Containers Overview

Accessing Singularity on HPC

Singularity is installed on the operating systems of all HPC compute nodes, so can be easily accessed either from an interactive session or batch script without worrying about software modules. 

Building a Container

Local Builds

Building a container locally requires root authority which users do not have on HPC. This means you must use a Mac or Linux workstation where you have sudo privileges and Singularity installed. The Sylabs website has instructions that can help users get started on building their own containers. Additionally, Nvidia provides an HPC Container Maker which lets you build a recipe without having to know all the syntax. You will just include the blocks you need (e.g., Cuda or Infiniband) and it will create the recipe that you can use for a build on your local workstation.

Remote Builds

To bypass the issue of needing root privileges to build your container, Singularity Hub lets you build and keep containers in the cloud as well as share them with other users. You maintain your recipes there and each time you need to pull one, it gets built remotely and is retrieved to your workstation. This conveniently allows you to build containers directly from HPC. 

As an example, if you want to build a container in your account, first go to https://cloud.sylabs.io, generate an access token (API key), and save it to your clipboard. Next, log in to an interactive terminal session and find your recipe file. In this example, we'll use the recipe:

BootStrap: docker
From: nersc/ubuntu-mpi:14.04

%runscript
    echo "This is what happens when you run the container..."

Then, assuming the recipe is stored in our home directory, we can build it remotely using:

$ singularity remote login   # paste in your API key at the prompt
$ singularity build --remote ~/nersc.sif ~/nersc.recipe

This will produce a .sif file in your home directory that is ready for use.

Singularity, Nvidia, and GPU's

Pulling Images

The general form to pull and convert a NGC image to a local Singularity image file is:

$ singularity build <local_image> docker://nvcr.io/<registry>/<app:tag>

This Singularity build command will download the app:tag NGC Docker image, convert it to Singularity format, and save it to the local filename local_image. For example, to pull the namd NGC container tagged with version 2.12-171025 to a local file named namd.simg saved to your home directory:

$ singularity build ~/namd.simg docker://nvcr.io/hpc/namd:2.12-171025

Running

Directory access:

Singularity containers are themselves ostensibly read only. In order to provide application input and output host directories are generally bound to the container, this is accomplished through the Singularity -B flag. The format of this flag is -B <host_src_dir>:<container_dst_dir>. Once a host directory, host_src_dir, is bound into the container you may interact with this directory from within the container, located at container_dst_dir, the same as you would outside the container.

You may also make use of the --pwd <container_dir> flag, which will be used to set the present working directory of the command to be run within the container.

Ocelote does not support filesystem overlay and as such the container_dst_dir must exist within the image for a bind to be successful. To get around the inability to bind arbitrary directories $HOME and /tmp are mounted in automatically and may be used for application I/O.

GPU support:

All NGC containers are optimized for NVIDIA GPU acceleration so you will always want to add the --nv flag to enable NVIDIA GPU support within the container.

Standard run command:

The Singularity command below represents the canonical form that will be used on the Ocelote cluster.

$ singularity exec --nv --pwd <work_dir> <image.simg> <cmd>   # <work_dir> should be set to either $HOME or /tmp

Containers Available on HPC

We support the use of HPC and ML/DL containers available on NVIDIA GPU Cloud (NGC). Many of the popular HPC applications including NAMD, LAMMPS and GROMACS containers are optimized for performance and available to run in Singularity on Ocelote or Puma. The containers and respective README files can be found in /contrib/singularity/nvidia. 

Tutorials

• The Sylabs GitHub site has files and instructions for creating sample containers.
• Our Github repository has Singularity examples available that can be run on HPC.

Simple Example

The lolcow image is often used as the standard "hello world!" introduction to containers and is described in Singularity's documentation. To follow their example, first start by logging into an interactive terminal session and pull the image:

$  singularity pull docker://godlovedc/lolcow
INFO:    Converting OCI blobs to SIF format
INFO:    Starting build...
Getting image source signatures
[...]
Writing manifest to image destination
Storing signatures
INFO:    Creating SIF file...
INFO:    Build complete: /home/uxx/netid/.singularity/cache/oci-tmp/a692b57abc43035b197b10390ea2c12855d21649f2ea2cc28094d18b93360eeb/lolcow_latest.sif

This will pull the image from Docker Hub and save it in your home in a hidden directory .singularity. Next, run the image simply using singularity run

$ singularity run lolcow_latest.sif
 ______________________________________
/ Perilous to all of us are the devices \
| of an art deeper than we ourselves    |
| possess.                              |
|                                       |
| -- Gandalf the Grey [J.R.R. Tolkien,  |
\ "Lord of the Rings"]                  /
 ---------------------------------------
        \   ^__^
         \  (oo)\_______
            (__)\       )\/\
                ||----w |
                ||     ||

Running Singularity in a Batch Job

Running a job with Singularity is as easy as running other jobs, simply include your resource requests, and include any commands necessary to execute your workflow. For more detailed information on creating and running jobs, see our SLURM documentation or Puma Quick Start. An example script might look like:

