Ubuntu 18.04 重灌

下載 iso 檔，利用 rufus 寫入 usb
搜尋工具/Disks 確認硬碟位置

備份資料含檔案屬性

$ sudo cp -a /path/source/. /path/dest

gitea 備份和還原

/var/lib/gitea

$ sudo cp /etc/gitea/app.ini .

$ sudo -u git gitea dump -c /etc/gitea/app.ini

$ unzip gitea-dump-xxxx.zip

$ cd gitea-dump-xxxx

$ mv data/conf/app.ini /etc/gitea/conf/app.ini

$ mv data/* /var/lib/gitea/data/

$ mv log/* /var/lib/gitea/log/

$ mv repos/* /var/lib/gitea/repositories/

$ chown -R gitea:gitea /etc/gitea/conf/app.ini /var/lib/gitea

$ mysql --default-character-set=utf8mb4 -u$USER -p$PASS $DATABASE <gitea-db.sql

$ service gitea restart

mariadb backup and restore

/var/lib/mysql

$ sudo mysql

> SHOW DATABASES;

$ sudo mysqldump --all-databases > all.sql

$ sudo mysql --one-database db_name < all.sql

$ sudo cp -a /opt/tomcat /backup/opt/tomcat

$ sudo cp -a /opt/nvidia /backup/opt/nvidia

$ sudo cp -a /etc/nginx /backup/etc/nginx

$ sudo cp -a /etc/systemd /backup/etc/systemd

$ sudo cp -a /etc/udev/rules.d /backup/udev

$ sudo vi /etc/default/grub

$ sudo update-grub

nvidia driver 更新

$ sudo apt-get install gcc

$ sudo apt-get install make

下載 https://www.nvidia.com/download/driverResults.aspx/168347/en-us 驅動

$ chmod 755 NVIDIA-Linux-x86_64-460.32.03.run

$ sudo ./ NVIDIA-Linux-x86_64-460.32.03.run

但會說 nvidia-drm 已經啟動

所以要卸載 舊驅動

$ sudo apt-get --purge remove "*cublas*" "*cufft*" "*curand*" \

 "*cusolver*" "*cusparse*" "*npp*" "*nvjpeg*" "cuda*" "nsight*"

$ sudo apt-get purge 'nvidia*'

$ sudo apt-get autoremove

$ sudo reboot

$ sudo ./ NVIDIA-Linux-x86_64-460.32.03.run

使用 --no-opengl-files 參數，以免使用DeepStream時，只開啟一下黑幕，出現下列錯誤

cuGraphicsGLRegisterBuffer failed with error(304) gst_eglglessink_cuda_init texture = 1

$ sudo ./ NVIDIA-Linux-x86_64-460.32.03.run --no-opengl-files --dkms --no-drm

因為 ubuntu 還是用了一個驅動，所以不能直接更新驅動

但可以設定停用此驅動，小心回答問題，在執行一遍

Would you like ton register the kernel module sources with DKMS? 回答 Yes

Install NVIDIA’s 32-bit compatibility libraries? 回答 No

$ nvidia-smi

安裝 CUDA, 不要使用 deb(network), 好像在版本安裝上會出錯

$ wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-ubuntu1804.pin

$ sudo mv cuda-ubuntu1804.pin /etc/apt/preferences.d/cuda-repository-pin-600

$ wget https://developer.download.nvidia.com/compute/cuda/11.1.1/local_installers/cuda-repo-ubuntu1804-11-1-local_11.1.1-455.32.00-1_amd64.deb

$ sudo dpkg -i cuda-repo-ubuntu1804-11-1-local_11.1.1-455.32.00-1_amd64.deb

$ sudo apt-key add /var/cuda-repo-ubuntu1804-11-1-local/7fa2af80.pub

$ sudo apt-get update

$ sudo apt-get -y install cuda

$ sudo apt-get -y install cuda-11-1

指定版本很重要，不然會裝最新的

$ nvidia-smi

Failed to initialize NVML: Driver/library version mismatch

重新開機即可

但顯示出來的 Driver 和 CUDA Version 都會改變

TensorRT 安裝

https://developer.nvidia.com/nvidia-tensorrt-download

由此進入選擇所需版本，並選擇 deb 版

由此選擇開啟文件

進入 TensorRT Installation Guide

https://docs.nvidia.com/deeplearning/tensorrt/archives/tensorrt-723/install-guide/index.html

