[https://nvbugs/6418912][fix] Prevent BufferIndexHolder test deadlock

[chienchunhung]

Summary

• Configure BufferIndexHolderLifecycleTest with two send and receive slots so move-assignment coverage can hold both the source and destination indices simultaneously.
• Assert the two-slot precondition before acquiring indices, turning any future fixture regression into an immediate failure instead of a one-hour timeout.

Root cause

MoveAssignReleasesPriorThenTransfers acquired two buffer indices before constructing either RAII holder. Both pools default to one slot, so the second acquisition waited indefinitely on the pool condition variable and the move assignment was never exercised.

This was reported by NVBUG 6418912 in A30-CPP-Post-Merge-2; both the send and receive variants timed out.

Why this escaped earlier

PR #14768 added the test, but its recorded pre-merge runs did not execute the A30 C++ post-merge stage. The final PR head used /bot skip after separate performance verification in #14916. The fixture also depended implicitly on CI-provided concurrency environment values, so it could appear healthy anywhere those values provided at least two slots.

This change makes the required capacity explicit and removes that environment dependence.

Validation

• pre-commit run --files cpp/tests/unit_tests/batch_manager/bufferIndexHolderTest.cpp
• Combined validation PR [https://nvbugs/6418912][fix] Fix KV-cache lifecycle test failures #16019 included this PR together with [https://nvbugs/6418912][test] Fix detached-block eviction tests #16011, [https://nvbugs/6418912][test] Fix KV cache reuse accounting tests #16012, and the independent sequence-link fix in [https://nvbugs/6418912][fix] Clear detached KV cache sequence links #16037.
• Targeted A30-CPP-1 pipeline #46619 completed SUCCESS on integration head 1248a186.
• The batch-manager report completed 525 gtests with 0 failures; the BufferIndexHolder cases no longer timed out.
• [https://nvbugs/6418912][fix] Clear detached KV cache sequence links #16037 documents the final end-to-end evidence and fixes the unrelated FillKvCacheWithRequestsAndCompleteOneByOne/5 crash exposed after this timeout was removed.

Earlier standalone pipelines #46503, #46532, and #46562 did not contain all companion fixes, so their later KV-cache failures do not invalidate this BufferIndexHolder fix.

[chienchunhung]

/bot run --disable-fail-fast --stage-list "A30-CPP-Post-Merge-2"

[tensorrt-cicd]

PR_Github #57806 [ run ] triggered by Bot. Commit: 2478323 Link to invocation

[tensorrt-cicd]

PR_Github #57806 [ run ] completed with state SUCCESS. Commit: 2478323
/LLM/main/L0_MergeRequest_PR pipeline #46503 (Partly Tested) completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[chienchunhung]

/bot run --disable-fail-fast --stage-list "A30-CPP-Post-Merge-2"

[tensorrt-cicd]

PR_Github #57836 [ run ] triggered by Bot. Commit: 2478323 Link to invocation

[tensorrt-cicd]

PR_Github #57836 [ run ] completed with state SUCCESS. Commit: 2478323
/LLM/main/L0_MergeRequest_PR pipeline #46532 (Partly Tested) completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[chienchunhung]

/bot run

[tensorrt-cicd]

PR_Github #57865 [ run ] triggered by Bot. Commit: 2478323 Link to invocation

[tensorrt-cicd]

PR_Github #57865 [ run ] completed with state FAILURE. Commit: 2478323
/LLM/main/L0_MergeRequest_PR pipeline #46562 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[chienchunhung]

/bot run --disable-fail-fast --stage-list "DGX_B200-PyTorch-9"

[tensorrt-cicd]

PR_Github #58042 [ run ] triggered by Bot. Commit: 2478323 Link to invocation

[chienchunhung]

Closing this PR as it is subsumed by #16019

[tensorrt-cicd]

PR_Github #58042 [ run ] completed with state ABORTED. Commit: 2478323

Link to invocation