feat: multipart blob sync support (#14138)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* Flexible blob uploads: GRAPHQL, presigned, and multipart flows with
per‑part URLs, abort/complete operations, presigned proxy endpoints, and
nightly cleanup of expired pending uploads.

* **API / Schema**
* GraphQL additions: new types, mutations, enum and error to manage
upload lifecycle (create, complete, abort, get part URL).

* **Database**
* New blob status enum and columns (status, upload_id); listing now
defaults to completed blobs.

* **Localization**
  * Added user-facing message: "Blob is invalid."

* **Tests**
* Expanded unit and end‑to‑end coverage for upload flows, proxy
behavior, multipart and provider integrations.

<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
This commit is contained in:
DarkSky
2025-12-23 22:09:21 +08:00
committed by GitHub
parent a9937e18b6
commit 76524084d1
36 changed files with 2880 additions and 33 deletions
@@ -506,6 +506,10 @@ export const USER_FRIENDLY_ERRORS = {
message: ({ spaceId, blobId }) =>
`Blob ${blobId} not found in Space ${spaceId}.`,
},
blob_invalid: {
type: 'invalid_input',
message: 'Blob is invalid.',
},
expect_to_publish_doc: {
type: 'invalid_input',
message: 'Expected to publish a doc, not a Space.',
@@ -437,6 +437,12 @@ export class BlobNotFound extends UserFriendlyError {
}
}
export class BlobInvalid extends UserFriendlyError {
constructor(message?: string) {
super('invalid_input', 'blob_invalid', message);
}
}
export class ExpectToPublishDoc extends UserFriendlyError {
constructor(message?: string) {
super('invalid_input', 'expect_to_publish_doc', message);
@@ -1166,6 +1172,7 @@ export enum ErrorNames {
INVALID_HISTORY_TIMESTAMP,
DOC_HISTORY_NOT_FOUND,
BLOB_NOT_FOUND,
BLOB_INVALID,
EXPECT_TO_PUBLISH_DOC,
EXPECT_TO_REVOKE_PUBLIC_DOC,
EXPECT_TO_GRANT_DOC_USER_ROLES,
@@ -19,8 +19,12 @@ class Redis extends IORedis implements OnModuleInit, OnModuleDestroy {
this.on('error', this.errorHandler);
}
onModuleDestroy() {
this.disconnect();
async onModuleDestroy() {
try {
await this.quit();
} catch {
this.disconnect();
}
}
override duplicate(override?: Partial<RedisOptions>): IORedis {
@@ -0,0 +1,80 @@
import test from 'ava';
import { S3StorageProvider } from '../providers/s3';
import { SIGNED_URL_EXPIRED } from '../providers/utils';
const config = {
region: 'auto',
credentials: {
accessKeyId: 'test',
secretAccessKey: 'test',
},
};
function createProvider() {
return new S3StorageProvider(config, 'test-bucket');
}
test('presignPut should return url and headers', async t => {
const provider = createProvider();
const result = await provider.presignPut('key', {
contentType: 'text/plain',
});
t.truthy(result);
t.true(result!.url.length > 0);
t.true(result!.url.includes('X-Amz-Algorithm=AWS4-HMAC-SHA256'));
t.deepEqual(result!.headers, { 'Content-Type': 'text/plain' });
const now = Date.now();
t.true(result!.expiresAt.getTime() >= now + SIGNED_URL_EXPIRED * 1000 - 2000);
t.true(result!.expiresAt.getTime() <= now + SIGNED_URL_EXPIRED * 1000 + 2000);
});
test('presignUploadPart should return url', async t => {
const provider = createProvider();
const result = await provider.presignUploadPart('key', 'upload-1', 3);
t.truthy(result);
t.true(result!.url.length > 0);
t.true(result!.url.includes('X-Amz-Algorithm=AWS4-HMAC-SHA256'));
});
test('createMultipartUpload should return uploadId', async t => {
