Skip to content

Enable Presto Multi-Container Profiling#260

Open
karthikeyann wants to merge 3 commits into
rapidsai:mainfrom
karthikeyann:fea-multicontainer-profile
Open

Enable Presto Multi-Container Profiling#260
karthikeyann wants to merge 3 commits into
rapidsai:mainfrom
karthikeyann:fea-multicontainer-profile

Conversation

@karthikeyann
Copy link
Copy Markdown
Contributor

This PR enables profiling with multiple containers.
Each worker container is profiled separately, and written to separate files with worker id as suffix in their profile file name.

@karthikeyann karthikeyann requested a review from a team as a code owner March 20, 2026 04:15
@karthikeyann
Copy link
Copy Markdown
Contributor Author

@paul-aiyedun please review this PR.

Copy link
Copy Markdown
Contributor

@paul-aiyedun paul-aiyedun left a comment

Choose a reason for hiding this comment

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

Changes overall look good to me. I had a question about the 10 seconds wait. Also, can we remove the --single-container option now? I believe this was added because of the profiling issue.

Comment on lines +56 to +57
# wait for 10 seconds
sleep 10
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

The current implementation works without a 10 seconds wait. Why is this now needed?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

IIRC,nsys stop needs time to write the files. Otherwise it errors out sometimes.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

This function is executed for each query. Adding a 10 second delay here could be significant, especially when running benchmarks for the entire suite.

fi

if [[ "$PROFILE" == "ON" && $NUM_WORKERS -gt 1 && "$SINGLE_CONTAINER" == "false" ]]; then
echo "Error: multi-worker --profile argument is only currently supported with the --single-container option"
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Can we remove the --single-container option, since it was added as a workaround for this issue?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants