han.zhao/test_gpu_scripts
1
0
Fork 0
Code Issues Pull Requests 2 Actions Packages Projects Releases Wiki Activity

Validate NCCL 2.27 multinode GDR performance

Browse Source
This commit is contained in:
cs 2026-05-23 15:58:21 +08:00
parent c660e04c99
commit 1f907e9691
8 changed files with 474 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

62
configs/multinode_nccl_nccl227_16g.yaml Normal file
View File

@ -0,0 +1,62 @@
tools:
install_dir: /opt/gpu-test-tools
report:
output_dir: ./reports
format: md
multinode_nccl:
enabled: true
mode: large-message-nccl-2.27.7
hosts:
- name: nccl-gpu-1
addr: 172.72.8.12
slots: 8
- name: nccl-gpu-2
addr: 172.72.8.16
slots: 8
ssh_user: root
ssh_preflight: true
mpirun_path: /usr/mpi/gcc/openmpi-4.1.9a1/bin/mpirun
mpi_ld_preload: null
extra_ld_library_path:
- /usr/mpi/gcc/openmpi-4.1.9a1/lib
- /tmp/nccl-2.27.7-cuda12.4/usr/lib/x86_64-linux-gnu
- /usr/local/cuda-12.4/targets/x86_64-linux/lib
nccl_tests_dir: null
tests:
- all_reduce_perf
- alltoall_perf
topologies:
- nodes: 2
gpus_per_node: 8
label: 2 nodes x 8 GPUs NCCL 2.27.7 16G
begin_size: 16G
end_size: 16G
step_factor: 2
warmup_iters: 1
iters: 3
gpus_per_rank: 1
timeout_sec: 1200
debug: INFO
socket_ifname: bond0
oob_tcp_ifname: bond0
plm_rsh_args: "-o StrictHostKeyChecking=accept-new -o ConnectTimeout=10 -o ServerAliveInterval=30"
ib_gid_index: 3
ib_sl: 5
ib_tc: 136
ib_hca: mlx5_0,mlx5_1,mlx5_6,mlx5_7
ib_timeout: 22
qps_per_connection: 4
min_nchannels: 4
net_plugin: none
nvls_enable: 1
split_data_on_qps: 1
extra_env:
NCCL_DEBUG_SUBSYS: INIT,NET
NCCL_NET_GDR_LEVEL: 5
NCCL_NET_GDR_READ: 1
NCCL_DMABUF_ENABLE: 0
min_peak_busbw_gbps:
allreduce: 480
alltoall: 75

62
configs/multinode_nccl_nccl227_diagnostic.yaml Normal file
View File

@ -0,0 +1,62 @@
tools:
install_dir: /opt/gpu-test-tools
report:
output_dir: ./reports
format: md
multinode_nccl:
enabled: true
mode: diagnostic-nccl-2.27.7
hosts:
- name: nccl-gpu-1
addr: 172.72.8.12
slots: 8
- name: nccl-gpu-2
addr: 172.72.8.16
slots: 8
ssh_user: root
ssh_preflight: true
mpirun_path: /usr/mpi/gcc/openmpi-4.1.9a1/bin/mpirun
mpi_ld_preload: null
extra_ld_library_path:
- /usr/mpi/gcc/openmpi-4.1.9a1/lib
- /tmp/nccl-2.27.7-cuda12.4/usr/lib/x86_64-linux-gnu
- /usr/local/cuda-12.4/targets/x86_64-linux/lib
nccl_tests_dir: null
tests:
- all_reduce_perf
- alltoall_perf
topologies:
- nodes: 2
gpus_per_node: 8
label: 2 nodes x 8 GPUs NCCL 2.27.7
begin_size: 256M
end_size: 256M
step_factor: 2
warmup_iters: 1
iters: 3
gpus_per_rank: 1
timeout_sec: 600
debug: INFO
socket_ifname: bond0
oob_tcp_ifname: bond0
plm_rsh_args: "-o StrictHostKeyChecking=accept-new -o ConnectTimeout=10 -o ServerAliveInterval=30"
ib_gid_index: 3
ib_sl: 5
ib_tc: 136
