Skip to content

[adapters] Regression in gather_output_to_host.#6647

Merged
blp merged 1 commit into
mainfrom
gather_output_to_host_regression
Jul 16, 2026
Merged

[adapters] Regression in gather_output_to_host.#6647
blp merged 1 commit into
mainfrom
gather_output_to_host_regression

Conversation

@ryzhyk

@ryzhyk ryzhyk commented Jul 16, 2026

Copy link
Copy Markdown
Contributor

Switching to a streaming exchange in gather_output_to_host introduced a regression where the exchange was disabled on all but one host (the one that the stream is assigned to). As a result these hosts didn't collaborate in delivering outputs to the assigned host.

This bug caused several multihost tests to fail in CI (test_pandas_struct / test_pandas_simple / test_url_encoding). The tests are passing (at least locally) with this fix.

This fix does have a performance implication: views without connectors attached to them are still gathered on one host just to get discarded.

Switching to a streaming exchange in gather_output_to_host introduced a
regression where the exchange was disabled on all but one host (the one that
the stream is assigned to). As a result these hosts didn't collaborate in
delivering outputs to the assigned host.

Signed-off-by: Leonid Ryzhyk <ryzhyk@gmail.com>
@ryzhyk
ryzhyk requested a review from blp July 16, 2026 06:30
@ryzhyk ryzhyk added connectors Issues related to the adapters/connectors crate multihost Related to multihost or distributed pipelines labels Jul 16, 2026

@mythical-fred mythical-fred left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Fix matches the described symptom and the existing multihost tests (test_pandas_struct/test_pandas_simple/test_url_encoding) are the regression net. Perf caveat is called out in the comment.

@blp

blp commented Jul 16, 2026

Copy link
Copy Markdown
Member

I filed #6649 so that the optimization suggested in the comment doesn't get lost so easily.

Merged via the queue into main with commit 7640782 Jul 16, 2026
1 check passed
@blp
blp deleted the gather_output_to_host_regression branch July 16, 2026 16:13
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

connectors Issues related to the adapters/connectors crate multihost Related to multihost or distributed pipelines

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants