opendal_layer_timeout/
lib.rs

1// Licensed to the Apache Software Foundation (ASF) under one
2// or more contributor license agreements.  See the NOTICE file
3// distributed with this work for additional information
4// regarding copyright ownership.  The ASF licenses this file
5// to you under the Apache License, Version 2.0 (the
6// "License"); you may not use this file except in compliance
7// with the License.  You may obtain a copy of the License at
8//
9//   http://www.apache.org/licenses/LICENSE-2.0
10//
11// Unless required by applicable law or agreed to in writing,
12// software distributed under the License is distributed on an
13// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14// KIND, either express or implied.  See the License for the
15// specific language governing permissions and limitations
16// under the License.
17
18//! Timeout layer implementation for Apache OpenDAL.
19
20#![cfg_attr(docsrs, feature(doc_cfg))]
21#![deny(missing_docs)]
22
23use std::future::Future;
24use std::sync::Arc;
25use std::time::Duration;
26
27use opendal_core::raw::*;
28use opendal_core::*;
29
30/// Add timeout for every operation to avoid slow or unexpected hang operations.
31///
32/// For example, a dead connection could hang a databases sql query. TimeoutLayer
33/// will break this connection and returns an error so users can handle it by
34/// retrying or print to users.
35///
36/// # Notes
37///
38/// `TimeoutLayer` treats all operations in two kinds:
39///
40/// - Non IO Operation like `stat`, `delete` they operate on a single file. We control
41///   them by setting `timeout`.
42/// - IO Operation like `read`, `Reader::read` and `Writer::write`, they operate on data directly, we
43///   control them by setting `io_timeout`.
44///
45/// # Default
46///
47/// - timeout: 60 seconds
48/// - io_timeout: 10 seconds
49///
50/// # Panics
51///
52/// TimeoutLayer will drop the future if the timeout is reached. This might cause the internal state
53/// of the future to be broken. If underlying future moves ownership into the future, it will be
54/// dropped and will neven return back.
55///
56/// For example, while using `TimeoutLayer` with `RetryLayer` at the same time, please make sure
57/// timeout layer showed up before retry layer.
58///
59/// ```no_run
60/// # use std::time::Duration;
61/// #
62/// # use opendal_core::services;
63/// # use opendal_core::Operator;
64/// # use opendal_core::Result;
65/// # use opendal_layer_retry::RetryLayer;
66/// # use opendal_layer_timeout::TimeoutLayer;
67/// #
68/// # fn main() -> Result<()> {
69/// let op = Operator::new(services::Memory::default())?
70///     // This is fine, since timeout happen during retry.
71///     .layer(TimeoutLayer::default().with_io_timeout(Duration::from_nanos(1)))
72///     .layer(RetryLayer::default())
73///     // This is wrong. Since timeout layer will drop future, leaving retry layer in a bad state.
74///     .layer(TimeoutLayer::default().with_io_timeout(Duration::from_nanos(1)))
75///     .finish();
76/// # Ok(())
77/// # }
78/// ```
79///
80/// # Examples
81///
82/// The following examples will create a timeout layer with 10 seconds timeout for all non-io
83/// operations, 3 seconds timeout for all io operations.
84///
85/// ```no_run
86/// # use std::time::Duration;
87/// #
88/// # use opendal_core::services;
89/// # use opendal_core::Operator;
90/// # use opendal_core::Result;
91/// # use opendal_layer_timeout::TimeoutLayer;
92/// #
93/// # fn main() -> Result<()> {
94/// let _ = Operator::new(services::Memory::default())?
95///     .layer(
96///         TimeoutLayer::default()
97///             .with_timeout(Duration::from_secs(10))
98///             .with_io_timeout(Duration::from_secs(3)),
99///     )
100///     .finish();
101/// # Ok(())
102/// # }
103/// ```
104///
105/// # Implementation Notes
106///
107/// TimeoutLayer is using [`tokio::time::timeout`] to implement timeout for operations. And IO
108/// Operations insides `reader`, `writer` will use `Pin<Box<tokio::time::Sleep>>` to track the
109/// timeout.
110///
111/// This might introduce a bit overhead for IO operations, but it's the only way to implement
112/// timeout correctly. We used to implement timeout layer in zero cost way that only stores
113/// a [`Instant`] and check the timeout by comparing the instant with current time.
114/// However, it doesn't work for all cases.
115///
116/// For examples, users TCP connection could be in [Busy ESTAB](https://blog.cloudflare.com/when-tcp-sockets-refuse-to-die) state. In this state, no IO event will be emitted. The runtime
117/// will never poll our future again. From the application side, this future is hanging forever
118/// until this TCP connection is closed for reaching the linux [net.ipv4.tcp_retries2](https://man7.org/linux/man-pages/man7/tcp.7.html) times.
