[问答]

GRID 3.0使用M60 GPU在ESXI 6.0.2上成功安装但无法通过nvidia-smi进行验证

问答对人有帮助，内容完整，我也想知道答案 0 我们有一个干净的SuperMicro服务器并安装了VMWare ESXI 6.0.2 build 360759.进入“维护模式”，然后按照步骤安装最新指南文档中的NVDIA主机驱动程序，重新启动并关闭“维护模式”。一旦它恢复，我们通过各种命令进行验证，它们都验证了，除非使用nvidia-smi命令返回：无法初始化NVML：未知错误注意：相同的硬件在GRID 2.0中正常工作硬件/软件清单： Supermicro Chassis 1028GQ-TRT 双至强E5-2600v3 2.60 256 gib内存（4）安装了NVIDIA M-60卡并处于图形模式（2）480 gig SSD硬盘 ESXI 6.0.2 build 360759 NVIDIA-的vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585.vib 安装步骤： •在干净的系统上安装ESXI 6 •启用SSH •尚未设置虚拟机或数据存储区 •通过vSphere在服务器上设置时钟：ntp.org.pool •在Configuration / Software / Advanced Settings / VMkernel / Boot下检查：“VMkernel.Boot.disableACSCheck “然后点击”确定“ •进入维护模式 •从此链接下的最新产品版本下的NVIDIA许可中心下载网格软件： •https://nvidia.flexnetoperations.com/control/nvda/viewRecentProductReleases •抓住4月4日发布的适用于vSphere 6.0的Grid 3.0 •复制：NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585.vib到tmp文件夹 •SSH进入服务器 •Ran：esxcli软件振动安装-v /tmp/NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585.vib 结果：安装结果消息：操作成功完成。需要重新启动：false 已安装VIB：NVIDIA_bootbank_NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585 VIB已删除： VIB跳过：？重新启动服务器并关闭维护模式？ SSH进入服务器并验证安装验证结果： [root @ localhost：〜] esxcli软件振动列表\| grep -i nvidia NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver 361.40-1OEM.600.0.0.2494585 NVIDIA VMwareAccepted 2016-05-04 [root @ localhost：〜] vmkload_mod -l \| grep nvidia nvidia 0 10012 [root @ localhost：〜] esxcfg-module -l \| grep nvidia nvidia 0 10012 [root @ localhost：〜] nvidia-smi 无法初始化NVML：未知错误在ESXI上没有passthru的设置它只是一个没有虚拟机或数据存储的干净服务器如果我们继续并继续尝试创建VM并通过vCenter设置添加M60卡，则不会列出任何配置文件。我们已经在另外两台具有不同CPU的服务器上进行了此操作，但是相同的软件和M60卡已经多次完成相同的结果。请帮忙。谢谢！亚历克斯以上来自于谷歌翻译以下为原文 We have a clean SuperMicro server and installed VMWare ESXI 6.0.2 build 360759. Entered "Maintenance Mode", then followed the steps and installed the NVDIA Host Driver from the latest guide docs, rebooted and turned off "Maintenance Mode". Once it came back up and we ssh'd in to verify it via various commands they all verified except when using the nvidia-smi command which returns: Failed to initialize NVML: Unknown Error NOTE: The same hardware worked properly in GRID 2.0 Hardware/Software list: Supermicro Chassis 1028GQ-TRT Dual Xeon E5-2600v3 2.60 256 gib memory (4) NVIDIA M-60 cards installed and in graphics mode (2) 480 gig SSD drives ESXI 6.0.2 build 360759 NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585.vib Steps taken to instal: •Installed ESXI 6 on a clean system •Enabled SSH •No vm’s or Datastores setup yet •Setup clock on server: ntp.org.pool via vSphere •Checked off under Configuration/Software/Advanced Settings/VMkernel/Boot: ”VMkernel.Boot.disableACSCheck " and clicked "OK" •Entered Maintenance Mode •Downloaded Grid software from NVIDIA License center under Recent Product Releases from this