1 基本的板卡环境：
1.1 硬件环境
使用Atlas200i A2加速卡，自研底板。（我方已经自研过好几框自研底板，在未使用集群情况下使用npu推理正常。）
1.2 软件环境
（CANN版本：8.0.0.beta1）
1.2.1 uname –a打印
Linux node1 5.10.0+ #1 SMP Mon Apr 15 00:00:00 CST 2024 aarch64 aarch64 aarch64 GNU/Linux
1.2.2 cat /etc/os-release打印
NAME="openEuler"
ID="openEuler"
VERSION_ID="22.03"
ANSI_COLOR="0;31"
VERSION="22.03 LTS dracut-055-4.oe2203"
PRETTY_NAME="openEuler 22.03 LTS dracut-055-4.oe2203 (Initramfs)"
DRACUT_VERSION="055-4.oe2203"
1.2.3 npu-smi info打印
+--------------------------------------------------------------------------------------------------------+
| npu-smi 24.1.rc1 Version: 24.1.rc1 |
+-------------------------------+-----------------+------------------------------------------------------+
| NPU Name | Health | Power(W) Temp(C) Hugepages-Usage(page) |
| Chip Device | Bus-Id | AICore(%) Memory-Usage(MB) |
+===============================+=================+======================================================+
| 0 310B1 | OK | 8.2 49 15 / 15 |
| 0 0 | NA | 0 1197 / 11578 |
+===============================+=================+======================================================+
1.2.4 固件版本查看
/var/davinci/driver/upgrade-tool --device_index -1 --component -1 --version
{
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(11).
{"device_id":0, "component":hboot1a, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(12).
{"device_id":0, "component":hboot1b, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(13).
{"device_id":0, "component":hboot2, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(14).
{"device_id":0, "component":ddr, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(16).
{"device_id":0, "component":hsm, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(18).
{"device_id":0, "component":hlink, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(25).
{"device_id":0, "component":Sys_Base_Config, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(30).
{"device_id":0, "component":atf, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(31).
{"device_id":0, "component":Usr_Base_Config, "version":7.1.0.6.220}
}
2 安装问题
//安装的组件版本及过程，问题描述。
2.1 dockre版本：
docker info
Client:
Version: 24.0.6
Context: default
Debug Mode: false
Server:
Containers: 37
Running: 34
Paused: 0
Stopped: 3
Images: 42
Server Version: 24.0.6
Storage Driver: overlay2
Backing Filesystem: extfs
Supports d_type: true
Using metacopy: false
Native Overlay Diff: true
userxattr: false
Logging Driver: json-file
Cgroup Driver: cgroupfs
Cgroup Version: 1
Plugins:
Volume: local
Network: bridge host ipvlan macvlan null overlay
Log: awslogs fluentd gcplogs gelf journald json-file local logentries splunk syslog
Swarm: inactive
Runtimes: ascend io.containerd.runc.v2 runc
Default Runtime: runc
Init Binary: docker-init
containerd version: 7880925980b188f4c97b462f709d0db8e8962aff
runc version: v1.1.9-0-gccaecfc
init version: de40ad0
Security Options:
seccomp
Profile: builtin
Kernel Version: 5.10.0+
Operating System: openEuler 22.03 LTS dracut-055-4.oe2203 (Initramfs) (containerized)
OSType: linux
Architecture: aarch64
CPUs: 4
Total Memory: 11.31GiB
Name: node1
ID: 58f99f10-2560-4039-b3f7-30d1c7ad7ed7
Docker Root Dir: /var/lib/docker
Debug Mode: false
Experimental: false
Insecure Registries:
127.0.0.0/8
Live Restore Enabled: false
Product License: Community Engine
2.2 ascend-docker-runtime版本：
-sh-5.1# /usr/local/Ascend/Ascend-Docker-Runtime/ascend-docker-runtime --version
runc version 1.1.9
commit: v1.1.9-0-gccaecfc
spec: 1.0.2-dev
go: go1.20.7
libseccomp: 2.5.1
-sh-5.1#
2.3 docker配置：
cat /etc/docker/daemon.json
{
"log-opts": {
"max-size": "5m",
"max-file":"3"
},
