Singularity on the cluster
Contents
Table of Contents:
- Introduction
- Docker vs Singularity
- Singularity on the cluster
- Getting existing images onto the cluster
- Interacting with images
- Singularity containers as SLURM jobs
- Building Singularity images
- References
Introduction
Containerization of workloads has become popular, particularly using Docker. However, Docker is not suitable for HPC applications due to security reasons. There are a couple of alternatives for HPC containers, with Singularity being the one that covers a large set of cases. Singularity has been deployed on the cluster, and can also import Docker containers.
This page provides information on how to use Singularity on the cluster. Singularity enables users to have full control of their operating system environment. This allows a non-privileged user to “swap out” the Linux operating system and environment on the host machine for a Linux OS and computing environment that they can control. For instance, if the host system runs CentOS Linux but your application requires Ubuntu Linux with a specific software stack, you can create an Ubuntu image, install your software into that image, copy the created image to the cluster, and run your application on that host in its native Ubuntu environment.
Singularity leverages the resources of the host system, such as high-speed interconnect (e.g., InfiniBand), high-performance parallel file systems (e.g., Lustre /n/holyscratch01 and /n/holylfs filesystems), GPUs, and other resources (e.g., licensed Intel compilers).
Note for Windows and MacOS: Singularity only supports Linux containers. You cannot create images that use Windows or MacOS (this is a restriction of the containerization model rather than Singularity).
Why Singularity?
There are some important differences between Docker and Singularity:
- Docker and Singularity have their own container formats.
- Docker containers may be imported to run via Singularity.
- Docker containers need root privileges for full functionality which is not suitable for a shared HPC environment.
- Singularity allows working with containers as a regular user.
Singularity on the cluster
Singularity is available only on the compute nodes on the cluster. Therefore, to use it you need to either start an interactive job or submit a batch-job to the available SLURM queues.
In the below examples we illustrate the interactive use of Singularity in an interactive bash shell.
[user@holylogin01 ~]$ salloc -p test -c 1 -t 00-01:00 --mem=4000 [user@holyseas02 ~]$
Check Singularity version:
[user@holyseas02 ~]$ which singularity /usr/bin/singularity [user@holyseas02 ~]$ singularity --version singularity version 3.1.1-1.el7
The most up-to-date help on Singularity comes from the command itself.
[user@holyseas02 ~]$ singularity --help
Linux container platform optimized for High Performance Computing (HPC) and
Enterprise Performance Computing (EPC)
Usage:
singularity [global options...]
Description:
Singularity containers provide an application virtualization layer enabling
mobility of compute via both application and environment portability. With
Singularity one is capable of building a root file system that runs on any
other Linux system where Singularity is installed.
Options:
-d, --debug print debugging information (highest verbosity)
-h, --help help for singularity
-q, --quiet suppress normal output
-s, --silent only print errors
-t, --tokenfile string path to the file holding your sylabs
authentication token (default
"/n/home12/nweeks/.singularity/sylabs-token")
-v, --verbose print additional information
--version version for singularity
Available Commands:
apps List available apps within a container
build Build a Singularity image
cache Manage the local cache
capability Manage Linux capabilities for users and groups
exec Run a command within a container
help Help about any command
inspect Show metadata for an image
instance Manage containers running as services
key Manage OpenPGP keys
oci Manage OCI containers
pull Pull an image from a URI
push Push a container to a Library URI
run Run the user-defined default command within a container
run-help Show the user-defined help for an image
search Search a Library for images
shell Run a shell within a container
sign Attach a cryptographic signature to an image
test Run the user-defined tests within a container
verify Verify cryptographic signatures attached to an image
version Show the version for Singularity
Examples:
$ singularity help
Getting existing images onto the cluster
Singularity uses container images which you can scp or rsync to the cluster as you would do with any other file. See Copying Data to & from the cluster using SCP or SFTP for more information.
Note: For larger Singularity images, please use the available scratch filesystems, such as /n/holyscratch01/my_lab/username and /n/holylfs/LABS/my_lab/username.
You can also use the pull or build commands to download pre-built images from external resources, such as Singularity Hub, the Sylabs Container Library or Docker Hub. For instance, you can download a native Singularity image with its default name from Singularity Hub with:
[user@holyseas02 ~]$ singularity pull shub://vsoch/hello-world 62.32 MiB / 62.32 MiB [================================] 100.00% 43.02 MiB/s 1
The downloaded image file is hello-world_latest.sif.
You can also pull the image with a customized name; e.g., hello.sif:
[user@holyseas02 ~]$ singularity pull --name hello.sif shub://vsoch/hello-world 62.32 MiB / 62.32 MiB [================================] 100.00% 57.26 MiB/s 1s
Similarly, you can pull images from Docker Hub:
[user@holyseas02 ~]$ singularity pull docker://godlovedc/lolcow INFO: Starting build... Getting image source signatures Copying blob sha256:9fb6c798fa41e509b58bccc5c29654c3ff4648b608f5daa67c1aab6a7d02c118 ... Writing manifest to image destination Storing signatures INFO: Creating SIF file... INFO: Build complete: lolcow_latest.sif
See official Singularity documentation for more information.
BioContainers
Cluster nodes automount a CernVM-File System at /cvmfs/singularity.galaxyproject.org/. This provides a universal file system namespace to Singularity images for the BioContainers project, which comprises container images automatically generated from Bioconda software packages. The Singularity images are organized into a directory hierarchy following the convention:
/cvmfs/singularity.galaxyproject.org/FIRST_LETTER/SECOND_LETTER/PACKAGE_NAME:VERSION--CONDA_BUILD
For example:
/cvmfs/singularity.galaxyproject.org/s/a/samtools:1.11--h6270b1f_0
The Bioconda package index lists all software available in /cvmfs/singularity.galaxyproject.org/, while the BioContainers registry provides a searchable interface.
