[CI] Modify 4-card container startup config and move test case

[EmmonsCurse]

Motivation

• Improve stability and compatibility of 4-card CI jobs with RDMA-enabled environments
• Ensure proper device mounting and resource limits for multi-card inference tests

Modifications

• Update 4-card container startup configuration:
  • Add RDMA device auto-detection and injection (RDMA_DEVICES)
  • Mount /dev/infiniband/rdma_cm for RDMA communication
  • Add --ulimit memlock=-1:-1 to support RDMA memory locking
  • Add required capabilities (SYS_PTRACE, IPC_LOCK)
  • Adjust shared memory configuration (--shm-size=64g)
• Add new environment variables for router, connector, and RDMA ports
• Enable CUDA cleanup flag (CLEAN_CUDA=1)
• Reorganize e2e test case:
  • Move test_ernie_03b_pd_router_v1_rdma_tp2.py into 4cards_cases directory for clearer test grouping

Usage or Command

N/A

Accuracy Tests

N/A

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[EmmonsCurse]

/skip-ci ci_iluvatar
/skip-ci ci_hpu
/skip-ci build_xpu
/skip-ci coverage
/skip-ci stable_test
/skip-ci base_test
/skip-ci pre_ce_test
/skip-ci logprob_test

[PaddlePaddle-bot]

🤖 AI Code Review | 2026-04-13

📋 Review 摘要

PR 概述：优化 4 卡 CI 容器启动配置，支持 RDMA 环境

变更范围：.github/workflows/_gpu_4cards_case_test.yml、tests/e2e/4cards_cases/

影响面 Tag：[CI]

📝 PR 规范检查

PR 标题和描述均符合规范，Motivation 和 Modifications 章节清晰。

问题

未发现阻塞性问题。

总体评价

本次 PR 合理地添加了 RDMA 设备挂载、必要的 Docker capabilities（SYS_PTRACE、IPC_LOCK）、共享内存配置（64G）和 ulimit 设置，以支持 RDMA 环境下的 4 卡测试。测试文件的导入路径和 GPU 设备分配修正也是正确的。

💡 改进建议（非阻塞性）

虽然不在本次 PR 变更范围内，但注意到 .github/workflows/_gpu_4cards_case_test.yml:158 的 PORTS 数组未包含本次新增的 FD_ROUTER_PORT、FD_CONNECTOR_PORT、FD_RDMA_PORT，可能需要后续补充端口清理逻辑。