Skip to content

fix(transfer): clean up failed RDMA handshakes#363

Open
xiaguan wants to merge 1 commit into
masterfrom
fix/rdma-handshake-cleanup
Open

fix(transfer): clean up failed RDMA handshakes#363
xiaguan wants to merge 1 commit into
masterfrom
fix/rdma-handshake-cleanup

Conversation

@xiaguan

@xiaguan xiaguan commented Jun 23, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • make RC handshake completion roll back connecting state and pending QPs when completion fails
  • validate remote memory metadata before committing connection state
  • split RcSession worker-owned I/O state from the session handle so failed pending sessions can release worker/QP resources
  • report RDMA fetch summary bandwidth as GB/s

Fixes #233.

Verification

  • cargo fmt --check
  • cargo test -p pegaflow-transfer
  • cargo clippy --workspace --all-targets -- -D warnings
  • pre-commit hooks during commit amend, including cargo test --release

Review

  • toxic reviewer pass: ready for review

@xiaguan xiaguan force-pushed the fix/rdma-handshake-cleanup branch from 3d24d96 to 7035cb1 Compare June 23, 2026 03:54
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.

Clean up RDMA handshake state when connect fails

1 participant