有一个nfs盘挂载到本机，本机docker想用这个nfs盘存储。





此时docker自动用起了devicemapper 。

docker 用 devicemapper 这个后备存储有个默认卷的限制，是10g。现在都用overlay2了。

##  dd

解决，首先在nfs盘里面dd出一个500G大文件系统出来：
```
 dd if=/dev/zero of=/home/kevin_xie/yifeinfs/docker.data bs=1G count=500
```

```
dd 用来复制、抽取、转换文件，这里复制、抽取的大小可以以具体的块数指定，同时复制的时候可以进行编码或格式转换；
dd 的主要选项：
if=file                 输入文件名，这里的文件不能是目录；
of=file                输出文件名，这里的文件不能是目录；    
ibs=bytes         读入多少个字节，在这里就是指定块的大小，即输入缓冲区 的大小；
obs=bytes        一 次读入多少个字节，在这里就是指定块的大小，即输出缓冲区的大小；
bs=bytes          同时设置读写块的大小，可代替 ibs 和 obs；
cbs=bytes         一次转换多少个个字节，即转换缓冲区大；
skip=blocks      从输入文件开头跳过多少个块后再开始复制，也就是指定都从哪里开始复制；
seek=blocks     从输出文件开头跳过多少个块后再开始复制，也就是指定目的文件要空多少个块才开始写入(通常只有当输出文件是磁盘或磁带时才有效)
count=blocks     拷贝输入文件的多少个块；
conv=conversion[,conversion...]  指定格式转换的参数
ascii                转换 EBCDIC 为 ASCII；
ebcdic             转换 ASCII 为 EBCDIC；
ibm                  转换 ASCII 为 alternate EBCDIC；
block              把每一行转换为长度为 cbs 的记录，不足部分用空格填充；
unblock          使每一行的长度都为 cbs ，不足部分用空格填充；     
lcase               把大写字符转换为小写字符；
ucase              把小写字符转换为大写字符；
noerror            出错时不停止；
notrunc            不截断输出文件；
sync 把每个块都写满，数据不足的部分用空字符填充；
```

##  格式化dd出来的大文件

把大文件格式化为xfs文件系统：
```
mkfs.xfs docker.data
```


不小心弄错了就用这个指令强制覆盖成xfs文件系统：
```
 mkfs -t xfs -f docker.data
```

-t 　　 指定文件系统类型（如ext2，ext3，ext4等等），则会从/etc/mke2fs.conf文件中读取默认配置

-f 　　 强制覆盖


##  挂载

然后把这个大文件系统挂载到/var/lib/docker/下面去，之后在/var/lib/docker/产生的文件就会实际存储在docker.data里面去。
```
mount /home/kevin_xie/yifeinfs/docker.data /var/lib/docker/
```

![在这里插入图片描述](/static/img/images/00181.png)

Docker no space left on device 10G限制 nfs块 devicemapper

docker run 的时候增加参数 `--storage-opt size=30G`




```
docker run -it --gpus all --storage-opt size=30G kevinchina/deeplearning:cuda10.1pytorch1.7.1baseenv0.1 bash
```

> 设置容器磁盘大小 在 docker 启动时候 加载参数 --storage-opt dm.basesize=100G 用来调整默认容器的

> rootfs大小； 在创建容器的时候使用参数--storage-opt size=30G来设置改容器的 rootfs 大小。

> 

> 作者：momo1023 链接：https://www.jianshu.com/p/d984723d9b31 来源：简书

> 著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。

docker 容器 磁盘 10G 限制 大小

# cpu





miniconda3的一个基础镜像，其中安装fastapi和opencv。

kevinchina/deeplearning:miniconda3base0.2
```shell
FROM kevinchina/deeplearning:wrfbaseenv
COPY deps /deps
RUN rm -rf /etc/yum.repos.d/* && \
    cp /deps/CentOS-Base.repo /etc/yum.repos.d/ && \
    yum clean all
RUN yum install wget opencv -y
RUN wget http://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda/Miniconda3-latest-Linux-x86_64.sh -O /tmp/conda.sh && \
    /bin/bash /tmp/conda.sh -b && \
    cp -f /deps/.condarc ~/ && rm /tmp/conda.sh
ENV PATH ~/miniconda3/bin:$PATH
RUN ~/miniconda3/bin/conda init bash && \
    . ~/.bashrc && \
    conda create -n py38 python=3.8.12 -y
RUN . ~/.bashrc && \
    conda activate py38 && \
    pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U && \
    pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple && \
    pip install fastapi uvicorn python-multipart requests opencv-python numpy
RUN cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
```