ib_hca: mlx5_0,mlx5_1,mlx5_6,mlx5_7
ib_timeout: 22
qps_per_connection: 4
min_nchannels: 4
net_plugin: none
nvls_enable: 1
split_data_on_qps: 1
extra_env:
NCCL_DEBUG_SUBSYS: INIT,NET
NCCL_NET_GDR_LEVEL: 5
NCCL_NET_GDR_READ: 1
NCCL_DMABUF_ENABLE: 0
min_peak_busbw_gbps:
allreduce: 480
alltoall: 75

62
configs/multinode_nccl_nccl227_sweep.yaml Normal file
View File

@ -0,0 +1,62 @@
tools:
install_dir: /opt/gpu-test-tools
report:
output_dir: ./reports
format: md
multinode_nccl:
enabled: true
mode: sweep-nccl-2.27.7
hosts:
- name: nccl-gpu-1
addr: 172.72.8.12
slots: 8
- name: nccl-gpu-2
addr: 172.72.8.16
slots: 8
ssh_user: root
ssh_preflight: true
mpirun_path: /usr/mpi/gcc/openmpi-4.1.9a1/bin/mpirun
mpi_ld_preload: null
extra_ld_library_path:
- /usr/mpi/gcc/openmpi-4.1.9a1/lib
- /tmp/nccl-2.27.7-cuda12.4/usr/lib/x86_64-linux-gnu
- /usr/local/cuda-12.4/targets/x86_64-linux/lib
nccl_tests_dir: null
tests:
- all_reduce_perf
- alltoall_perf
topologies:
- nodes: 2
gpus_per_node: 8
label: 2 nodes x 8 GPUs NCCL 2.27.7 sweep
begin_size: 1M
end_size: 4G
step_factor: 4
warmup_iters: 2
iters: 5
gpus_per_rank: 1
timeout_sec: 1200
debug: INFO
socket_ifname: bond0
oob_tcp_ifname: bond0
plm_rsh_args: "-o StrictHostKeyChecking=accept-new -o ConnectTimeout=10 -o ServerAliveInterval=30"
ib_gid_index: 3
ib_sl: 5
ib_tc: 136
ib_hca: mlx5_0,mlx5_1,mlx5_6,mlx5_7
ib_timeout: 22
qps_per_connection: 4
min_nchannels: 4
net_plugin: none
nvls_enable: 1
split_data_on_qps: 1
extra_env:
NCCL_DEBUG_SUBSYS: INIT,NET
NCCL_NET_GDR_LEVEL: 5
NCCL_NET_GDR_READ: 1
NCCL_DMABUF_ENABLE: 0
min_peak_busbw_gbps:
allreduce: 480
alltoall: 75

7
modules/report.py
View File

@ -498,13 +498,14 @@ class ReportGenerator:
if net: if net:
diag_rows.append((topo, net)) diag_rows.append((topo, net))
if diag_rows: if diag_rows:
lines.append("| Topology | NCCL Network | GPU Direct RDMA | GDR Disabled HCAs |") lines.append("| Topology | NCCL Network | GPU Direct RDMA | GDR Enabled HCAs | GDR Disabled HCAs |")
lines.append("|----------|--------------|-----------------|-------------------|") lines.append("|----------|--------------|-----------------|------------------|-------------------|")
for topo, net in diag_rows: for topo, net in diag_rows:
networks = ", ".join(net.get("networks") or []) or "unknown" networks = ", ".join(net.get("networks") or []) or "unknown"
gdr = net.get("gpu_direct_rdma", "UNKNOWN") gdr = net.get("gpu_direct_rdma", "UNKNOWN")
enabled = ", ".join(net.get("gdr_enabled_hcas") or []) or "-"
disabled = ", ".join(net.get("gdr_disabled_hcas") or []) or "-" disabled = ", ".join(net.get("gdr_disabled_hcas") or []) or "-"
lines.append(f"| {topo.get('label', '')} | {networks} | {gdr} | {disabled} |") lines.append(f"| {topo.get('label', '')} | {networks} | {gdr} | {enabled} | {disabled} |")
lines.append("") lines.append("")
failed_topos = [topo for topo in data.get("topologies", []) if topo.get("status") == "FAIL"] failed_topos = [topo for topo in data.get("topologies", []) if topo.get("status") == "FAIL"]
if failed_topos: if failed_topos:

66
reports_multinode_nccl_16g_2x8_nccl227.md Normal file
View File

@ -0,0 +1,66 @@
# GPU Test Report
- **Date:** 2026-05-23T07:56:26.791384
- **Host:** aikubeworker0012
## Overall Acceptance Verdict
**Result: FAIL**
Missing required evidence:
- GPU Info
- Health Check
- Memory Bandwidth
- Compute Throughput
- NVLink/NVSwitch
- NCCL
- Stress Test
- RDMA
- DCGM
- Training
## Summary
| Test | Result |
|------|--------|
| Multi-node NCCL | FAIL |
## Multi-node NCCL / Cross Leaf
Source: nccl-tests-mpirun | Mode: large-message-nccl-2.27.7
- **Hosts:** nccl-gpu-1(172.72.8.12), nccl-gpu-2(172.72.8.16)
- **Preflight:** PASS
### Multi-node NCCL allreduce
| Topology | Peak Bus BW | Peak Size | Avg Bus BW | Threshold | Status |
|----------|-------------|-----------|------------|-----------|--------|
| 2 nodes x 8 GPUs NCCL 2.27.7 16G | 237.86 GB/s | 16G | 238.56 GB/s | >= 480 GB/s | FAIL |
| Topology | NCCL Network | GPU Direct RDMA | GDR Enabled HCAs | GDR Disabled HCAs |
|----------|--------------|-----------------|------------------|-------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 16G | IB | ENABLED | mlx5_0, mlx5_1, mlx5_6, mlx5_7 | - |
| Topology | Return Code | Error / Output Tail |
|----------|-------------|---------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 16G | 0 | aikubeworker0016:1019342:1020412 [4] NCCL INFO comm 0x559f14871c30 rank 12 nranks 16 cudaDev 4 busId 9a000 - Destroy COMPLETE # Out of bounds values : 0 OK # Avg bus bandwidth : 238.555 # # Collective test concluded: all_reduce_perf # |
### Multi-node NCCL alltoall
| Topology | Peak Bus BW | Peak Size | Avg Bus BW | Threshold | Status |
|----------|-------------|-----------|------------|-----------|--------|
| 2 nodes x 8 GPUs NCCL 2.27.7 16G | 28.62 GB/s | 16G | 28.62 GB/s | >= 75 GB/s | FAIL |
| Topology | NCCL Network | GPU Direct RDMA | GDR Enabled HCAs | GDR Disabled HCAs |
|----------|--------------|-----------------|------------------|-------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 16G | IB | ENABLED | mlx5_0, mlx5_1, mlx5_6, mlx5_7 | - |
| Topology | Return Code | Error / Output Tail |
|----------|-------------|---------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 16G | 0 | E aikubeworker0016:1020609:1021756 [5] NCCL INFO comm 0x55f920e55d90 rank 13 nranks 16 cudaDev 5 busId ab000 - Destroy COMPLETE # Out of bounds values : 0 OK # Avg bus bandwidth : 28.6222 # # Collective test concluded: alltoall_perf # |
**Overall: FAIL**
---
*Generated by GPU Test Suite v0.2.0*

101
reports_multinode_nccl_diagnosis_20260523.md
View File

@ -8,9 +8,11 @@
## 当前结论 ## 当前结论
这不是单纯 “IB 不通” 的问题。底层 CUDA RDMA perftest 可以跑到接近单端口 400Gb/s 的水平，但 NCCL 在实际 2 节点通信时把 GPU Direct RDMA 禁用了，导致 NCCL 带宽显著低于验收阈值。 这不是单纯 “IB 不通” 的问题。底层 CUDA RDMA perftest 可以跑到接近单端口 400Gb/s 的水平；最初使用 pip 包里的 NCCL 2.21.5 时，NCCL 在实际 2 节点通信中把 GPU Direct RDMA 禁用了，导致带宽显著偏低。
同时，`nccl-gpu-2` 的 SSH 入口曾因未认证连接过多触发 `MaxStartups` 随机拒绝，导致 `mpirun` 拉起远端 rank 失败。已经做了临时 SSHD 缓解并拿到有效的 2 节点 x 8 GPU allreduce/alltoall 报告；当前剩余核心问题是 NCCL GDR 仍被禁用。 后续临时切换到 apt 包解压出的 NCCL 2.27.7+cuda12.4 后，NCCL GDR 已经恢复启用，2 节点 x 8 GPU allreduce 从 `67.42 GB/s` 提升到 `237.86 GB/s`，alltoall 从 `9.56 GB/s` 提升到 `28.62 GB/s`。当前剩余问题不再是 GDR disabled，而是 GDR enabled 后仍低于当前配置里的验收阈值。
同时，`nccl-gpu-2` 的 SSH 入口曾因未认证连接过多触发 `MaxStartups` 随机拒绝，导致 `mpirun` 拉起远端 rank 失败。已经做了临时 SSHD 缓解并拿到有效的 2 节点 x 8 GPU allreduce/alltoall 报告。
## 已完成的修正 ## 已完成的修正
@ -22,6 +24,8 @@
6. 增加诊断配置 `configs/multinode_nccl_diagnostic.yaml`，固定跑 2 节点 x 8 GPU、256M、`NCCL_DEBUG=INFO` 和 `NCCL_DEBUG_SUBSYS=INIT,NET`。 6. 增加诊断配置 `configs/multinode_nccl_diagnostic.yaml`，固定跑 2 节点 x 8 GPU、256M、`NCCL_DEBUG=INFO` 和 `NCCL_DEBUG_SUBSYS=INIT,NET`。
7. 在 `nccl-gpu-2` 上临时提高 SSHD `MaxStartups` 并缩短 `LoginGraceTime`，缓解未认证连接过多导致的 SSH 随机拒绝。 7. 在 `nccl-gpu-2` 上临时提高 SSHD `MaxStartups` 并缩短 `LoginGraceTime`，缓解未认证连接过多导致的 SSH 随机拒绝。
8. 将 OpenMPI OOB TCP 控制通道固定到 `bond0`，并加入 `plm_rsh_args`，减少 `mpirun` 远端启动受 SSH/host key/接口选择影响的概率。 8. 将 OpenMPI OOB TCP 控制通道固定到 `bond0`，并加入 `plm_rsh_args`，减少 `mpirun` 远端启动受 SSH/host key/接口选择影响的概率。
9. 从 NVIDIA apt 源下载但不安装 `libnccl2=2.27.7-1+cuda12.4`，解压到两台机器 `/tmp/nccl-2.27.7-cuda12.4`，用 `LD_LIBRARY_PATH` 临时覆盖 NCCL 运行库验证。
10. 增强报告解析，能够区分 `GPU Direct RDMA ENABLED` 和 `DISABLED`，并列出 enabled/disabled HCA。
## 关键证据 ## 关键证据
@ -59,6 +63,8 @@ CUDA_VISIBLE_DEVICES=0 ib_write_bw -d mlx5_0 -i 1 --use_cuda=0 --use_cuda_dmabuf
### 3. NCCL 单卡跨节点仍禁用 GDR ### 3. NCCL 单卡跨节点仍禁用 GDR
使用 pip NCCL 2.21.5 时，
已经尝试： 已经尝试：
- `NCCL_NET_GDR_LEVEL=SYS` - `NCCL_NET_GDR_LEVEL=SYS`
@ -77,9 +83,27 @@ NCCL INFO NET/IB : GPU Direct RDMA Disabled for HCA 0 'mlx5_0'
256M allreduce 约 `13.4 GB/s`，明显低于 400Gb/s IB 端口能力。 256M allreduce 约 `13.4 GB/s`，明显低于 400Gb/s IB 端口能力。
### 3.1 NCCL 2.27.7 恢复 GDR
临时使用：
```bash
LD_LIBRARY_PATH=/usr/mpi/gcc/openmpi-4.1.9a1/lib:/tmp/nccl-2.27.7-cuda12.4/usr/lib/x86_64-linux-gnu:/usr/local/cuda-12.4/targets/x86_64-linux/lib
```
2 节点 x 1 GPU 日志显示：
```text
NCCL version 2.27.7+cuda12.4
NET/IB : GPU Direct RDMA Enabled for HCA 0 'mlx5_0'
Channel ... via NET/IB/0/GDRDMA
```
256M allreduce 从 NCCL 2.21.5 的约 `13.4 GB/s` 提升到 `45.2 GB/s`。判断：NCCL 2.21.5 与当前 driver/OFED/H100 组合存在 GDR 判定或注册路径兼容问题；升级 NCCL 是有效修复方向。
### 4. 脚本 2 节点 x 8 GPU 诊断结果 ### 4. 脚本 2 节点 x 8 GPU 诊断结果
原始报告：`reports_multinode_nccl_diagnostic_2x8_sshfix.md` 原始报告：`reports_multinode_nccl_diagnostic_2x8_sshfix.md`，使用 pip NCCL 2.21.5。
| Operation | Topology | Peak Bus BW | Threshold | Status | NCCL Network | GPU Direct RDMA | | Operation | Topology | Peak Bus BW | Threshold | Status | NCCL Network | GPU Direct RDMA |
|-----------|----------|-------------|-----------|--------|--------------|-----------------| |-----------|----------|-------------|-----------|--------|--------------|-----------------|
@ -94,6 +118,31 @@ allreduce 失败原因是带宽不达标，且报告捕获到 GDR 被 NCCL 禁
allreduce 和 alltoall 本轮均正常完成，`returncode=0`、`wrong=0`，失败原因是带宽低于阈值，不是正确性失败。 allreduce 和 alltoall 本轮均正常完成，`returncode=0`、`wrong=0`，失败原因是带宽低于阈值，不是正确性失败。
### 4.1 NCCL 2.27.7 诊断结果
256M 诊断报告：`reports_multinode_nccl_diagnostic_2x8_nccl227_v2.md`
| Operation | Topology | Peak Bus BW | Threshold | Status | NCCL Network | GPU Direct RDMA |
|-----------|----------|-------------|-----------|--------|--------------|-----------------|
| allreduce | 2 nodes x 8 GPUs | `212.19 GB/s` | `>= 480 GB/s` | FAIL | IB | ENABLED |
| alltoall | 2 nodes x 8 GPUs | `28.37 GB/s` | `>= 75 GB/s` | FAIL | IB | ENABLED |
1M 到 4G sweep 报告：`reports_multinode_nccl_sweep_2x8_nccl227.md`
| Operation | Peak Bus BW | Peak Size | Threshold | Status | GPU Direct RDMA |
|-----------|-------------|-----------|-----------|--------|-----------------|
| allreduce | `237.26 GB/s` | `4G` | `>= 480 GB/s` | FAIL | ENABLED |
| alltoall | `28.78 GB/s` | `1G` | `>= 75 GB/s` | FAIL | ENABLED |
16G 大包报告：`reports_multinode_nccl_16g_2x8_nccl227.md`
| Operation | Peak Bus BW | Peak Size | Threshold | Status | GPU Direct RDMA |
|-----------|-------------|-----------|-----------|--------|-----------------|
| allreduce | `237.86 GB/s` | `16G` | `>= 480 GB/s` | FAIL | ENABLED |
| alltoall | `28.62 GB/s` | `16G` | `>= 75 GB/s` | FAIL | ENABLED |
解释：NCCL 2.27.7 已经修复 GDR 禁用问题，且性能提升明显；但在当前跨节点/跨 Leaf 环境和当前阈值下仍不达标。allreduce 约稳定在 `238 GB/s`，alltoall 约稳定在 `28-29 GB/s`。
### 5. SSHD MaxStartups 阻塞已临时缓解 ### 5. SSHD MaxStartups 阻塞已临时缓解
`nccl-gpu-2` 曾显示： `nccl-gpu-2` 曾显示：
@ -109,13 +158,20 @@ maxstartups 10:30:100
kex_exchange_identification: Connection closed by remote host kex_exchange_identification: Connection closed by remote host
``` ```
已临时改为： 先临时改为：
```text ```text
MaxStartups 120:30:240 MaxStartups 120:30:240
LoginGraceTime 20 LoginGraceTime 20
``` ```
后续外部未认证连接继续上涨到 `110 of 120-240 startups`，测试窗口进一步临时改为：
```text
MaxStartups 500:30:1000
LoginGraceTime 5
```
改完后从 0012 连续 SSH 0016 5 次成功，2 节点 `mpirun hostname` 成功，2 节点 x 8 GPU allreduce/alltoall 也都能跑出有效结果。 改完后从 0012 连续 SSH 0016 5 次成功，2 节点 `mpirun hostname` 成功，2 节点 x 8 GPU allreduce/alltoall 也都能跑出有效结果。
### 6. `nvidia_peermem` legacy 模式实验无效 ### 6. `nvidia_peermem` legacy 模式实验无效
@ -139,18 +195,27 @@ NET/IB : GPU Direct RDMA Disabled for HCA 0 'mlx5_0'
## 当前阻塞 ## 当前阻塞
### 阻塞 1：NCCL 禁用 GPU Direct RDMA ### 阻塞 1：当前生产 NCCL 版本过旧，GDR 被禁用
现象： 现象：
- IB 能被 NCCL 识别：`Using network IB` - pip NCCL 2.21.5：`GPU Direct RDMA Disabled`，2x8 allreduce `67.42 GB/s`
- 400Gb/s HCA 被 NCCL 选中：`mlx5_0, mlx5_1, mlx5_6, mlx5_7` - 临时 NCCL 2.27.7：`GPU Direct RDMA Enabled`，2x8 allreduce `237.86 GB/s`
- 但 NCCL 明确禁用 GDR：`GPU Direct RDMA Disabled` - 因此，生产测试环境应避免继续使用 pip NCCL 2.21.5 作为多机 NCCL 验收运行库
- perftest 的经典 CUDA RDMA 又能跑到 `387.16 Gb/s`
判断：底层 RDMA 能力存在，但 NCCL 的 GDR 判定/注册路径没有打通。优先排查 NCCL 与 NVIDIA driver、OFED、`nvidia_peermem`、NCCL net plugin/内部 IB 后端之间的兼容性。 判断：底层 RDMA 能力存在，GDR 禁用主要由旧 NCCL 版本触发。建议正式安装并固定 NCCL 2.27.7+cuda12.4 或更新的已验证版本。
### 阻塞 2：`nccl-gpu-2` SSH 存在外部连接压力 ### 阻塞 2：GDR enabled 后带宽仍低于当前阈值
现象：
- 2x8 16G allreduce：`237.86 GB/s`，阈值 `>= 480 GB/s`
- 2x8 16G alltoall：`28.62 GB/s`，阈值 `>= 75 GB/s`
- 已使用 4 个 400Gb/s HCA：`mlx5_0, mlx5_1, mlx5_6, mlx5_7`
判断：需要确认当前 PDF/config 阈值是否适用于跨 Leaf 两节点场景；如果阈值确实要求跨 Leaf 也达到这些数值，则还需要继续查链路聚合、多 rail 使用、交换网络、NCCL net plugin/SHARP 或 rail mapping。
### 阻塞 3：`nccl-gpu-2` SSH 存在外部连接压力
现象： 现象：
@ -164,13 +229,19 @@ NET/IB : GPU Direct RDMA Disabled for HCA 0 'mlx5_0'
## 建议下一步 ## 建议下一步
1. 从网络/安全侧处理 `172.239.10.85` 等来源的 SSH 未认证连接压力，或者保留更高的 `MaxStartups` 配置作为测试窗口临时策略。 1. 从网络/安全侧处理 `172.239.10.85` 等来源的 SSH 未认证连接压力，或者保留更高的 `MaxStartups` 配置作为测试窗口临时策略。
2. 尝试安装或启用匹配当前 OFED/driver 的 NCCL net plugin；当前日志显示 `No plugin found (libnccl-net.so)`，NCCL 使用的是 internal network plugin。 2. 正式安装并固定已验证的 NCCL 2.27.7+cuda12.4 或更新版本，不要依赖 pip NCCL 2.21.5；当前 `/tmp/nccl-2.27.7-cuda12.4` 只是临时解压验证。
3. 用同版本软件栈补测 `nccl-tests` + NCCL net plugin 后的 GDR 状态，核心判据是报告里 `GPU Direct RDMA` 从 `DISABLED` 变成未禁用，且 2x8 带宽显著抬升。 3. 尝试安装或启用匹配当前 OFED/driver 的 NCCL net plugin/SHARP；当前日志显示 `Could not find: libnccl-net.so`，NCCL 使用的是 internal IB plugin。
4. 如果仍禁用 GDR，再继续查 NVIDIA driver 580.159.03、OFED 26.01、NCCL 2.21.5 与 H100/IB NDR 组合的兼容矩阵。 4. 核对跨 Leaf 链路的 rail mapping、交换机端口速率、路由和拥塞计数，确认 4 个 400Gb/s HCA 是否都在跨节点通信中充分利用。
5. GDR 修复后重跑完整多机配置：2 节点 x 8 GPU，至少覆盖 `all_reduce_perf` 和 `alltoall_perf`，消息大小从 `1K` 到 `16G`。 5. 确认当前 `allreduce >= 480 GB/s`、`alltoall >= 75 GB/s` 阈值是否应直接用于跨 Leaf 两节点场景；如果是，继续按链路和 NCCL rail 聚合方向排查。
## 当前可交付物 ## 当前可交付物
- `configs/multinode_nccl_diagnostic.yaml`：多机多卡诊断配置 - `configs/multinode_nccl_diagnostic.yaml`：多机多卡诊断配置
- `configs/multinode_nccl_nccl227_diagnostic.yaml`：NCCL 2.27.7 256M 诊断配置
- `configs/multinode_nccl_nccl227_sweep.yaml`：NCCL 2.27.7 1M 到 4G sweep 配置
- `configs/multinode_nccl_nccl227_16g.yaml`：NCCL 2.27.7 16G 大包配置
- `reports_multinode_nccl_diagnostic_2x8_sshfix.md`：脚本生成的原始 2x8 诊断报告 - `reports_multinode_nccl_diagnostic_2x8_sshfix.md`：脚本生成的原始 2x8 诊断报告
- `reports_multinode_nccl_diagnostic_2x8_nccl227_v2.md`：NCCL 2.27.7 256M 诊断报告
- `reports_multinode_nccl_sweep_2x8_nccl227.md`：NCCL 2.27.7 1M 到 4G sweep 报告
- `reports_multinode_nccl_16g_2x8_nccl227.md`：NCCL 2.27.7 16G 大包报告
- `reports_multinode_nccl_diagnosis_20260523.md`：本中文诊断总结 - `reports_multinode_nccl_diagnosis_20260523.md`：本中文诊断总结

66
reports_multinode_nccl_diagnostic_2x8_nccl227_v2.md Normal file
View File

@ -0,0 +1,66 @@
# GPU Test Report
- **Date:** 2026-05-23T07:53:24.460277
- **Host:** aikubeworker0012
## Overall Acceptance Verdict
**Result: FAIL**
Missing required evidence:
- GPU Info
- Health Check
- Memory Bandwidth
- Compute Throughput
- NVLink/NVSwitch
- NCCL
- Stress Test
- RDMA
- DCGM
- Training
## Summary
| Test | Result |
|------|--------|
| Multi-node NCCL | FAIL |
## Multi-node NCCL / Cross Leaf
Source: nccl-tests-mpirun | Mode: diagnostic-nccl-2.27.7
- **Hosts:** nccl-gpu-1(172.72.8.12), nccl-gpu-2(172.72.8.16)
- **Preflight:** PASS
### Multi-node NCCL allreduce
| Topology | Peak Bus BW | Peak Size | Avg Bus BW | Threshold | Status |
|----------|-------------|-----------|------------|-----------|--------|
| 2 nodes x 8 GPUs NCCL 2.27.7 | 212.19 GB/s | 256M | 211.75 GB/s | >= 480 GB/s | FAIL |
| Topology | NCCL Network | GPU Direct RDMA | GDR Enabled HCAs | GDR Disabled HCAs |
|----------|--------------|-----------------|------------------|-------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 | IB | ENABLED | mlx5_0, mlx5_1, mlx5_6, mlx5_7 | - |
| Topology | Return Code | Error / Output Tail |
|----------|-------------|---------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 | 0 | 0016:1009332:1009965 [2] NCCL INFO comm 0x56388eec2e40 rank 10 nranks 16 cudaDev 2 busId 3a000 - Destroy COMPLETE aikubeworker0012:2144366:2144531 [5] NCCL INFO comm 0x556e4fcf5280 rank 5 nranks 16 cudaDev 5 busId ab000 - Destroy COMPLETE |
### Multi-node NCCL alltoall
| Topology | Peak Bus BW | Peak Size | Avg Bus BW | Threshold | Status |
|----------|-------------|-----------|------------|-----------|--------|
| 2 nodes x 8 GPUs NCCL 2.27.7 | 28.37 GB/s | 256M | 28.32 GB/s | >= 75 GB/s | FAIL |
| Topology | NCCL Network | GPU Direct RDMA | GDR Enabled HCAs | GDR Disabled HCAs |
|----------|--------------|-----------------|------------------|-------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 | IB | ENABLED | mlx5_0, mlx5_1, mlx5_6, mlx5_7 | - |
| Topology | Return Code | Error / Output Tail |
|----------|-------------|---------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 | 0 | 0012:2144547:2144713 [4] NCCL INFO comm 0x55896a1dae20 rank 4 nranks 16 cudaDev 4 busId 9a000 - Destroy COMPLETE aikubeworker0016:1010164:1010881 [2] NCCL INFO comm 0x565344db7790 rank 10 nranks 16 cudaDev 2 busId 3a000 - Destroy COMPLETE |