跳至 4.1. Debian Installation

跟著步驟安裝

設定環境變數

PATH

LD_LIBRARY_PATH

pycuda 只能搭配 python 3.7

$ sudo apt-get install python3.7

$ sudo apt-get install python3.7-dev

$ python3.7 -m pip install 'pycuda>=2019.1.1'

Install Nvidia Docker

$ curl https://get.docker.com | sh && sudo systemctl --now enable docker

$ distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \

   && curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add - \

   && curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

$ curl -s -L https://nvidia.github.io/nvidia-container-runtime/experimental/$distribution/nvidia-container-runtime.list | sudo tee /etc/apt/sources.list.d/nvidia-container-runtime.list

$ sudo apt-get update

$ sudo apt-get install -y nvidia-docker2

$ sudo systemctl restart docker

$ sudo docker run --rm --gpus all nvidia/cuda:11.0-base nvidia-smi

$ sudo groupadd docker

$ sudo usermod -a -G docker $USER

$ sudo reboot

Install TensorRT 7.2 OSS

參考 https://github.com/NVIDIA/TensorRT/tree/master

Install TensorRT 7.2 OSS

$ git clone -b master https://github.com/nvidia/TensorRT TensorRT_OSS-7.2.3.4

$ cd TensorRT_OSS-7.2.3.4/

$ git submodule update --init --recursive

$ cd /pathto/TensorRT-7.2.2.3/

$ export TRT_LIBPATH=`pwd`

$ cd /pathto/TensorRT_OSS-7.2.3.4/

$ ./docker/build.sh --file docker/ubuntu-18.04.Dockerfile --tag tensorrt-ubuntu-1804 --cuda 11.1

$ ./docker/launch.sh --tag tensorrt-ubuntu-1804 --gpus all

trtuser@c2936e108d43:/workspace$ cd $TRT_OSSPATH

trtuser@c2936e108d43:/workspace/TensorRT$ mkdir -p build && cd build

trtuser@c2936e108d43:/workspace/TensorRT/build$ cmake .. -DTRT_LIB_DIR=$TRT_LIBPATH -DTRT_OUT_DIR=`pwd`/out

trtuser@c2936e108d43:/workspace/TensorRT/build$ make -j$(nproc)

trtuser@c2936e108d43:/workspace/TensorRT/build$ exit

$ cd ..

$ mkdir backup

$ mv TensorRT-7.2.2.3/targets/x86_64-linux-gnu/lib/libnvinfer_plugin.so.7.2.2 backup

$ cp TensorRT_OSS-7.2.3.4/build/out/libnvinfer_plugin.so.7.2.3 TensorRT-7.2.2.3/targets/x86_64-linux-gnu/lib/

install xpra

https://github.com/Xpra-org/xpra/blob/master/docs/Build/Debian.md

$ git clone https://github.com/Xpra-org/xpra.git

$ sudo ./setup.py install

Exception: ERROR: cannot find a valid pkg-config entry for nvjpeg-11.4 using PKG_CONFIG_PATH=(empty)

$ vi setup.py

/if nvjpeg_ENABLED:

在底下不遠處有兩處 for v in ("11.4", "11.3"...):

皆改成 for v in ("11.1"): 即可

另外 ld: connot find -lcuda

$ sudo ln -s /usr/local/cuda-11.1/lib64/stubs/libcuda.so /usr/lib

pip 更新出錯

$ pip install --upgrade --no-cache-dir pip

$ python3 -m pip install --upgrade --no-cache-dir pip -i https://pypi.python.org/simple