"data-root": "/var/lib/docker",
"exec-opts": ["native.cgroupdriver=cgroupfs"],
"runtimes": {
"ascend": {
"path": "/usr/local/Ascend/Ascend-Docker-Runtime/ascend-docker-runtime",
"runtimeArgs": []
}
}
}
2.4 由于系统是欧拉，ascend-device-plugin是ubuntu，驱动文件分散到各目录，所以新建了目录：/usr/lib64-ascend将需要的驱动放置到该目录
pwd
/usr/lib64-ascend
-sh-5.1# ls
libaicpu_processer.so libascend_aac_adp.so libascend_aac_sbr_dec.so libascend_drv_baselib.so libascend_protobuf.so.3.13.0.0 libdrvdsmi.so libslog.so libtsdclient.so
libaicpu_prof.so libascend_aac_comm.so libascend_aac_sbr_enc.so libascend_drvdmp.so libc_sec.so libdrvdsmi_efuse.so libtsd_common.so
libaicpu_scheduler.so libascend_aac_dec.so libascend_aicpu_datadump.so libascend_drvupgrade.so libdcmi.so libdrvgpiocfg.so libtsd_eventclient.so
libaicpu_sharder.so libascend_aac_enc.so libascend_camera_mpi_mfusion.so libascend_hal.so libdevmmap.so libmmpa.so libtsdaemon.so
-sh-5.1#
直接使用docker运行ascend-device-plugin镜像(v7.3.1和v26.0.0)
```bash
docker run --rm -it --privileged \
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
-v /usr/lib64-ascend:/usr/lib64-ascend \
-v /dev:/dev \
-v /sys:/sys \
-v /usr/local/sbin/npu-smi:/usr/local/sbin/npu-smi \
-v /tmp:/tmp \
-e LD_LIBRARY_PATH=/usr/lib64-ascend:/usr/local/Ascend/driver/lib64 \
ascend-k8sdeviceplugin:v7.3.1 \
/lib/ld-linux-aarch64.so.1 /usr/local/sbin/npu-smi info
```
报错：`[ERROR] DRV(1,npu-smi):2026-06-02-10:02:55.123.951 [hdc_core.c:396][hdc] [hdcPcieInit 396] HDC init failed, driver may not load.`
目前在k8s集群(1.23.17)中运行ascend-device-plugin和直接用docker形式运行，都是报错：HDC初始化失败
点赞
收藏
回复
分享
举报
1 基本的板卡环境：
1.1 硬件环境
使用Atlas200i A2加速卡，自研底板。（我方已经自研过好几框自研底板，在未使用集群情况下使用npu推理正常。）
1.2 软件环境
（CANN版本：8.0.0.beta1）
1.2.1 uname –a打印
Linux node1 5.10.0+ #1 SMP Mon Apr 15 00:00:00 CST 2024 aarch64 aarch64 aarch64 GNU/Linux
1.2.2 cat /etc/os-release打印
NAME="openEuler"
ID="openEuler"
VERSION_ID="22.03"
ANSI_COLOR="0;31"
VERSION="22.03 LTS dracut-055-4.oe2203"
PRETTY_NAME="openEuler 22.03 LTS dracut-055-4.oe2203 (Initramfs)"
DRACUT_VERSION="055-4.oe2203"
1.2.3 npu-smi info打印
+--------------------------------------------------------------------------------------------------------+
| npu-smi 24.1.rc1 Version: 24.1.rc1 |
+-------------------------------+-----------------+------------------------------------------------------+
| NPU Name | Health | Power(W) Temp(C) Hugepages-Usage(page) |
| Chip Device | Bus-Id | AICore(%) Memory-Usage(MB) |
+===============================+=================+======================================================+
| 0 310B1 | OK | 8.2 49 15 / 15 |
| 0 0 | NA | 0 1197 / 11578 |
+===============================+=================+======================================================+
1.2.4 固件版本查看
/var/davinci/driver/upgrade-tool --device_index -1 --component -1 --version
{
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(11).
{"device_id":0, "component":hboot1a, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(12).
{"device_id":0, "component":hboot1b, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(13).
{"device_id":0, "component":hboot2, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(14).
{"device_id":0, "component":ddr, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(16).
{"device_id":0, "component":hsm, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(18).
{"device_id":0, "component":hlink, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(25).
{"device_id":0, "component":Sys_Base_Config, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(30).
{"device_id":0, "component":atf, "version":7.1.0.6.220}
Get component version(7.1.0.6.220) succeed for deviceId(0), componentType(31).