**Overall: FAIL**
---
*Generated by GPU Test Suite v0.2.0*

66
reports_multinode_nccl_sweep_2x8_nccl227.md Normal file
View File

@ -0,0 +1,66 @@
# GPU Test Report
- **Date:** 2026-05-23T07:54:48.990378
- **Host:** aikubeworker0012
## Overall Acceptance Verdict
**Result: FAIL**
Missing required evidence:
- GPU Info
- Health Check
- Memory Bandwidth
- Compute Throughput
- NVLink/NVSwitch
- NCCL
- Stress Test
- RDMA
- DCGM
- Training
## Summary
| Test | Result |
|------|--------|
| Multi-node NCCL | FAIL |
## Multi-node NCCL / Cross Leaf
Source: nccl-tests-mpirun | Mode: sweep-nccl-2.27.7
- **Hosts:** nccl-gpu-1(172.72.8.12), nccl-gpu-2(172.72.8.16)
- **Preflight:** PASS
### Multi-node NCCL allreduce
| Topology | Peak Bus BW | Peak Size | Avg Bus BW | Threshold | Status |
|----------|-------------|-----------|------------|-----------|--------|
| 2 nodes x 8 GPUs NCCL 2.27.7 sweep | 237.26 GB/s | 4G | 150.62 GB/s | >= 480 GB/s | FAIL |
| Topology | NCCL Network | GPU Direct RDMA | GDR Enabled HCAs | GDR Disabled HCAs |
|----------|--------------|-----------------|------------------|-------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 sweep | IB | ENABLED | mlx5_0, mlx5_1, mlx5_6, mlx5_7 | - |
| Topology | Return Code | Error / Output Tail |
|----------|-------------|---------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 sweep | 0 | aikubeworker0012:2145024:2145189 [0] NCCL INFO comm 0x561f7dc1f780 rank 0 nranks 16 cudaDev 0 busId 18000 - Destroy COMPLETE # Out of bounds values : 0 OK # Avg bus bandwidth : 150.624 # # Collective test concluded: all_reduce_perf # |
### Multi-node NCCL alltoall
| Topology | Peak Bus BW | Peak Size | Avg Bus BW | Threshold | Status |
|----------|-------------|-----------|------------|-----------|--------|
| 2 nodes x 8 GPUs NCCL 2.27.7 sweep | 28.78 GB/s | 1G | 23.57 GB/s | >= 75 GB/s | FAIL |
| Topology | NCCL Network | GPU Direct RDMA | GDR Enabled HCAs | GDR Disabled HCAs |
|----------|--------------|-----------------|------------------|-------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 sweep | IB | ENABLED | mlx5_0, mlx5_1, mlx5_6, mlx5_7 | - |
| Topology | Return Code | Error / Output Tail |
|----------|-------------|---------------------|
| 2 nodes x 8 GPUs NCCL 2.27.7 sweep | 0 | r0012:2145213:2145384 [7] NCCL INFO comm 0x558d54228110 rank 7 nranks 16 cudaDev 7 busId db000 - Destroy COMPLETE aikubeworker0016:1014703:1015544 [0] NCCL INFO comm 0x55ed6d99d8e0 rank 8 nranks 16 cudaDev 0 busId 18000 - Destroy COMPLETE |
**Overall: FAIL**
---
*Generated by GPU Test Suite v0.2.0*