Aggregate LNet counters across multiple EFA interfaces in mount_fsx.sh

[davidk-mir]

Purpose

Relates to #989, which added the EFA transport verification to mount_fsx.sh.

verify_efa_transport() reads LNet counters with awk '/send_count:/ {print $2}'. lnetctl net show -v --net efa prints one send_count:/recv_count: line per EFA interface, so on instances with more than one EFA interface the pipeline emits a newline-separated list of numbers instead of a single integer. The subsequent [[ "$send_after" -le "$send_before" ]] comparison then fails with a bash "integer expression expected" error, aborting the verification (and, via the ERR trap, the lifecycle script) — on exactly the instance families most likely to use EFA.

Changes

• Sum the per-interface counters inside awk ({sum += $2} END {print sum + 0}) for the before/after send_count and recv_count reads, so the comparison always receives a single integer regardless of how many EFA interfaces the instance exposes. sum + 0 prints 0 rather than an empty string when no lines match, preserving the existing comparison semantics.

Test Plan

Environment:

• AWS Service: SageMaker HyperPod (Slurm), FSx for Lustre created with EFA enabled
• Instance type: instances with multiple EFA interfaces (p5/p6 families), plus a single-EFA-interface control
• Number of nodes: multi-node cluster

Test commands:

bash -n 1.architectures/5.sagemaker-hyperpod/LifecycleScripts/base-config/mount_fsx.sh

# on an EFA-enabled node with the FSx mount:
lnetctl net show -v --net efa | awk '/send_count:/ {sum += $2} END {print sum + 0}'

Test Results

• Before: on a multi-EFA-interface instance, verify_efa_transport() aborts with integer expression expected — the awk pipeline emits one value per interface as a newline-separated list.
• After: counters aggregate to a single integer, the sum increases under generated I/O, and the verification passes/fails based on actual EFA traffic as intended. Behavior on single-interface instances is unchanged.

Checklist

• I have read the contributing guidelines.
• I am working against the latest main branch.
• I have searched existing open and recently merged PRs to confirm this is not a duplicate.
• The contribution is self-contained with documentation and scripts.
• External dependencies are pinned to a specific version or tag (no latest). (N/A — no dependencies added)
• A README is included or updated with prerequisites, instructions, and known issues. (N/A — focused bug fix)
• New test cases follow the expected directory structure. (N/A — not a test case)

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.