119#[derive(Clone)]
120pub struct TimeoutLayer {
121    timeout: Duration,
122    io_timeout: Duration,
123}
124
125impl Default for TimeoutLayer {
126    fn default() -> Self {
127        Self {
128            timeout: Duration::from_secs(60),
129            io_timeout: Duration::from_secs(10),
130        }
131    }
132}
133
134impl TimeoutLayer {
135    /// Create a new [`TimeoutLayer`] with default settings.
136    pub fn new() -> Self {
137        Self::default()
138    }
139
140    /// Set timeout for TimeoutLayer with given value.
141    ///
142    /// This timeout is for all non-io operations like `stat`, `delete`.
143    pub fn with_timeout(mut self, timeout: Duration) -> Self {
144        self.timeout = timeout;
145        self
146    }
147
148    /// Set io timeout for TimeoutLayer with given value.
149    ///
150    /// This timeout is for all io operations like `read`, `Reader::read` and `Writer::write`.
151    pub fn with_io_timeout(mut self, timeout: Duration) -> Self {
152        self.io_timeout = timeout;
153        self
154    }
155}
156
157impl<A: Access> Layer<A> for TimeoutLayer {
158    type LayeredAccess = TimeoutAccessor<A>;
159
160    fn layer(&self, inner: A) -> Self::LayeredAccess {
161        let info = inner.info();
162        info.update_executor(|exec| {
163            Executor::with(TimeoutExecutor::new(exec.into_inner(), self.io_timeout))
164        });
165
166        TimeoutAccessor {
167            inner,
168
169            timeout: self.timeout,
170            io_timeout: self.io_timeout,
171        }
172    }
173}
174
175#[doc(hidden)]
176#[derive(Debug)]
177pub struct TimeoutAccessor<A: Access> {
178    inner: A,
179
180    timeout: Duration,
181    io_timeout: Duration,
182}
183
184impl<A: Access> TimeoutAccessor<A> {
185    async fn timeout<F: Future<Output = Result<T>>, T>(&self, op: Operation, fut: F) -> Result<T> {
186        tokio::time::timeout(self.timeout, fut).await.map_err(|_| {
187            Error::new(ErrorKind::Unexpected, "operation timeout reached")
188                .with_operation(op)
189                .with_context("timeout", self.timeout.as_secs_f64().to_string())
190                .set_temporary()
191        })?
192    }
193
194    async fn io_timeout<F: Future<Output = Result<T>>, T>(
195        &self,
196        op: Operation,
197        fut: F,
198    ) -> Result<T> {
199        tokio::time::timeout(self.io_timeout, fut)
200            .await
201            .map_err(|_| {
202                Error::new(ErrorKind::Unexpected, "io timeout reached")
203                    .with_operation(op)
204                    .with_context("timeout", self.io_timeout.as_secs_f64().to_string())
205                    .set_temporary()
206            })?
207    }
208}
209
210impl<A: Access> LayeredAccess for TimeoutAccessor<A> {
211    type Inner = A;
212    type Reader = TimeoutWrapper<A::Reader>;
213    type Writer = TimeoutWrapper<A::Writer>;
214    type Lister = TimeoutWrapper<A::Lister>;
215    type Deleter = TimeoutWrapper<A::Deleter>;
216
217    fn inner(&self) -> &Self::Inner {
218        &self.inner
219    }
220
221    async fn create_dir(&self, path: &str, args: OpCreateDir) -> Result<RpCreateDir> {
222        self.timeout(Operation::CreateDir, self.inner.create_dir(path, args))
223            .await
224    }
225
226    async fn read(&self, path: &str, args: OpRead) -> Result<(RpRead, Self::Reader)> {
227        self.io_timeout(Operation::Read, self.inner.read(path, args))
228            .await
229            .map(|(rp, r)| (rp, TimeoutWrapper::new(r, self.io_timeout)))
230    }
231
232    async fn write(&self, path: &str, args: OpWrite) -> Result<(RpWrite, Self::Writer)> {
233        self.io_timeout(Operation::Write, self.inner.write(path, args))
234            .await
235            .map(|(rp, r)| (rp, TimeoutWrapper::new(r, self.io_timeout)))
236    }
237
238    async fn copy(&self, from: &str, to: &str, args: OpCopy) -> Result<RpCopy> {
239        self.timeout(Operation::Copy, self.inner.copy(from, to, args))
240            .await
241    }
242
243    async fn rename(&self, from: &str, to: &str, args: OpRename) -> Result<RpRename> {
244        self.timeout(Operation::Rename, self.inner.rename(from, to, args))
245            .await
246    }
247
248    async fn stat(&self, path: &str, args: OpStat) -> Result<RpStat> {
249        self.timeout(Operation::Stat, self.inner.stat(path, args))
250            .await
251    }
252
