moomindani opened a new pull request, #72304:
URL: https://github.com/apache/airflow/pull/72304

   `DatabricksHook.get_run` and its async twin read only the first page of 
`2.2/jobs/runs/get`. The Jobs API returns at most 100 entries of a run's 
`tasks` and `job_clusters` per page and hands back a `next_page_token` for the 
rest, so on a job with more tasks than that the run came back carrying an 
arbitrary subset of them — the entries are not returned in declaration order, 
so which ones went missing was not predictable.
   
   The visible effect is in `extract_failed_task_errors[_async]`, which walks 
`run_info["tasks"]` to attach each failed task's own error to the message the 
task fails with: a failure on a later page falls back to the generic run state 
message instead.
   
   `get_run_tasks` already paginated, so it now delegates to `get_run` and the 
two cannot drift apart again.
   
   Verified against a live workspace, using a job of 101 `condition_task` 
entries (no compute) plus one failing notebook task:
   
   * before: `get_run()["tasks"]` returned 100 entries while `get_run_tasks()` 
returned 103, and the keys missing from the former were an arbitrary trio
   * after: both return the complete list, with the run-level fields unchanged
   * `job_clusters` is paginated too rather than repeated on each page — a job 
with two job clusters returns both on page 1 and none on page 2 — so merging it 
is correct
   
   Both new unit tests fail without the change (`call_count 1 == 2`), and the 
pre-existing `get_run_tasks` multi-page test passes unchanged.
   
   ---
   
   ##### Was generative AI tooling used to co-author this PR?
   
   - [X] Yes — Claude Code (Opus 5)
   
   Generated-by: Claude Code (Opus 5) following [the 
guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions)
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to