在上一个基础上，安装pytorch cpu版本 1.8

kevinchina/deeplearning:miniconda3base0.3
```shell
FROM kevinchina/deeplearning:miniconda3base0.2
RUN . ~/.bashrc && conda activate py38 && conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cpuonly -c pytorch
```

在上一个基础上，安装 yolo需求的包， 创建app文件夹。

成为 ： kevinchina/deeplearning:miniconda3base0.6
```
FROM kevinchina/deeplearning:miniconda3base0.3
RUN mkdir /app
WORKDIR /app
COPY requirements.txt .
COPY deps/Arial.ttf /root/.config/Ultralytics/Arial.ttf
RUN . ~/.bashrc && conda activate py38 && pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U
RUN . ~/.bashrc && conda activate py38 && pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
RUN . ~/.bashrc && conda activate py38 && pip install -r requirements.txt
```

```
matplotlib>=3.2.2
Pillow
PyYAML==5.3.1
scipy>=1.4.1
tqdm>=4.41.0
seaborn>=0.11.0
pandas
Shapely==1.8.0
numpy==1.19.2
tensorboard
```



# selenium

在kevinchina/deeplearning:miniconda3base0.2基础上装selenium

kevinchina/deeplearning:miniconda3base0.5
```shell
FROM kevinchina/deeplearning:miniconda3base0.2
RUN . ~/.bashrc && \
    conda activate py38 && \
    pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple && \
    pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U && \
    pip install  pandas requests selenium
```



# GPU CUDA11.1  pytorch1.8

cuda11.1+pytorch1.8+opencv

kevinchina/deeplearning:cuda11.1pytorch1.8baseenv0.1
```shell
FROM kevinchina/deeplearning:cuda11.1pytorch1.8baseenv
RUN cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
```

下面是kevinchina/deeplearning:cuda11.1pytorch1.8baseenv0.5：

将目前YOLOv5需要的环境安装了起来。
```
FROM kevinchina/deeplearning:cuda11.1pytorch1.8baseenv0.1
RUN mkdir /app
WORKDIR /app
COPY requirements.txt .
RUN . ~/.bashrc && conda activate py38 && pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U
RUN . ~/.bashrc && conda activate py38 && pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
COPY deps/Arial.ttf /root/.config/Ultralytics/Arial.ttf
RUN . ~/.bashrc && conda activate py38 && pip install numpy==1.19.2 wandb  tensorboard -i https://pypi.tuna.tsinghua.edu.cn/simple
```

```
matplotlib>=3.2.2
Pillow
PyYAML==5.3.1
scipy>=1.4.1
tqdm>=4.41.0
seaborn>=0.11.0
pandas
Shapely==1.8.0
numpy==1.19.2
tensorboard
```

# GPU CUDA10.2 pytorch1.8

cuda10.2+pytorch1.8+opencv

kevinchina/deeplearning:cuda10.2pytorch1.8baseenv0.1
```shell
FROM nvidia/cuda:10.2-cudnn7-devel-centos7
COPY deps /deps
RUN rm -rf /etc/yum.repos.d/* && cp /deps/CentOS-Base.repo /etc/yum.repos.d/ && yum clean all
RUN yum install wget opencv vim -y
RUN wget http://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda/Miniconda3-latest-Linux-x86_64.sh -O /tmp/conda.sh && \
    bash /tmp/conda.sh -b && rm /tmp/conda.sh && \
    cp -f /deps/.condarc /root/
RUN ~/miniconda3/bin/conda init bash && . ~/.bashrc && conda create -n py38 python=3.8.12 -y
RUN . ~/.bashrc && conda activate py38 && pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U && pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple && pip install fastapi uvicorn python-multipart requests opencv-python numpy
RUN . ~/.bashrc && conda activate py38 && conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=10.2 -c pytorch -y
RUN cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
```