#!/bin/bash
#SBATCH --job-name singularity-job
#SBATCH --account=your_pi
#SBATCH --partition=standard
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --time=01:00:00

date
singularity exec --nv dockerTF.img python TFlow_example.py
date

Example Recipe Files

CentOS with Tensorflow

BootStrap: yum
OSVersion: 7
MirrorURL: http://mirror.centos.org/centos-%{OSVERSION}/%{OSVERSION}/os/$basearch/
Include: yum
# best to build up container using kickstart mentality.
# ie, to add more packages to image,
# re-run bootstrap command again.
# bootstrap on existing image will build on top of it, not overwriting it/restarting from scratch
# singularity .def file is like kickstart file
# unix commands can be run, but if there is any error, the bootstrap process ends
%setup
   # commands to be executed on host outside container during bootstrap
%post
     # commands to be executed inside container during bootstrap
     # add python and install some packages
    yum -y install vim wget python3 epel-release
     # install tensorflow
    pip3 install --upgrade pip
    pip3 install tensorflow-gpu==2.0.0-rc1
     # create bind points for storage.
    mkdir /xdisk
    mkdir /groups
    exit 0
# %runscript
   # commands to be executed when the container runs
# %test
   # commands to be executed within container at close of bootstrap process
   python --version

To build and test a container from the recipe from an interactive session on a GPU node:

$ singularity build centosTflow.sif centosTflow.def # Remember, you will need to either build this on a workstation where you have root privileges or will need to user a --remote build
$ singularity exec --nv centosTFlow.simg python3 TFlow_example.py

As a tensorflow example, you could use the following script:

#Linear Regression Example with TensorFlow v2 library 
 
from __future__ import absolute_import, division, print_function
#
import tensorflow as tf
import numpy as np
rng = np.random
#
# Parameters.
learning_rate = 0.01
training_steps = 1000
display_step = 50
#
# Training Data.
X = np.array([3.3,4.4,5.5,6.71,6.93,4.168,9.779,6.182,7.59,2.167,
              7.042,10.791,5.313,7.997,5.654,9.27,3.1])
Y = np.array([1.7,2.76,2.09,3.19,1.694,1.573,3.366,2.596,2.53,1.221,
              2.827,3.465,1.65,2.904,2.42,2.94,1.3])
n_samples = X.shape[0]
#
# Weight and Bias, initialized randomly.
W = tf.Variable(rng.randn(), name="weight")
b = tf.Variable(rng.randn(), name="bias")
 
# Linear regression (Wx + b).
def linear_regression(x):
    return W * x + b
 
# Mean square error.
def mean_square(y_pred, y_true):
    return tf.reduce_sum(tf.pow(y_pred-y_true, 2)) / (2 * n_samples)
 
# Stochastic Gradient Descent Optimizer.
optimizer = tf.optimizers.SGD(learning_rate)
#
# Optimization process. 
def run_optimization():
# Wrap computation inside a GradientTape for automatic differentiation.
    with tf.GradientTape() as g:
        pred = linear_regression(X)
        loss = mean_square(pred, Y)
 
    # Compute gradients.
    gradients = g.gradient(loss, [W, b])    
 
    # Update W and b following gradients.
    optimizer.apply_gradients(zip(gradients, [W, b]))
#
# Run training for the given number of steps.
for step in range(1, training_steps + 1):
    # Run the optimization to update W and b values.
    run_optimization()    
 
    if step % display_step == 0:
        pred = linear_regression(X)
        loss = mean_square(pred, Y)
        print("step: %i, loss: %f, W: %f, b: %f" % (step, loss, W.numpy(), b.numpy()))

MPI

Singularity supports MPI pretty well since, by default, the network is the same inside and outside the container. The more complicated bit is making sure that the container has the right set of MPI libraries. MPI is an open specification, but there are several different implementations (OpenMPI, MVAPICH2, and Intel MPI to name three) with some non-overlapping feature sets. If the host and container are running different MPI implementations, or even different versions of the same implementation, hilarity may ensue. 

The general rule is that you want the version MPI inside the container to be the same version or newer than the host. You may be thinking that this is not good for the portability of your container and you are right. Containerizing MPI applications is not terribly difficult with Singularity, but it comes at the cost of additional requirements for the host system. 

In this example, the infiniband pieces are installed and then the MVAPICH version of MPI. When the job is run, the script will need to load the correct module with the matching version of MVAPICH.

BootStrap: debootstrap
OSVersion: xenial
MirrorURL: http://us.archive.ubuntu.com/ubuntu/


%runscript
    echo "This is what happens when you run the container..."


%post
    echo "Hello from inside the container"
    sed -i 's/$/ universe/' /etc/apt/sources.list
    apt update
    apt -y --allow-unauthenticated install vim build-essential wget gfortran bison libibverbs-dev libibmad-dev libibumad-dev librdmacm-dev libmlx5-dev libmlx4-dev
    wget http://mvapich.cse.ohio-state.edu/download/mvapich/mv2/mvapich2-2.1.tar.gz
    tar xvf mvapich2-2.1.tar.gz
    cd mvapich2-2.1
    ./configure --prefix=/usr/local
    make -j4
    make install
    /usr/local/bin/mpicc examples/hellow.c -o /usr/bin/hellow