Will the configured timeout take effect on ByteStream::into_async_read?

[hzxa21]

We use ByteStream::into_async_read() to construct an AsyncRead object for the AWS S3 GetObjectOuput so that we can implement streaming read when reading from S3:

let resp = self.client.get_object().bucket(&self.bucket).key(path).send().await?
let stream = resp.body.into_async_read();

// Caller will use read_exact to fetch data in a streaming way when needed
let ret = stream.read_exact(buf).await?;

Does the TimeoutConfig configured for the aws client applicable to both getting back the original resp or does it work for both getting back the resp and also for each poll for the stream?

Example TimeoutConfig:

TimeoutConfig::builder()
      .connect_timeout(Duration::from_millis(3100))
      .read_timeout(Duration::from_millis(10000))
      .operation_attempt_timeout(Duration::from_millis(20000))
      .build()

Based on the above TimeoutConfig, I think can get a TimeoutError if self.client.get_object().bucket(&self.bucket).key(path).send() takes too long to become ready. But I wonder what the behaive is for the following scenarios when using the stream created via into_async_read():

1. The caller doesn't call and poll read_exact on the stream for a long time.
2. The caller constantly calls and polls read_exact on the stream but it remains pending for a long time.
3. The caller constantly calls and polls read_exact on the stream. It becomes ready for the first call but remains pending for a long time in later calls.

aws-sdk version: 0.28
aws-config version: 0.55

[jdisanti]

The current timeouts don't apply to the streaming response body at all. If you need a timeout on that, you'll have to set one up yourself for the time being.

[hzxa21]

The current timeouts don't apply to the streaming response body at all. If you need a timeout on that, you'll have to set one up yourself for the time being.

Got it. Thanks for your prompt reply!

[github-actions[bot]]

Hello! Reopening this discussion to make it searchable.