1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing,
// software distributed under the License is distributed on an
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied. See the License for the
// specific language governing permissions and limitations
// under the License.
use std::sync::Arc;
use futures::select;
use futures::Future;
use futures::FutureExt;
use crate::raw::*;
use crate::*;
/// MultipartWrite is used to implement [`oio::Write`] based on multipart
/// uploads. By implementing MultipartWrite, services don't need to
/// care about the details of uploading parts.
///
/// # Architecture
///
/// The architecture after adopting [`MultipartWrite`]:
///
/// - Services impl `MultipartWrite`
/// - `MultipartWriter` impl `Write`
/// - Expose `MultipartWriter` as `Accessor::Writer`
///
/// # Notes
///
/// `MultipartWrite` has an oneshot optimization when `write` has been called only once:
///
/// ```no_build
/// w.write(bs).await?;
/// w.close().await?;
/// ```
///
/// We will use `write_once` instead of starting a new multipart upload.
///
/// # Requirements
///
/// Services that implement `BlockWrite` must fulfill the following requirements:
///
/// - Must be a http service that could accept `AsyncBody`.
/// - Don't need initialization before writing.
/// - Block ID is generated by caller `BlockWrite` instead of services.
/// - Complete block by an ordered block id list.
pub trait MultipartWrite: Send + Sync + Unpin + 'static {
/// write_once is used to write the data to underlying storage at once.
///
/// MultipartWriter will call this API when:
///
/// - All the data has been written to the buffer and we can perform the upload at once.
fn write_once(&self, size: u64, body: Buffer) -> impl Future<Output = Result<()>> + MaybeSend;
/// initiate_part will call start a multipart upload and return the upload id.
///
/// MultipartWriter will call this when:
///
/// - the total size of data is unknown.
/// - the total size of data is known, but the size of current write
/// is less then the total size.
fn initiate_part(&self) -> impl Future<Output = Result<String>> + MaybeSend;
/// write_part will write a part of the data and returns the result
/// [`MultipartPart`].
///
/// MultipartWriter will call this API and stores the result in
/// order.
///
/// - part_number is the index of the part, starting from 0.
fn write_part(
&self,
upload_id: &str,
part_number: usize,
size: u64,
body: Buffer,
) -> impl Future<Output = Result<MultipartPart>> + MaybeSend;
/// complete_part will complete the multipart upload to build the final
/// file.
fn complete_part(
&self,
upload_id: &str,
parts: &[MultipartPart],
) -> impl Future<Output = Result<()>> + MaybeSend;
/// abort_part will cancel the multipart upload and purge all data.
fn abort_part(&self, upload_id: &str) -> impl Future<Output = Result<()>> + MaybeSend;
}
/// The result of [`MultipartWrite::write_part`].
///
/// services implement should convert MultipartPart to their own represents.
///
/// - `part_number` is the index of the part, starting from 0.
/// - `etag` is the `ETag` of the part.
/// - `checksum` is the optional checksum of the part.
#[derive(Clone)]
pub struct MultipartPart {
/// The number of the part, starting from 0.
pub part_number: usize,
/// The etag of the part.
pub etag: String,
/// The checksum of the part.
pub checksum: Option<String>,
}
struct WriteInput<W: MultipartWrite> {
w: Arc<W>,
executor: Executor,
upload_id: Arc<String>,
part_number: usize,
bytes: Buffer,
}
/// MultipartWriter will implements [`oio::Write`] based on multipart
/// uploads.
pub struct MultipartWriter<W: MultipartWrite> {
w: Arc<W>,
executor: Executor,
upload_id: Option<Arc<String>>,
parts: Vec<MultipartPart>,
cache: Option<Buffer>,
next_part_number: usize,
tasks: ConcurrentTasks<WriteInput<W>, MultipartPart>,
}
/// # Safety
///
/// wasm32 is a special target that we only have one event-loop for this state.
impl<W: MultipartWrite> MultipartWriter<W> {
/// Create a new MultipartWriter.
pub fn new(inner: W, executor: Option<Executor>, concurrent: usize) -> Self {
let w = Arc::new(inner);
let executor = executor.unwrap_or_default();
Self {
w,
executor: executor.clone(),
upload_id: None,
parts: Vec::new(),
cache: None,
next_part_number: 0,
tasks: ConcurrentTasks::new(executor, concurrent, |input| {
Box::pin({
async move {
let fut = input.w.write_part(
&input.upload_id,
input.part_number,
input.bytes.len() as u64,
input.bytes.clone(),
);
match input.executor.timeout() {
None => {
let result = fut.await;
(input, result)
}
Some(timeout) => {
let result = select! {
result = fut.fuse() => {
result
}
_ = timeout.fuse() => {
Err(Error::new(
ErrorKind::Unexpected, "write part timeout")
.with_context("upload_id", input.upload_id.to_string())
.with_context("part_number", input.part_number.to_string())
.set_temporary())
}
};
(input, result)
}
}
}
})
}),
}
}
fn fill_cache(&mut self, bs: Buffer) -> usize {
let size = bs.len();
assert!(self.cache.is_none());
self.cache = Some(bs);
size
}
}
impl<W> oio::Write for MultipartWriter<W>
where
W: MultipartWrite,
{
async fn write(&mut self, bs: Buffer) -> Result<()> {
let upload_id = match self.upload_id.clone() {
Some(v) => v,
None => {
// Fill cache with the first write.