link: •https://nvidia.flexnetoperations.com/control/nvda/viewRecentProductReleases •Grabbed the April 4th release of Grid 3.0 for vSphere 6.0 •Copied: NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585.vib to tmp folder •SSH’d into server •Ran: esxcli software vib install -v /tmp/NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585.vib Result: Installation Result Message: Operation finished successfully. Reboot Required: false VIBs Installed: NVIDIA_bootbank_NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver_361.40-1OEM.600.0.0.2494585 VIBs Removed: VIBs Skipped: ?Rebooted Server and turned off Maintenance mode ?SSH'd into server and verified install Verify Results: [root@localhost:~] esxcli software vib list \| grep -i nvidia NVIDIA-vGPU-VMware_ESXi_6.0_Host_Driver 361.40-1OEM.600.0.0.2494585 NVIDIA VMwareAccepted 2016-05-04 [root@localhost:~] vmkload_mod -l \| grep nvidia nvidia 0 10012 [root@localhost:~]esxcfg-module -l \| grep nvidia nvidia 0 10012 [root@localhost:~] nvidia-smi Failed to initialize NVML: Unknown Error No passthru's setup on ESXI it's just a clean server with no vm's or datastores If we continue and go forward try to create VM's and setup via vCenter to add the M60 card, none of the profiles are listed. We've gone through this on 2 other servers with different CPU's but the same software and M60 cards several times with the exact same results. Please help. Thanks! Alex 0
2018-9-27 16:11:14　　评论淘帖0 邀请回答您可以邀请以下用户，快速回答问题 × 60user188 该类别下有 8 个回答。邀请回答 wo4456 该类别下有 7 个回答。邀请回答 jingfu888 该类别下有 7 个回答。邀请回答 muuwfwr 该类别下有 6 个回答。邀请回答 myf888 该类别下有 5 个回答。邀请回答 zsqzsqzs 该类别下有 5 个回答。邀请回答 cmlzm 该类别下有 5 个回答。邀请回答鑫12345 该类别下有 5 个回答。邀请回答 nnmnnm 该类别下有 5 个回答。邀请回答 60user30 该类别下有 4 个回答。邀请回答 www036 该类别下有 4 个回答。邀请回答 mao5091 该类别下有 4 个回答。邀请回答 ey8616 该类别下有 4 个回答。邀请回答 testd021 该类别下有 3 个回答。邀请回答邓刚1234 该类别下有 3 个回答。邀请回答 zmyecho 该类别下有 3 个回答。邀请回答 uwjfjsdfwer 该类别下有 3 个回答。邀请回答 zjjcn 该类别下有 3 个回答。邀请回答 h1654155275.6441 该类别下有 3 个回答。邀请回答 zzy0407 该类别下有 3 个回答。邀请回答举报张博相关推荐 • 计算模式禁止GRID M60 3533 • 特斯拉M60，VMware vSphere 6.5U1，GPU在设备管理器中丢失 2865 • NVIDIA-SMI：监控GPU的绝佳起点 2751 • 计算M.禁止在特斯拉vGRID M60-2Q 6493 • ESXi 6.5 + Tesla M60无法启动大众 5010 • Nvidia Tesla M60 PassThought不能使用 2722 • 查询vGPU和nvidia-smi命令什么是C+G型 10702 • NVIDIA Vmware vSphere-6.5升级无法启动GPU 4963 • Tesla M60能直通ESXi 6.5 VM 5327 • nvidia-smi在Cisco b200-m4上返回失败的结果 2875 20个回答

答案对人有帮助，有参考价值 0 从这个案例中学到的总结： •与虚拟机管理程序一起使用时，服务器上的BIOS设置不正确可能导致MMIO地址问题导致无法识别GRID GPU。 Øhttp://nvidia.custhelp.com/app/answers/detail/a_id/4119/ 以上来自于谷歌翻译以下为原文 Summary of what has been learned from this case: •Incorrect BIOS settings on a server when used with a hypervisor can cause MMIO address issues that result in GRID GPUs failing to be recognized. ohttp://nvidia.custhelp.com/app/answers/detail/a_id/4119/

2018-9-27 20:53:54 评论举报邓罕攀