253    async fn delete(&self) -> Result<(RpDelete, Self::Deleter)> {
254        self.timeout(Operation::Delete, self.inner.delete())
255            .await
256            .map(|(rp, r)| (rp, TimeoutWrapper::new(r, self.io_timeout)))
257    }
258
259    async fn list(&self, path: &str, args: OpList) -> Result<(RpList, Self::Lister)> {
260        self.io_timeout(Operation::List, self.inner.list(path, args))
261            .await
262            .map(|(rp, r)| (rp, TimeoutWrapper::new(r, self.io_timeout)))
263    }
264
265    async fn presign(&self, path: &str, args: OpPresign) -> Result<RpPresign> {
266        self.timeout(Operation::Presign, self.inner.presign(path, args))
267            .await
268    }
269}
270
271struct TimeoutExecutor {
272    exec: Arc<dyn Execute>,
273    timeout: Duration,
274}
275
276impl TimeoutExecutor {
277    fn new(exec: Arc<dyn Execute>, timeout: Duration) -> Self {
278        Self { exec, timeout }
279    }
280}
281
282impl Execute for TimeoutExecutor {
283    fn execute(&self, f: BoxedStaticFuture<()>) {
284        self.exec.execute(f)
285    }
286
287    fn timeout(&self) -> Option<BoxedStaticFuture<()>> {
288        Some(Box::pin(tokio::time::sleep(self.timeout)))
289    }
290}
291
292#[doc(hidden)]
293pub struct TimeoutWrapper<R> {
294    inner: R,
295
296    timeout: Duration,
297}
298
299impl<R> TimeoutWrapper<R> {
300    fn new(inner: R, timeout: Duration) -> Self {
301        Self { inner, timeout }
302    }
303
304    #[inline]
305    async fn io_timeout<F: Future<Output = Result<T>>, T>(
306        timeout: Duration,
307        op: &'static str,
308        fut: F,
309    ) -> Result<T> {
310        tokio::time::timeout(timeout, fut).await.map_err(|_| {
311            Error::new(ErrorKind::Unexpected, "io operation timeout reached")
312                .with_operation(op)
313                .with_context("timeout", timeout.as_secs_f64().to_string())
314                .set_temporary()
315        })?
316    }
317}
318
319impl<R: oio::Read> oio::Read for TimeoutWrapper<R> {
320    async fn read(&mut self) -> Result<Buffer> {
321        let fut = self.inner.read();
322        Self::io_timeout(self.timeout, Operation::Read.into_static(), fut).await
323    }
324}
325
326impl<R: oio::Write> oio::Write for TimeoutWrapper<R> {
327    async fn write(&mut self, bs: Buffer) -> Result<()> {
328        let fut = self.inner.write(bs);
329        Self::io_timeout(self.timeout, Operation::Write.into_static(), fut).await
330    }
331
332    async fn close(&mut self) -> Result<Metadata> {
333        let fut = self.inner.close();
334        Self::io_timeout(self.timeout, Operation::Write.into_static(), fut).await
335    }
336
337    async fn abort(&mut self) -> Result<()> {
338        let fut = self.inner.abort();
339        Self::io_timeout(self.timeout, Operation::Write.into_static(), fut).await
340    }
341}
342
343impl<R: oio::List> oio::List for TimeoutWrapper<R> {
344    async fn next(&mut self) -> Result<Option<oio::Entry>> {
345        let fut = self.inner.next();
346        Self::io_timeout(self.timeout, Operation::List.into_static(), fut).await
347    }
348}
349
350impl<R: oio::Delete> oio::Delete for TimeoutWrapper<R> {
351    async fn delete(&mut self, path: &str, args: OpDelete) -> Result<()> {
352        let fut = self.inner.delete(path, args);
353        Self::io_timeout(self.timeout, Operation::Delete.into_static(), fut).await
354    }
355
356    async fn close(&mut self) -> Result<()> {
357        let fut = self.inner.close();
358        Self::io_timeout(self.timeout, Operation::Delete.into_static(), fut).await
359    }
360}
361
362#[cfg(test)]
363mod tests {
364    use std::future::pending;
365
366    use futures::StreamExt;
367    use tokio::time::sleep;
368    use tokio::time::timeout;
369
370    use super::*;
371
372    #[derive(Debug, Clone, Default)]
373    struct MockService;
374
375    impl Access for MockService {
376        type Reader = oio::Reader;
377        type Writer = oio::Writer;
378        type Lister = oio::Lister;
379        type Deleter = oio::Deleter;
380
381        fn info(&self) -> Arc<AccessorInfo> {
382            let am = AccessorInfo::default();
383            am.set_native_capability(Capability {
384                read: true,
385                delete: true,
386                ..Default::default()
387            });
388
389            am.into()
390        }
391
392        /// This function will build a reader that always return pending.
