit frontend, we called an endpoint to tell us the pre-signed URL to upload the file to. Once we had that URL back, we proceeded with the upload directly from the frontend to the cloud provider. With multipart uploads, our ambition is again to upload directly from the frontend to our provider. For this to work, we will change the logic in the endpoint.

We will pass the file size to the endpoint when we request the pre-signed upload URLs. Based on the file size, our logic will decide whether we will do a single file or multipart upload. When we create an S3 client object, we get back some parameters from the provider which give us minimum, maximum and recommended file part size. So, to look at a concrete example. Let's say we want to upload a 16 MB video and the recommended part size is 5 MB. In this case, we will need four parts: the first 3 parts will be 5 MB and the final one, 1 MB. Typically, the minimum part size is not enforced by the provider for the final part in a multipart upload.

Now we know what we are doing, let's get coding!

SvelteKit S3 Multipart Upload: presigned‑urls.json Endpoint Code #

This is a substantial refactor on the previous code for the file at src/routes/api/presigned-urls.json/+server.js:

    
1 import {
2   authoriseAccount,
3   generatePresignedPartUrls,
4   getS3Client,
5   initiateMultipartUpload,
6   presignedUrls,
7 } from '$lib/utilities/storage';
8 
9 export async function POST({ request, setHeaders }) {
10   const { key, size } = await request.json();
11 
12   try {
13     const { absoluteMinimumPartSize, recommendedPartSize, s3ApiUrl } = await authoriseAccount();
14     if (s3ApiUrl) {
15       const client = getS3Client({ s3ApiUrl });
16       if (absoluteMinimumPartSize && size > absoluteMinimumPartSize) {
17         const uploadId = await initiateMultipartUpload({ client, key });
18         if (recommendedPartSize) {
19           const partSize =
20             size < recommendedPartSize ? absoluteMinimumPartSize : recommendedPartSize;
21           const partCount = Math.ceil(size / partSize);
22           if (uploadId) {
23             const multipartUploadUrls = await generatePresignedPartUrls({
24               client,
25               key,
26               uploadId,
27               partCount,
28             });
29 
30             const { readSignedUrl, writeSignedUrl } = await presignedUrls(key);
31 
32             setHeaders({
33               'Content-Type': 'application/json',
34             });
35 
36             return new Response(
37               JSON.stringify({
38                 multipartUploadUrls,
39                 partCount,
40                 partSize,
41                 readSignedUrl,
42                 writeSignedUrl,
43                 uploadId,
44               }),
45             );
46           }
47         }
48       }
49 
50       const { readSignedUrl, writeSignedUrl } = await presignedUrls(key);
51 
52       setHeaders({
53         'Content-Type': 'application/json',
54       });
55 
56       return new Response(
57         JSON.stringify({
58           partCount: 1,
59           readSignedUrl,
60           writeSignedUrl,
61         }),
62       );
63     }
64   } catch (error) {
65     const message = `Error in route api/presigned-urls.json: ${error}`;
66     console.error(message);
67     throw new Error(message);
68   }
69 }

  

At the top of the file, you can see, we now import the functions we have just exported from the utilities file. In line 13, we get the file size parameters we talked about. We use them in line 16 to work out if we will do a multipart upload or single. For a single upload we jump to line 50 and the code is not too different to what we had last time. We just add a partCount field in the response, to let the front end code know we only have one part (line 53).

For multipart uploads, we work out how big each of the parts is based on the recommendedPartSize provided by our authorization response. Once we have that, it is just a case of generating the pre-signed URLs and returning these to the frontend with some extra meta we will find handy.

🚛 Complete Multipart Upload Endpoint #

Once the parts have been uploaded, we need to let the provider know, so they can piece the parts together. We will have a separate endpoint for this. Let’s create the file now at src/routes/api/complete-multipart-upload.json/+server.js, pasting in the content below:

    
import { authoriseAccount, completeMultipartUpload, getS3Client } from '$lib/utilities/storage';
export async function POST({ request, setHeaders }) {
  const { key, parts, uploadId } = await request.json();
  try {
    const { s3ApiUrl } = await authoriseAccount();
    if (s3ApiUrl) {
      const client = getS3Client({ s3ApiUrl });
      await completeMultipartUpload({ parts, client, key, uploadId });
      return new Response();
    }
    setHeaders({
      'Content-Type': 'application/json',
    });
    return new Response(JSON.stringify({ message: 'unauthorised' }));
  } catch (error) {
    const message = `Error in route api/complete-multipart-upload.json: ${error}`;
    console.error(message);
    throw new Error(message);
  }
}

  

That’s all the endpoint code in place now. Let's move on to the client page next.