const provider = createProvider();
let receivedCommand: any;
const sendStub = async (command: any) => {
receivedCommand = command;
return { UploadId: 'upload-1' };
};
(provider as any).client = { send: sendStub };
const now = Date.now();
const result = await provider.createMultipartUpload('key', {
contentType: 'text/plain',
});
t.is(result?.uploadId, 'upload-1');
t.true(result!.expiresAt.getTime() >= now + SIGNED_URL_EXPIRED * 1000 - 2000);
t.true(result!.expiresAt.getTime() <= now + SIGNED_URL_EXPIRED * 1000 + 2000);
t.is(receivedCommand.input.Key, 'key');
t.is(receivedCommand.input.ContentType, 'text/plain');
});
test('completeMultipartUpload should order parts', async t => {
const provider = createProvider();
let called = false;
const sendStub = async (command: any) => {
called = true;
t.deepEqual(command.input.MultipartUpload.Parts, [
{ ETag: 'a', PartNumber: 1 },
{ ETag: 'b', PartNumber: 2 },
]);
};
(provider as any).client = { send: sendStub };
await provider.completeMultipartUpload('key', 'upload-1', [
{ partNumber: 2, etag: 'b' },
{ partNumber: 1, etag: 'a' },
]);
t.true(called);
});
@@ -118,7 +118,7 @@ export const StorageJSONSchema: JSONSchema = {
urlPrefix: {
type: 'string' as const,
description:
'The presigned url prefix for the cloudflare r2 storage provider.\nsee https://developers.cloudflare.com/waf/custom-rules/use-cases/configure-token-authentication/ to configure it.\nExample value: "https://storage.example.com"\nExample rule: is_timed_hmac_valid_v0("your_secret", http.request.uri, 10800, http.request.timestamp.sec, 6)',
'The custom domain URL prefix for the cloudflare r2 storage provider.\nWhen `enabled=true` and `urlPrefix` + `signKey` are provided, the server will:\n- Redirect GET requests to this custom domain with an HMAC token.\n- Return upload URLs under `/api/storage/*` for uploads.\nPresigned/upload proxy TTL is 1 hour.\nsee https://developers.cloudflare.com/waf/custom-rules/use-cases/configure-token-authentication/ to configure it.\nExample value: "https://storage.example.com"\nExample rule: is_timed_hmac_valid_v0("your_secret", http.request.uri, 10800, http.request.timestamp.sec, 6)',
},
signKey: {
type: 'string' as const,
@@ -135,4 +135,12 @@ export const StorageJSONSchema: JSONSchema = {
};
export type * from './provider';
export { applyAttachHeaders, autoMetadata, sniffMime, toBuffer } from './utils';
export {
applyAttachHeaders,
autoMetadata,
PROXY_MULTIPART_PATH,
PROXY_UPLOAD_PATH,
sniffMime,
STORAGE_PROXY_ROOT,
toBuffer,
} from './utils';
@@ -25,12 +25,51 @@ export interface ListObjectsMetadata {
export type BlobInputType = Buffer | Readable | string;
export type BlobOutputType = Readable;
export interface PresignedUpload {
url: string;
headers?: Record<string, string>;
expiresAt: Date;
}
export interface MultipartUploadInit {
uploadId: string;
expiresAt: Date;
}
export interface MultipartUploadPart {
partNumber: number;
etag: string;
}
export interface StorageProvider {
put(
key: string,
body: BlobInputType,
metadata?: PutObjectMetadata
): Promise<void>;
presignPut?(
key: string,
metadata?: PutObjectMetadata
): Promise<PresignedUpload | undefined>;
createMultipartUpload?(
key: string,
metadata?: PutObjectMetadata
): Promise<MultipartUploadInit | undefined>;
presignUploadPart?(
key: string,
uploadId: string,
partNumber: number