{"device_id":0, "component":Usr_Base_Config, "version":7.1.0.6.220}
}
2 安装问题
//安装的组件版本及过程，问题描述。
2.1 dockre版本：
docker info
Client:
Version: 24.0.6
Context: default
Debug Mode: false
Server:
Containers: 37
Running: 34
Paused: 0
Stopped: 3
Images: 42
Server Version: 24.0.6
Storage Driver: overlay2
Backing Filesystem: extfs
Supports d_type: true
Using metacopy: false
Native Overlay Diff: true
userxattr: false
Logging Driver: json-file
Cgroup Driver: cgroupfs
Cgroup Version: 1
Plugins:
Volume: local
Network: bridge host ipvlan macvlan null overlay
Log: awslogs fluentd gcplogs gelf journald json-file local logentries splunk syslog
Swarm: inactive
Runtimes: ascend io.containerd.runc.v2 runc
Default Runtime: runc
Init Binary: docker-init
containerd version: 7880925980b188f4c97b462f709d0db8e8962aff
runc version: v1.1.9-0-gccaecfc
init version: de40ad0
Security Options:
seccomp
Profile: builtin
Kernel Version: 5.10.0+
Operating System: openEuler 22.03 LTS dracut-055-4.oe2203 (Initramfs) (containerized)
OSType: linux
Architecture: aarch64
CPUs: 4
Total Memory: 11.31GiB
Name: node1
ID: 58f99f10-2560-4039-b3f7-30d1c7ad7ed7
Docker Root Dir: /var/lib/docker
Debug Mode: false
Experimental: false
Insecure Registries:
127.0.0.0/8
Live Restore Enabled: false
Product License: Community Engine
2.2 ascend-docker-runtime版本：
-sh-5.1# /usr/local/Ascend/Ascend-Docker-Runtime/ascend-docker-runtime --version
runc version 1.1.9
commit: v1.1.9-0-gccaecfc
spec: 1.0.2-dev
go: go1.20.7
libseccomp: 2.5.1
-sh-5.1#
2.3 docker配置：
cat /etc/docker/daemon.json
{
"log-opts": {
"max-size": "5m",
"max-file":"3"
},
"data-root": "/var/lib/docker",
"exec-opts": ["native.cgroupdriver=cgroupfs"],
"runtimes": {
"ascend": {
"path": "/usr/local/Ascend/Ascend-Docker-Runtime/ascend-docker-runtime",
"runtimeArgs": []
}
}
}
2.4 由于系统是欧拉，ascend-device-plugin是ubuntu，驱动文件分散到各目录，所以新建了目录：/usr/lib64-ascend将需要的驱动放置到该目录
pwd
/usr/lib64-ascend
-sh-5.1# ls
libaicpu_processer.so libascend_aac_adp.so libascend_aac_sbr_dec.so libascend_drv_baselib.so libascend_protobuf.so.3.13.0.0 libdrvdsmi.so libslog.so libtsdclient.so
libaicpu_prof.so libascend_aac_comm.so libascend_aac_sbr_enc.so libascend_drvdmp.so libc_sec.so libdrvdsmi_efuse.so libtsd_common.so
libaicpu_scheduler.so libascend_aac_dec.so libascend_aicpu_datadump.so libascend_drvupgrade.so libdcmi.so libdrvgpiocfg.so libtsd_eventclient.so
libaicpu_sharder.so libascend_aac_enc.so libascend_camera_mpi_mfusion.so libascend_hal.so libdevmmap.so libmmpa.so libtsdaemon.so
-sh-5.1#
直接使用docker运行ascend-device-plugin镜像(v7.3.1和v26.0.0)
```bash
docker run --rm -it --privileged \
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
-v /usr/lib64-ascend:/usr/lib64-ascend \
-v /dev:/dev \
-v /sys:/sys \
-v /usr/local/sbin/npu-smi:/usr/local/sbin/npu-smi \
-v /tmp:/tmp \
-e LD_LIBRARY_PATH=/usr/lib64-ascend:/usr/local/Ascend/driver/lib64 \
ascend-k8sdeviceplugin:v7.3.1 \
/lib/ld-linux-aarch64.so.1 /usr/local/sbin/npu-smi info
```
报错：`[ERROR] DRV(1,npu-smi):2026-06-02-10:02:55.123.951 [hdc_core.c:396][hdc] [hdcPcieInit 396] HDC init failed, driver may not load.`
目前在k8s集群(1.23.17)中运行ascend-device-plugin和直接用docker形式运行，都是报错：HDC初始化失败