🧑🏽 Client Homepage Svelte Code #

There’s not too much to change vs. the single file upload code. We'll start by adding a completeMultipartUpload function which calls that last endpoint we created. Add this block to src/routes/+page.svelte:

    
async function completeMultipartUpload({ key, parts, uploadId }) {
  try {
    await fetch('/api/complete-multipart-upload.json', {
      method: 'POST',
      credentials: 'omit',
      headers: {
        'Content-Type': 'application/json',
      },
      body: JSON.stringify({ key, parts, uploadId }),
    });
  } catch (error) {
    console.error(`Error in completeMultipartUpload on / route: ${error}`);
  }
}

  

Handle Submit #

Next we need to check in handleSubmit whether we have a single or multipart upload. If you are using this code in your own new project, you will probably want to refactor the block into separate functions, possibly in different files. Anyway, for now, paste in this block:

    
41   const handleSubmit = async () => {
42     try {
43       if (files.length === 0) {
44         errors.files = 'Select a file to upload first';
45         return;
46       }
47  
48       isSubmitting = true;
49       const { name: key, size, type } = files[0];
50  
51       // get signed upload URL
52       const response = await fetch('/api/presigned-urls.json', {
53         method: 'POST',
54         credentials: 'omit',
55         headers: {
56           'Content-Type': 'application/json',
57         },
58         body: JSON.stringify({ key, size }),
59       });
60       const json = await response.json();
61       const { multipartUploadUrls, partCount, partSize, readSignedUrl, writeSignedUrl, uploadId } =
62         json;
63       const reader = new FileReader();
64       if (partCount === 1) {
65         downloadUrl = readSignedUrl;
66  
67         // Upload (single part) file
68         reader.onloadend = async () => {
69           await fetch(writeSignedUrl, {
70             method: 'PUT',
71             body: reader.result,
72             headers: {
73               'Content-Type': type,
74             },
75           });
76           uploadComplete = true;
77           isSubmitting = false;
78         };
79         reader.readAsArrayBuffer(files[0]);
80       } else {
81         downloadUrl = readSignedUrl;
82         const lastIndex = multipartUploadUrls.length - 1;
83  
84         // Upload (multipartpart) file
85         reader.onloadend = async () => {
86           const uploadPromises = multipartUploadUrls.map((element, index) =>
87             fetch(element, {
88               method: 'PUT',
89               body:
90                 index !== lastIndex
91                   ? reader.result.slice(index * partSize, (index + 1) * partSize)
92                   : reader.result.slice(index * partSize),
93               headers: {
94                 'Content-Type': type,
95                 'Content-Length': index !== lastIndex ? partSize : size - index * partSize,
96               },
97             }),
98           );
99           const uploadResults = await Promise.all(uploadPromises);
100           const parts = uploadResults.map((element, index) => ({
101             ETag: element.headers.get('etag'),
102             PartNumber: index + 1,
103           }));
104           await completeMultipartUpload({ parts, key, uploadId });
105           uploadComplete = true;
106           isSubmitting = false;
107         };
108         reader.readAsArrayBuffer(files[0]);
109       }
110     } catch (error) {
111       console.error(`A;Error in handleSubmit on / route: ${error}`A;);
112     }
113   };
114 </script>

  

Notice in line 49 we now get the file size, so we can pass that to the pre-signed URL endpoint. The value we have is in bytes. For single part uploads, nothing really changes. So let's jump to the reader.onloadend block for multipart uploads starting at line 85.

We use JavaScript’s Promise API. That way, we do not need to wait for one part to finish uploading before we start on the next one. This allows for faster uploads. For larger files, where there will be dozens of parts, it would make sense to extend this code to throttle the downloads, so we only upload say three or four parts simultaneously and wait for one of those to finish before starting to upload a new part. We won't look at the detail of doing that here.

The code in lines 90 – 92 splits the file into chunks of the right size. We compute the part length and send it in the Content-Length header in line 95.

Multipart Upload Completion #

When we complete the multipart upload, to help piece together the parts, we send an ID to identify each part. That ID comes in the form of an ETag which is included in the multipart upload response header sent from our provider. We collate this data in lines 100 – 103 into the parts variable.

That parts object is passed to our completeMultipartUpload in this file and subsequently passed to the endpoint and the utility function.