): Promise<PresignedUpload | undefined>;
listMultipartUploadParts?(
key: string,
uploadId: string
): Promise<MultipartUploadPart[] | undefined>;
completeMultipartUpload?(
key: string,
uploadId: string,
parts: MultipartUploadPart[]
): Promise<void>;
abortMultipartUpload?(key: string, uploadId: string): Promise<void>;
head(key: string): Promise<GetObjectMetadata | undefined>;
get(
key: string,
@@ -1,10 +1,20 @@
import assert from 'node:assert';
import { Readable } from 'node:stream';
import { PutObjectCommand, UploadPartCommand } from '@aws-sdk/client-s3';
import { Logger } from '@nestjs/common';
import { GetObjectMetadata } from './provider';
import {
GetObjectMetadata,
PresignedUpload,
PutObjectMetadata,
} from './provider';
import { S3StorageConfig, S3StorageProvider } from './s3';
import {
PROXY_MULTIPART_PATH,
PROXY_UPLOAD_PATH,
SIGNED_URL_EXPIRED,
} from './utils';
export interface R2StorageConfig extends S3StorageConfig {
accountId: string;
@@ -39,8 +49,24 @@ export class R2StorageProvider extends S3StorageProvider {
this.key = this.encoder.encode(config.usePresignedURL?.signKey ?? '');
}
private async signUrl(url: URL): Promise<string> {
const timestamp = Math.floor(Date.now() / 1000);
private get shouldUseProxyUpload() {
const { usePresignedURL } = this.config;
return (
!!usePresignedURL?.enabled &&
!!usePresignedURL.signKey &&
this.key.length > 0
);
}
private parseWorkspaceKey(fullKey: string) {
const [workspaceId, ...rest] = fullKey.split('/');
if (!workspaceId || rest.length !== 1) {
return null;
}
return { workspaceId, key: rest.join('/') };
}
private async signPayload(payload: string) {
const key = await crypto.subtle.importKey(
'raw',
this.key,
@@ -51,14 +77,140 @@ export class R2StorageProvider extends S3StorageProvider {
const mac = await crypto.subtle.sign(
'HMAC',
key,
this.encoder.encode(`${url.pathname}${timestamp}`)
this.encoder.encode(payload)
);
const base64Mac = Buffer.from(mac).toString('base64');
return Buffer.from(mac).toString('base64');
}
private async signUrl(url: URL): Promise<string> {
const timestamp = Math.floor(Date.now() / 1000);
const base64Mac = await this.signPayload(`${url.pathname}${timestamp}`);
url.searchParams.set('sign', `${timestamp}-${base64Mac}`);
return url.toString();
}
private async createProxyUrl(
path: string,
canonicalFields: (string | number | undefined)[],
query: Record<string, string | number | undefined>
) {
const exp = Math.floor(Date.now() / 1000) + SIGNED_URL_EXPIRED;
const canonical = [
path,
...canonicalFields.map(field =>
field === undefined ? '' : field.toString()
),
exp.toString(),
].join('\n');
const token = await this.signPayload(canonical);
const url = new URL(`http://localhost${path}`);
for (const [key, value] of Object.entries(query)) {
if (value === undefined) continue;
url.searchParams.set(key, value.toString());
}
url.searchParams.set('exp', exp.toString());
url.searchParams.set('token', `${exp}-${token}`);
return { url: url.pathname + url.search, expiresAt: new Date(exp * 1000) };
}
override async presignPut(
key: string,
metadata: PutObjectMetadata = {}
): Promise<PresignedUpload | undefined> {
if (!this.shouldUseProxyUpload) {
return super.presignPut(key, metadata);
}
const parsed = this.parseWorkspaceKey(key);
if (!parsed) {
return super.presignPut(key, metadata);
}
const contentType = metadata.contentType ?? 'application/octet-stream';