if self.cache.is_none() {
self.fill_cache(bs);
return Ok(());
}
let upload_id = self.w.initiate_part().await?;
let upload_id = Arc::new(upload_id);
self.upload_id = Some(upload_id.clone());
upload_id
}
};
let bytes = self.cache.clone().expect("pending write must exist");
let part_number = self.next_part_number;
self.tasks
.execute(WriteInput {
w: self.w.clone(),
executor: self.executor.clone(),
upload_id: upload_id.clone(),
part_number,
bytes,
})
.await?;
self.cache = None;
self.next_part_number += 1;
self.fill_cache(bs);
Ok(())
}
async fn close(&mut self) -> Result<()> {
let upload_id = match self.upload_id.clone() {
Some(v) => v,
None => {
let (size, body) = match self.cache.clone() {
Some(cache) => (cache.len(), cache),
None => (0, Buffer::new()),
};
// Call write_once if there is no upload_id.
self.w.write_once(size as u64, body).await?;
self.cache = None;
return Ok(());
}
};
if let Some(cache) = self.cache.clone() {
let part_number = self.next_part_number;
self.tasks
.execute(WriteInput {
w: self.w.clone(),
executor: self.executor.clone(),
upload_id: upload_id.clone(),
part_number,
bytes: cache,
})
.await?;
self.cache = None;
self.next_part_number += 1;
}
loop {
let Some(result) = self.tasks.next().await.transpose()? else {
break;
};
self.parts.push(result)
}
if self.parts.len() != self.next_part_number {
return Err(Error::new(
ErrorKind::Unexpected,
"multipart part numbers mismatch, please report bug to opendal",
)
.with_context("expected", self.next_part_number)
.with_context("actual", self.parts.len())
.with_context("upload_id", upload_id));
}
self.w.complete_part(&upload_id, &self.parts).await
}
async fn abort(&mut self) -> Result<()> {
let Some(upload_id) = self.upload_id.clone() else {
return Ok(());
};
self.tasks.clear();
self.cache = None;
self.w.abort_part(&upload_id).await?;
Ok(())
}
}
#[cfg(test)]
mod tests {
use std::time::Duration;
use pretty_assertions::assert_eq;
use rand::thread_rng;
use rand::Rng;
use rand::RngCore;
use tokio::sync::Mutex;
use tokio::time::sleep;
use tokio::time::timeout;
use super::*;
use crate::raw::oio::Write;
struct TestWrite {
upload_id: String,
part_numbers: Vec<usize>,
length: u64,
}
impl TestWrite {
pub fn new() -> Arc<Mutex<Self>> {
let v = Self {
upload_id: uuid::Uuid::new_v4().to_string(),
part_numbers: Vec::new(),
length: 0,
};
Arc::new(Mutex::new(v))
}
}
impl MultipartWrite for Arc<Mutex<TestWrite>> {
async fn write_once(&self, size: u64, _: Buffer) -> Result<()> {
self.lock().await.length += size;
Ok(())
}
async fn initiate_part(&self) -> Result<String> {
let upload_id = self.lock().await.upload_id.clone();
Ok(upload_id)
}
async fn write_part(
&self,
upload_id: &str,
part_number: usize,
size: u64,
_: Buffer,
) -> Result<MultipartPart> {
{
let test = self.lock().await;
assert_eq!(upload_id, test.upload_id);
}
// Add an async sleep here to enforce some pending.
sleep(Duration::from_nanos(50)).await;
// We will have 10% percent rate for write part to fail.
if thread_rng().gen_bool(1.0 / 10.0) {
return Err(
Error::new(ErrorKind::Unexpected, "I'm a crazy monkey!").set_temporary()
);
}
{
let mut test = self.lock().await;
test.part_numbers.push(part_number);
test.length += size;
}
Ok(MultipartPart {
part_number,
etag: "etag".to_string(),
checksum: None,
})
}
async fn complete_part(&self, upload_id: &str, parts: &[MultipartPart]) -> Result<()> {
let test = self.lock().await;
assert_eq!(upload_id, test.upload_id);
assert_eq!(parts.len(), test.part_numbers.len());
Ok(())
}
async fn abort_part(&self, upload_id: &str) -> Result<()> {
let test = self.lock().await;
assert_eq!(upload_id, test.upload_id);
Ok(())
}
}
struct TimeoutExecutor {
exec: Arc<dyn Execute>,
}
impl TimeoutExecutor {
pub fn new() -> Self {
Self {
exec: Executor::new().into_inner(),
}
}
}
impl Execute for TimeoutExecutor {
fn execute(&self, f: BoxedStaticFuture<()>) {
self.exec.execute(f)
}
fn timeout(&self) -> Option<BoxedStaticFuture<()>> {
let time = thread_rng().gen_range(0..100);
Some(Box::pin(tokio::time::sleep(Duration::from_nanos(time))))
}
}
#[tokio::test]
async fn test_multipart_upload_writer_with_concurrent_errors() {
let mut rng = thread_rng();
let mut w = MultipartWriter::new(
TestWrite::new(),
Some(Executor::with(TimeoutExecutor::new())),
200,
);
let mut total_size = 0u64;
for _ in 0..1000 {
let size = rng.gen_range(1..1024);
total_size += size as u64;
let mut bs = vec![0; size];
rng.fill_bytes(&mut bs);
loop {
match timeout(Duration::from_nanos(10), w.write(bs.clone().into())).await {
Ok(Ok(_)) => break,
Ok(Err(_)) => continue,
Err(_) => {
continue;
}
}
}
}
loop {
match timeout(Duration::from_nanos(10), w.close()).await {
Ok(Ok(_)) => break,
Ok(Err(_)) => continue,
Err(_) => {
continue;
}
}
}
let actual_parts: Vec<_> = w.parts.into_iter().map(|v| v.part_number).collect();
let expected_parts: Vec<_> = (0..1000).collect();
assert_eq!(actual_parts, expected_parts);
let actual_size = w.w.lock().await.length;
assert_eq!(actual_size, total_size);
}
}