393        async fn read(&self, _: &str, _: OpRead) -> Result<(RpRead, Self::Reader)> {
394            Ok((RpRead::new(), Box::new(MockReader)))
395        }
396
397        /// This function will never return.
398        async fn delete(&self) -> Result<(RpDelete, Self::Deleter)> {
399            sleep(Duration::from_secs(u64::MAX)).await;
400
401            Ok((RpDelete::default(), Box::new(())))
402        }
403
404        async fn list(&self, _: &str, _: OpList) -> Result<(RpList, Self::Lister)> {
405            Ok((RpList::default(), Box::new(MockLister)))
406        }
407    }
408
409    #[derive(Debug, Clone, Default)]
410    struct MockReader;
411
412    impl oio::Read for MockReader {
413        fn read(&mut self) -> impl Future<Output = Result<Buffer>> {
414            pending()
415        }
416    }
417
418    #[derive(Debug, Clone, Default)]
419    struct MockLister;
420
421    impl oio::List for MockLister {
422        fn next(&mut self) -> impl Future<Output = Result<Option<oio::Entry>>> {
423            pending()
424        }
425    }
426
427    #[derive(Debug, Clone, Default)]
428    struct MockDeleter;
429
430    impl oio::Delete for MockDeleter {
431        fn delete(&mut self, _: &str, _: OpDelete) -> impl Future<Output = Result<()>> {
432            pending()
433        }
434
435        async fn close(&mut self) -> Result<()> {
436            Ok(())
437        }
438    }
439
440    #[tokio::test]
441    async fn test_operation_timeout() {
442        let srv = MockService;
443        let op = Operator::from_inner(Arc::new(srv))
444            .layer(TimeoutLayer::default().with_timeout(Duration::from_secs(1)));
445
446        let fut = async {
447            let res = op.delete("test").await;
448            assert!(res.is_err());
449            let err = res.unwrap_err();
450            assert_eq!(err.kind(), ErrorKind::Unexpected);
451            assert!(err.to_string().contains("timeout"))
452        };
453
454        timeout(Duration::from_secs(2), fut)
455            .await
456            .expect("this test should not exceed 2 seconds")
457    }
458
459    #[tokio::test]
460    async fn test_io_timeout() {
461        let srv = MockService;
462        let op = Operator::from_inner(Arc::new(srv))
463            .layer(TimeoutLayer::default().with_io_timeout(Duration::from_secs(1)));
464
465        let reader = op.reader("test").await.unwrap();
466
467        let res = reader.read(0..4).await;
468        assert!(res.is_err());
469        let err = res.unwrap_err();
470        assert_eq!(err.kind(), ErrorKind::Unexpected);
471        assert!(err.to_string().contains("timeout"))
472    }
473
474    #[tokio::test]
475    async fn test_list_timeout() {
476        let srv = MockService;
477        let op = Operator::from_inner(Arc::new(srv)).layer(
478            TimeoutLayer::default()
479                .with_timeout(Duration::from_secs(1))
480                .with_io_timeout(Duration::from_secs(1)),
481        );
482
483        let mut lister = op.lister("test").await.unwrap();
484
485        let res = lister.next().await.unwrap();
486        assert!(res.is_err());
487        let err = res.unwrap_err();
488        assert_eq!(err.kind(), ErrorKind::Unexpected);
489        assert!(err.to_string().contains("timeout"))
490    }
491
492    #[tokio::test]
493    async fn test_delete_io_timeout() {
494        use oio::Delete;
495
496        let mut deleter = TimeoutWrapper::new(MockDeleter, Duration::from_secs(1));
497
498        let res = deleter.delete("test", OpDelete::default()).await;
499        assert!(res.is_err());
500        let err = res.unwrap_err();
501        assert_eq!(err.kind(), ErrorKind::Unexpected);
502        assert!(err.to_string().contains("timeout"));
503    }
504
505    #[tokio::test]
506    async fn test_list_timeout_raw() {
507        use oio::List;
508
509        let acc = MockService;
510        let timeout_layer = TimeoutLayer::default()
511            .with_timeout(Duration::from_secs(1))
512            .with_io_timeout(Duration::from_secs(1));
513        let timeout_acc = timeout_layer.layer(acc);
514
515        let (_, mut lister) = Access::list(&timeout_acc, "test", OpList::default())
516            .await
517            .unwrap();
518
519        let res = lister.next().await;
520        assert!(res.is_err());
521        let err = res.unwrap_err();
522        assert_eq!(err.kind(), ErrorKind::Unexpected);
523        assert!(err.to_string().contains("timeout"));
524    }
525}
opendal_layer_timeout/lib.rs

opendal_layer_timeout/
lib.rs