const { url, expiresAt } = await this.createProxyUrl(
PROXY_UPLOAD_PATH,
[parsed.workspaceId, parsed.key, contentType, metadata.contentLength],
{
workspaceId: parsed.workspaceId,
key: parsed.key,
contentType,
contentLength: metadata.contentLength,
}
);
return {
url,
headers: { 'Content-Type': contentType },
expiresAt,
};
}
override async presignUploadPart(
key: string,
uploadId: string,
partNumber: number
): Promise<PresignedUpload | undefined> {
if (!this.shouldUseProxyUpload) {
return super.presignUploadPart(key, uploadId, partNumber);
}
const parsed = this.parseWorkspaceKey(key);
if (!parsed) {
return super.presignUploadPart(key, uploadId, partNumber);
}
return this.createProxyUrl(
PROXY_MULTIPART_PATH,
[parsed.workspaceId, parsed.key, uploadId, partNumber],
{
workspaceId: parsed.workspaceId,
key: parsed.key,
uploadId,
partNumber,
}
);
}
async proxyPutObject(
key: string,
body: Readable | Buffer | Uint8Array | string,
options: { contentType?: string; contentLength?: number } = {}
) {
return this.client.send(
new PutObjectCommand({
Bucket: this.bucket,
Key: key,
Body: body,
ContentType: options.contentType,
ContentLength: options.contentLength,
})
);
}
async proxyUploadPart(
key: string,
uploadId: string,
partNumber: number,
body: Readable | Buffer | Uint8Array | string,
options: { contentLength?: number } = {}
) {
const result = await this.client.send(
new UploadPartCommand({
Bucket: this.bucket,
Key: key,
UploadId: uploadId,
PartNumber: partNumber,
Body: body,
ContentLength: options.contentLength,
})
);
return result.ETag;
}
override async get(
key: string,
signedUrl?: boolean
@@ -2,15 +2,21 @@
import { Readable } from 'node:stream';
import {
AbortMultipartUploadCommand,
CompleteMultipartUploadCommand,
CreateMultipartUploadCommand,
DeleteObjectCommand,
GetObjectCommand,
HeadObjectCommand,
ListObjectsV2Command,
ListPartsCommand,
NoSuchKey,
NoSuchUpload,
NotFound,
PutObjectCommand,
S3Client,
S3ClientConfig,
UploadPartCommand,
} from '@aws-sdk/client-s3';
import { getSignedUrl } from '@aws-sdk/s3-request-presigner';
import { Logger } from '@nestjs/common';
@@ -19,6 +25,9 @@ import {
BlobInputType,
GetObjectMetadata,
ListObjectsMetadata,
MultipartUploadInit,
MultipartUploadPart,
PresignedUpload,
PutObjectMetadata,
StorageProvider,
} from './provider';
@@ -89,6 +98,196 @@ export class S3StorageProvider implements StorageProvider {
}
}
async presignPut(
key: string,
metadata: PutObjectMetadata = {}
): Promise<PresignedUpload | undefined> {
try {
const contentType = metadata.contentType ?? 'application/octet-stream';
const url = await getSignedUrl(
this.client,
new PutObjectCommand({
Bucket: this.bucket,
Key: key,
ContentType: contentType,
}),
{ expiresIn: SIGNED_URL_EXPIRED }
);
return {
url,
headers: { 'Content-Type': contentType },
expiresAt: new Date(Date.now() + SIGNED_URL_EXPIRED * 1000),
};
} catch (e) {
this.logger.error(
`Failed to presign put object (${JSON.stringify({
key,
bucket: this.bucket,
metadata,
})}`
);
throw e;
}
}
async createMultipartUpload(
key: string,
metadata: PutObjectMetadata = {}
): Promise<MultipartUploadInit | undefined> {
try {
const contentType = metadata.contentType ?? 'application/octet-stream';
const response = await this.client.send(
new CreateMultipartUploadCommand({
Bucket: this.bucket,
Key: key,
ContentType: contentType,
})
);
if (!response.UploadId) {
return;
}
return {
uploadId: response.UploadId,
