Repository navigation
fix(fs-aws): forward body stream errors in readStream - #1743
Merged
Merged
Conversation
blacha
approved these changes
Nov 15, 2026
Merged
blacha
pushed a commit
that referenced
this pull request
Nov 15, 2026
🤖 I have created a release *beep* *boop* --- <details><summary>fs-aws: 11.6.4</summary> ## [11.6.4](fs-aws-v11.6.3...fs-aws-v11.6.4) (2026-11-15) ### Bug Fixes * **fs-aws:** forward body stream errors in readStream ([#1743](#1743)) ([36534fd](36534fd)) </details> --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
olitreadwell
pushed a commit
to olitreadwell/argo-tasks
that referenced
this pull request
Nov 16, 2026
…1445) ### Motivation `stac-validate` crashes when the connection to S3 drops while checksumming a file (`ECONNRESET`), failing collections that are actually valid. See TDE-2196. This needed two fixes: 1. `@chunkd/fs-aws` `readStream()` didn't pass body errors to the returned stream, so the error crashed the process instead of being catchable. Fixed in blacha/chunkd#1743 and released in `@chunkd/fs-aws` 11.6.4. 2. Once the error is catchable, the read should be retried rather than failing the collection. That's this PR. ### Modifications - Bump `@chunkd/fs-aws` to 11.6.4, which includes the `readStream` error fix. - Add a shared `hashFile(url)` helper in `src/utils/hash.ts` that hashes a file and retries the read on transient network errors (`ECONNRESET`, `ETIMEDOUT`, `EPIPE`) using the existing `retryOnError`: up to 3 attempts, 1s then 2s delay, re-opening the stream each time. Other errors (e.g. access denied) fail straight away as before. - Use `hashFile` everywhere a file was hashed with `hashStream(fsa.readStream(...))`: `stac-validate`, `copy` and `mapsheet-coverage`. ### Verification - Added unit tests for `hashFile`: - a read that fails with `ECONNRESET` is retried and returns the correct hash - a read that fails with a non-network error is not retried - Existing unit tests pass - Reproduced locally with a fake S3 endpoint and the chunkd fix: - connection dropped mid-download: retried, checksum passes on the second request - S3 Access Denied: fails immediately after one request, no retry
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
readStream()pipes the S3 response body into aPassThroughwith.pipe(), which doesn't forward errors. If the connection drops mid-download (e.g.ECONNRESET), the error is emitted on the AWS SDK'sChecksumStreamwith no listener, so Node throws an unhandled'error'event and the process exits. Error handlers on the returned stream never see it.When the response has no checksum header, the same drop makes the returned stream hang instead: it never ends or errors.
We hit this in argo-tasks
stac-validate, which crashed partway through checksumming a large collection.Modification
Listen for
erroron the response body and destroy the returned stream with anFsError, using the same wrapping as the existing.catch().Verification
ECONNRESETonChecksumStream; after, the error is emitted on the returned stream as anFsError