andygrove commented on issue #2023:
URL: 
https://github.com/apache/datafusion-ballista/issues/2023#issuecomment-4997784803

   **Cannot reproduce on current `main`.**
   
   Re-ran the exact repro from this issue against `apache/main` @ `0f6ec8c6`, 
on the same cluster shape and config as originally reported. **Q18 completed in 
750.5s and returned 100 rows** — no hang.
   
   | | As filed | Re-run |
   |---|---|---|
   | Ballista | 54.0.0-rc2 | `main` @ `0f6ec8c6` |
   | Cluster | 2 × (8 CPU / 56Gi), 48GB pool | same (6 GB/slot, one executor 
per node) |
   | Query | Q18, SF1000, p32, 1 iteration | same |
   | Config | `prefer_hash_join=false`, `enable_dynamic_filter_pushdown=false`, 
`adaptive.enabled=true` | same |
   | Outcome | hang (~11h, 1m CPU, 24 GB held) | **750.5s, 100 rows** |
   
   The original signature was both executors pinned at 1m CPU with memory held. 
In this run they sustained 1000–6000m CPU with GB-scale memory churn for the 
full 13 minutes, and the scheduler logged the `Adaptive Query Planning is 
EXPERIMENTAL` warning, confirming the adaptive path was exercised rather than 
silently skipped. The scheduler log was clean — no replan/cancel errors, so 
this run also did not hit #2063.
   
   This was not bisected, so which change resolved it is unknown.
   
   Closing as no-longer-reproducible; happy to reopen if it resurfaces.
   


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to