RetriableReadableStream ineffective now that it handles "aborted" event

[mheggeseth]

• Package Name: @azure/storage-blob
• Package Version: 12.26.0
• Operating system:
• nodejs
  • version: v20.18.1
• browser
  • name/version:
• typescript
  • version:
• Is the bug related to documentation in
  • README.md
  • source code documentation
  • SDK API docs on https://learn.microsoft.com

Describe the bug
The aborted event handling in sdk/storage/storage-blob/src/utils/RetriableReadableStream.ts added in #24141 renders the class ineffective for its purpose when the http response stream returned by the initial download call terminates because of a server timeout (specified in retryOptions.tryTimeoutInMs given to new BlobServiceClient) resulting in a connection reset and the http response emitting the aborted event which causes RetriableReadableStream to destroy itself, thereby not retrying the download from the byte offset at which it left off.

From Node's http.request docs:

In the case of a premature connection close after the response is received, the following events will be emitted in the following order:

• 'socket'
• 'response'
  • 'data' any number of times, on the res object
• (connection closed here)
• 'aborted' on the res object
• 'close'
• 'error' on the res object with an error with message 'Error: aborted' and code 'ECONNRESET'
• 'close' on the res object

To Reproduce
Steps to reproduce the behavior:

1. Create a new BlobServiceClient with retryOptions.tryTimeoutInMs=3000
2. Call BlobClient.downloadToFile for a blob that would take longer than 3s to download
3. The call to downloadToFile rejects because RetriableReadableStream destroys itself upon handling aborted rather than retrying

Expected behavior
Downloading the blob resumes from the byte offset at which the initial download left off after the per-request timeout.

[github-actions]

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @xgithubtriage.