# GPU CUDA10.2   pytorch1.6.0

```shell
FROM nvidia/cuda:10.2-cudnn7-devel-centos7
COPY deps /deps
RUN rm -rf /etc/yum.repos.d/* && cp /deps/CentOS-Base.repo /etc/yum.repos.d/ && yum clean all
RUN yum install wget opencv vim -y
RUN wget http://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda/Miniconda3-latest-Linux-x86_64.sh -O /tmp/conda.sh && \
    bash /tmp/conda.sh -b && rm /tmp/conda.sh && \
    cp -f /deps/.condarc /root/
RUN ~/miniconda3/bin/conda init bash && . ~/.bashrc && conda create -n py38 python=3.8.12 -y
RUN . ~/.bashrc && conda activate py38 && pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U && pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple && pip install fastapi uvicorn python-multipart requests opencv-python numpy
RUN . ~/.bashrc && conda activate py38 && conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.2 -c pytorch
RUN cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
```


# GPU CUDA10.1 pytorch1.7.1

cuda10.1+pytorch1.7.1+opencv

kevinchina/deeplearning:cuda10.1pytorch1.7.1baseenv0.1

```shell
FROM nvidia/cuda:10.1-cudnn7-devel-centos7
COPY deps /deps
RUN rm -rf /etc/yum.repos.d/* && cp /deps/CentOS-Base.repo /etc/yum.repos.d/ && yum clean all
RUN yum install wget opencv vim -y
RUN wget http://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda/Miniconda3-latest-Linux-x86_64.sh -O /tmp/conda.sh && \
    bash /tmp/conda.sh -b && rm /tmp/conda.sh && \
    cp -f /deps/.condarc /root/
RUN ~/miniconda3/bin/conda init bash && . ~/.bashrc && conda create -n py38 python=3.8.12 -y
RUN . ~/.bashrc && conda activate py38 && pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U && pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple && pip install fastapi uvicorn python-multipart requests opencv-python numpy
RUN . ~/.bashrc && conda activate py38 && conda install pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=10.1 -c pytorch -y
RUN cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
```

加一些库，加个VGG16的权重进去用。

kevinchina/deeplearning:cuda10.1pytorch1.7.1baseenv0.2
```
FROM kevinchina/deeplearning:cuda10.1pytorch1.7.1baseenv0.1
RUN . ~/.bashrc && conda activate py38 && pip install pymilvus pandas pillow clickhouse_driver opencv-python pillow requests
COPY vgg16-397923af.pth /root/.cache/torch/hub/checkpoints/
```

# YOLO

装了一些yolo的东西

kevinchina/deeplearning:cuda10.2torch1.8yolov0.1
```python
FROM kevinchina/deeplearning:cuda10.2pytorch1.8baseenv0.1
COPY requirements.txt .
COPY Arial.ttf /root/.config/Ultralytics/
RUN . ~/.bashrc && conda activate py38 && pip install -r requirements.txt
#pip install -r requirements.txt
# Base ----------------------------------------
matplotlib>=3.2.2
numpy>=1.18.5
Pillow
PyYAML==5.3.1
requests>=2.23.0
scipy>=1.4.1
tqdm>=4.41.0
pandas
requests
fastapi
uvicorn
python-multipart
seaborn>=0.11.0
pandas
Shapely==1.8.0
opencv-python
```

# paddle OCR 

docker push kevinchina/deeplearning:paddleocr0.1
```
FROM paddlepaddle/paddle:2.2.1
RUN rm -f /etc/localtime
RUN cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
RUN pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip -U
RUN pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
RUN pip install fastapi uvicorn python-multipart requests opencv-python numpy
RUN unset GREP_OPTIONS
COPY . /PaddleOCRAI
WORKDIR /PaddleOCRAI
RUN pip install -i https://pypi.tuna.tsinghua.edu.cn/simple pip install -r requirements.txt
RUN python setup.py install
#ENTRYPOINT nohup python ocr_fastapi_only_one.py >/ocr.log 2>&1 &
```

dockerfile 常用的环境编写 miniconda3 dockerfile python

dockerfile





COPY hl.tar.gz /wrf_env/

得到

```shell
(base) [root@fce482f7fc63 wrf_env]# ll
total 4
-rw-r--r--. 1 root root 143 Dec  9 03:35 hl.tar.gz
```

ADD hl.tar.gz /wrf_env1/

得到

```shell
(base) [root@fce482f7fc63 wrf_env1]# ll
total 0
drwxr-xr-x. 2 root root 24 Dec  9 03:32 test
(base) [root@fce482f7fc63 wrf_env1]# pwd
/wrf_env1
```

dockerfile ADD 和 COPY

centOS镜像的话，dockerfile 加一句：




```
RUN cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
```