Allowing Video Upload #

The final change is to update the user interface to accept video as well as image files:

    
165 <input
166   id="file"
167   aria-invalid={errors.files != null}
168   aria-describedby={errors.files != null ? 'files-error' : null}
169   type="file"
170   multiple
171   formenctype="multipart/form-data"
172   accept="image/*,video/*"
173   title="File"
174   on:change={handleChange}
175 />

  

Remember, you can change this to be more restrictive or, in fact, allow other types based on your own needs.

⛔️ CORS Update #

Because we want to look at a new header (the ETag header) from the client browser, we will need to update the bucket CORS policy. Check how to do this with your storage provider. If you are using Backblaze, you can update the backblaze-bucket-cors-rules.json file we introduced in the previous tutorial and submit this to Backblaze using the CLI.

    
1 [
2   {
3     "corsRuleName": "development",
4     "allowedOrigins": ["https://test.localhost.com:5173"],
5     "allowedHeaders": ["content-type", "range"],
6     "allowedOperations": ["s3_put"],
7     "exposeHeaders": ["etag", "x-amz-version-id"],
8     "maxAgeSeconds": 300
9   },
10   {
11     "corsRuleName": "production",
12     "allowedOrigins": ["https://example.com"],
13     "allowedHeaders": ["content-type", "range"],
14     "allowedOperations": ["s3_put"],
15     "exposeHeaders": ["etag", "x-amz-version-id"],
16     "maxAgeSeconds": 3600
17   }
18 ]

  

🙌🏽 SvelteKit S3 Multipart Upload: What we Learned #

In this post we looked at:

• how you can upload larger files to S3 compatible cloud storage,
• generating pre-signed URLs for multipart upload,
• how you can determine whether to use single or multipart upload and also calculate part size when choosing multipart upload.

I do hope there is at least one thing in this article which you can use in your work or a side project. As an extension, you might consider throttling uploads, especially when uploading very large files with many parts. You can also extend the UI to show existing uploads in the bucket and even generate download pre-signed links with custom parameters, like link validity. On top, consider adding code to abandon failed multipart uploads. This can potentially reduce costs.

You can see the full completed code for this tutorial on the Rodney Lab Git Hub repo  .

🏁 SvelteKit S3 Multipart Upload: Summary #

When should you use an S3 multipart upload? #

Multipart uploads make sense for larger files. Typically, this would be when uploading videos and files of similar sizes. Generally for images, small PDFs and so on, single part uploads are fine. As a rule of thumb, for files larger than 5 MB you will consider multipart uploads. Using multipart uploads allows your file to upload quicker. This is because the file can be split into parts, uploaded simultaneously. Consider a 16 MB file; instead of uploading 16 MB from start to finish, you can split the file into four chunks and start uploading all of them at the same time.

Can you upload large files to S3 with a pre-signed URL? #

Yes. Much like you can upload single files using pre-signed URLs, you can upload multipart files too. As an initial step, you need to work out how many parts you will use and get a pre-signed URL for each of these parts. You then upload the parts, looking out for the ETag header included in the server response when each part finishes uploading. You will use those ETags in a final call to let the provider know the upload is complete. The tags will help combine the parts into a single object (or file) in the cloud.

Is it possible to upload large files to S3 compatible storage from a SvelteKit app? #

Yes! We look at SvelteKit S3 multipart uploads in this tutorial. On top, we sort out determining when a multipart or single part upload makes most sense and how to upload directly from the client browser using pre-signed links for S3 compatible cloud storage providers.

🙏🏽 SvelteKit S3 Multipart Upload: Feedback #

Have you found the post useful? Would you prefer to see posts on another topic instead? Get in touch with ideas for new posts. Also, if you like my writing style, get in touch if I can write some posts for your company site on a consultancy basis. Read on to find ways to get in touch, further below. If you want to support posts similar to this one and can spare a few dollars, euros or pounds, please consider supporting me through Buy me a Coffee.

Rodney

@askRodney

🚚 how you can upload large files like video to S3 compatible storage with SvelteKit.

You can use this to let users upload content using your ❤️ Svelte app or to push your own media to the cloud.

Hope you find it useful!

https://t.co/kLFO52d8SG #askRodney #sveltekit @backblaze

— Rodney (@askRodney) November 19, 2021

Finally, feel free to share the post on your social media accounts for all your followers who will find it useful. As well as leaving a comment below, you can get in touch via @askRodney on Twitter and also askRodney on Telegram  . Also, see further ways to get in touch with Rodney Lab. I post regularly on SvelteKit as well as other topics. Also, subscribe to the newsletter to keep up-to-date with our latest projects.