expiresAt: new Date(Date.now() + SIGNED_URL_EXPIRED * 1000),
};
} catch (e) {
this.logger.error(
`Failed to create multipart upload (${JSON.stringify({
key,
bucket: this.bucket,
metadata,
})}`
);
throw e;
}
}
async presignUploadPart(
key: string,
uploadId: string,
partNumber: number
): Promise<PresignedUpload | undefined> {
try {
const url = await getSignedUrl(
this.client,
new UploadPartCommand({
Bucket: this.bucket,
Key: key,
UploadId: uploadId,
PartNumber: partNumber,
}),
{ expiresIn: SIGNED_URL_EXPIRED }
);
return {
url,
expiresAt: new Date(Date.now() + SIGNED_URL_EXPIRED * 1000),
};
} catch (e) {
this.logger.error(
`Failed to presign upload part (${JSON.stringify({ key, bucket: this.bucket, uploadId, partNumber })}`
);
throw e;
}
}
async listMultipartUploadParts(
key: string,
uploadId: string
): Promise<MultipartUploadPart[] | undefined> {
const parts: MultipartUploadPart[] = [];
let partNumberMarker: string | undefined;
try {
// ListParts is paginated by part number marker
// https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListParts.html
// R2 follows S3 semantics here.
while (true) {
const response = await this.client.send(
new ListPartsCommand({
Bucket: this.bucket,
Key: key,
UploadId: uploadId,
PartNumberMarker: partNumberMarker,
})
);
for (const part of response.Parts ?? []) {
if (!part.PartNumber || !part.ETag) {
continue;
}
parts.push({ partNumber: part.PartNumber, etag: part.ETag });
}
if (!response.IsTruncated) {
break;
}
if (response.NextPartNumberMarker === undefined) {
break;
}
partNumberMarker = response.NextPartNumberMarker;
}
return parts;
} catch (e) {
// the upload may have been aborted/expired by provider lifecycle rules
if (e instanceof NoSuchUpload || e instanceof NotFound) {
return undefined;
}
this.logger.error(`Failed to list multipart upload parts for \`${key}\``);
throw e;
}
}
async completeMultipartUpload(
key: string,
uploadId: string,
parts: MultipartUploadPart[]
): Promise<void> {
try {
const orderedParts = [...parts].sort(
(left, right) => left.partNumber - right.partNumber
);
await this.client.send(
new CompleteMultipartUploadCommand({
Bucket: this.bucket,
Key: key,
UploadId: uploadId,
MultipartUpload: {
Parts: orderedParts.map(part => ({
ETag: part.etag,
PartNumber: part.partNumber,
})),
},
})
);
} catch (e) {
this.logger.error(`Failed to complete multipart upload for \`${key}\``);
throw e;
}
}
async abortMultipartUpload(key: string, uploadId: string): Promise<void> {
try {
await this.client.send(
new AbortMultipartUploadCommand({
Bucket: this.bucket,
Key: key,
UploadId: uploadId,
})
);
} catch (e) {
this.logger.error(`Failed to abort multipart upload for \`${key}\``);
throw e;
}
}
async head(key: string) {
try {
const obj = await this.client.send(
@@ -164,7 +363,7 @@ export class S3StorageProvider implements StorageProvider {
body: obj.Body,
metadata: {
// always set when putting object
contentType: obj.ContentType!,
contentType: obj.ContentType ?? 'application/octet-stream',
contentLength: obj.ContentLength!,
lastModified: obj.LastModified!,
checksumCRC32: obj.ChecksumCRC32,
@@ -94,3 +94,7 @@ export function sniffMime(
}
export const SIGNED_URL_EXPIRED = 60 * 60; // 1 hour
export const STORAGE_PROXY_ROOT = '/api/storage';
export const PROXY_UPLOAD_PATH = `${STORAGE_PROXY_ROOT}/upload`;
export const PROXY_MULTIPART_PATH = `${STORAGE_PROXY_ROOT}/multipart`;