Difference between revisions of "Deploy a container using Singularity"

Revision as of 09:20, 10 June 2021

Create a Docker file and build a docker image on your local computer

• Install Docker
• Create a folder

$ mkdir test_container
$ cd test_container

• Open file named **Dockerfile** with a text editor

FROM ubuntu:latest

WORKDIR /home/project

RUN apt-get -y update
RUN apt-get install -y lsb-release python3 python3-pip \
&& pip install numpy==1.19 matplotlib tensorflow

• Build an image

$ docker build -t username/ubuntu-python3:latest .
Sending build context to Docker daemon  3.072kB
Step 1/4 : FROM ubuntu:latest
 ---> 7e0aa2d69a15
Step 2/4 : WORKDIR /home/project
 ---> Using cache
 ---> d9947182f6a8
Step 3/4 : RUN apt-get -y update
 ---> Using cache
 ---> 89a634ac491b
Step 4/4 : RUN apt-get install -y lsb-release python3 python3-pip && pip install numpy==1.19 matplotlib tensorflow
 ---> Using cache
 ---> d639b6f22ee5
Successfully built d639b6f22ee5
Successfully tagged username/ubuntu-python3:latest

$ docker images
 REPOSITORY                 TAG       IMAGE ID       CREATED          SIZE
 username/ubuntu-python3     latest    d639b6f22ee5   18 minutes ago   2.19GB

Deploy the image on the cluster

Method 1: Use the Docker image on Dockerhub

• Push the image to Dockerhub

[user@local]$  docker push dockeruser/ubuntu-python3:latest
The push refers to repository [docker.io/dockeruser/ubuntu-python3]
4648e930d81f: Pushed 
2e38cc729d73: Pushed 
f281ab5d2fac: Pushed 
2f140462f3bc: Mounted from library/ubuntu 
63c99163f472: Mounted from library/ubuntu 
ccdbb80308cc: Mounted from library/ubuntu 
latest: digest: sha256:cd8a34b30aabe432232787d1e93844cd01027c2b235fb88f106297ed26c1f2ca size: 1576

• Log into Euler

[user@local]$ ssh nethzuser@euler.ethz.ch

• To use Singularity, your NETHZ username has to be added to ID-HPC-SINGULARITY group

• Request a compute node with Singularity

[nethzuser@eu-login-01 ~]$ bsub -n 1 -R singularity -R light -Is bash
Generic job.
Job <174809810> is submitted to queue <light.5d>.
<<Waiting for dispatch ...>>
<<Starting on eu-ms-001-02>>
[nethzuser@eu-ms-001-02 ~]$

• Pull the Docker image with Singularity

[nethzuser@eu-ms-001-02 ~]$ cd $SCRATCH
[nethzuser@eu-ms-001-02 nethzuser]$ singularity pull docker://dockeruser/ubuntu-python3
INFO:    Converting OCI blobs to SIF format
INFO:    Starting build...
Getting image source signatures
Copying blob 345e3491a907 done  
...
INFO:    Creating SIF file...

[nethzuser@eu-ms-001-02 nethzuser]$ ll
ubuntu-python3_latest.sif

• Run the container as shell

[nethzuser@eu-ms-001-02 nethzuser]$ singularity shell ubuntu-python3_latest.sif
Singularity> 
Singularity> lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 20.04.2 LTS
Release:	20.04
Codename:	focal

Singularity> python3
Python 3.8.5 (default, May 27 2021, 13:30:53) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
2021-06-09 14:08:24.919069: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library  'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /.singularity.d/libs
2021-06-09 14:08:24.919094: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not  have a GPU set up on your machine.
>>> tf.__version__
'2.5.0'
>>> 
Singularity> exit
exit

As you can see, the container contains GCC 9.3.0 with Python 3.8.5 and TensorFlow 2.5.0. At the time of writing tutorial, the latest version of GCC is 8.2.0 and TensorFlow 2.4.0

[nethzuser@eu-ms-001-02 nethzuser]$ module avail gcc
--------------------------- /cluster/apps/lmodules/Core --------------------------------
  gcc/4.8.2    gcc/4.8.5 (L)    gcc/5.4.0    gcc/6.3.0    gcc/7.3.0    gcc/8.2.0 (D)

[nethzuser@eu-ms-001-02 nethzuser]$ module load gcc/6.3.0 python/3.8.5
[nethzuser@eu-ms-001-02 nethzuser]$ python
Python 3.8.5 (default, Oct  6 2020, 10:04:29) 
[GCC 6.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
2021-06-09 14:56:47.778183: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library  'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /cluster/apps/gcc-6.3.0/openblas-0.2.20-cot3cawsqf4pkxjwzjexaykbwn2ch3ii/lib:/cluster/apps/nss/gcc-6.3.0/python/3.8.5/x86_64/lib64:/cluster/spack/apps/linux-centos7-x86_64/gcc-4.8.5/gcc-6.3.0-sqhtfh32p5gerbkvi5hih7cfvcpmewvj/lib64:/cluster/spack/apps/linux-centos7-x86_64/gcc-4.8.5/gcc-6.3.0-sqhtfh32p5gerbkvi5hih7cfvcpmewvj/lib:/cluster/apps/lsf/10.1/linux2.6-glibc2.3-x86_64/lib
2021-06-09 14:56:47.778214: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not  have a GPU set up on your machine.
>>> tf.__version__
'2.4.0'
>>> 

Method 2: Convert a Docker image to a Singularity image

• Create a tarball of the Docker image

[user@local]$ docker save d639b6f22ee5 -o ubuntu-python3.tar
[user@local]$ ls
ubuntu-python3.tar

• Copy the tarball to $SCRATCH on Euler

[user@local]$ scp ubuntu-python3.tar netzuser@euler.ethz.ch:/cluster/scratch/nethzuser

• Log in to Euler

[user@local]$ ssh netzuser@euler.ethz.ch
[nethzuser@eu-login-01 ~]$ cd $SCRATCH
[nethzuser@eu-login-01 nethzuser]$ pwd
/cluster/scratch/nethzuser

[nethzuser@eu-login-01 nethzuser]$ pwd
/cluster/scratch/nethzuser

• Convert the tarbal to a Singularity image

[nethzuser@eu-login-01 nethzuser]$ singularity build --sandbox ubuntu-python3 docker-archive://ubuntu-python3.tar
INFO:    Starting build...
...
INFO:    Creating sandbox directory...
INFO:    Build complete: ubuntu-python3

• Run the Singularity box

[nethzuser@eu-login-01 nethzuser]$ singularity shell ubuntu-python3
Singularity> 
Singularity> gcc --version
gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Singularity> python3 --version
Python 3.8.5
ingularity> python3 -c "import tensorflow as tf; print(tf.__version__)"
2021-06-10 09:12:44.533833: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /.singularity.d/libs
2021-06-10 09:12:44.533867: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2.5.0
Singularity> exit
exit