NOTE: There will be a 10-30 second delay when first accessing /cvmfs/singularity.galaxyproject.org/ on a compute node on which it is not currently mounted; in addition, there will be a delay when accessing a Singularity image on a compute node where it has not already been accessed and cached to node-local storage.
Working with images
When working with images you could either start an interactive session, or submit a Singularity job to the available queues. For these examples, we will use a hello-world.sif in an interactive bash shell.
[user@holylogin01 ~]$ salloc -p test -c 1 -t 00-01:00 --mem=4000 [user@holyseas02 ~]$ singularity pull --name hello-world.sif shub://vsoch/hello-world 62.32 MiB / 62.32 MiB [================================] 100.00% 37.63 MiB/s 1s
Shell
With the shell command, you can start a new shell within the container image and interact with it as if it were a small virtual machine.
[user@holyseas02 ~]$ singularity shell hello-world.sif Singularity hello-world.sif:~/holylfs/pgk/SINGULARITY/vol2> pwd /n/home06/pkrastev/holylfs/pgk/SINGULARITY/vol2 Singularity hello-world.sif:~/holylfs/pgk/SINGULARITY/vol2> ls funny.sif gcc-7.2.0.sif hello-world.sif hello.sif lolcow.sif ubuntu.sif vsoch-hello-world-master-latest.sif Singularity hello-world.sif:~/holylfs/pgk/SINGULARITY/vol2> id uid=56139(pkrastev) gid=40273(rc_admin) groups=40273(rc_admin),10006(econh11),34539(fas_it),34540(cluster_users),402119(solexa_writers),402160(VPN_HELPMAN),402161(RT_Users),402854(wpdocs_users),403083(owncloud),403266(file-isi_microsoft-full-dlg),403284(gitlabint_users),403331(rc_class) Singularity hello-world.sif:~/holylfs/pgk/SINGULARITY/vol2>
Commands within a container
You can use the exec command to execute specific commands within the container. For instance, you can run the below command to display information about the native Linux OS of the image:
[user@holyseas02 ~]$ singularity exec hello-world.sif cat /etc/os-release NAME="Ubuntu" VERSION="14.04.5 LTS, Trusty Tahr" ID=ubuntu ID_LIKE=debian PRETTY_NAME="Ubuntu 14.04.5 LTS" VERSION_ID="14.04" HOME_URL="http://www.ubuntu.com/" SUPPORT_URL="http://help.ubuntu.com/" BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
Running containers
Singularity images contain run-scripts that can be triggered with the run command to perform specific actions when the container is run. This can be done either by using the run command, or by calling the container as if it were an executable, i.e.,
[user@holyseas02 ~]$ singularity run hello-world.sif RaawwWWWWWRRRR!!
or
[user@holyseas02 ~]$ ./hello-world.sif RaawwWWWWWRRRR!!
Sometimes you may have a container with several apps, each with its own set of run-scripts. You can use the apps command to list the available apps within the container. For instance, if you have an image named my_image.simg which has N apps (app_1, app_2,…, app_N) you can do:
[user@holyseas02 ~]$ singularity apps my_image.sif app_1 app_2 ... app_N
You can run a particular app with
[user@holyseas02 ~]$ singularity run --app app_2 my_image.sif
GPU example:
To access Nvidia GPU card driver installed inside of Singularity container you need to use –nv option while executing the container.
[user@holylogin01 ~]$ salloc -p gpu --gres=gpu:1 --mem 1000 -n 4 -t 600 [user@supermicgpu01 ~]$ singularity pull --name hello-world.sif shub://vsoch/hello-world [user@supermicgpu01 ~]$ singularity exec --nv hello-world.sif /bin/bash [user@supermicgpu01 ~]$ nvidia-smi +-----------------------------------------------------------------------------+ | NVIDIA-SMI 396.26 Driver Version: 396.26 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | |===============================+======================+======================| | 0 Tesla K20Xm Off | 00000000:88:00.0 Off | 0 | | N/A 37C P0 61W / 235W | 0MiB / 5700MiB | 65% Default | +-------------------------------+----------------------+----------------------+
To verify, if you have access to the requested GPUs, run “nvidia-smi”
Accessing files from a container
Files and directories on the cluster are accessible from within the container. By default, directories under /n, $HOME, $PWD, and /tmp are available at runtime.
You can specify additional directories to bind mount into your container with the --bind option. For instance, in the below example the /scratch directory on the host system is bind mounted to the /mnt directory inside the container:
[user@holyseas02 ~]$ echo 'Hello from inside the container!' > /scratch/hello.dat [user@holyseas02 ~]$ singularity exec --bind /scratch:/mnt hello-world.sif cat /mnt/hello.dat Hello from inside the container!
Singularity containers as SLURM jobs
You can also use Singularity images within a non-interactive batch script as you would any other command. If your image contains a run-script then you can use singularity run to execute the run-script in the job. You can also use singularity exec to execute arbitrary commands (or scripts) within the image. Below is an example batch-job submission script using the hello-world.simg to print out information about the native OS of the image.
#!/bin/bash #SBATCH -J singularity_test #SBATCH -o singularity_test.out #SBATCH -e singularity_test.err #SBATCH -p shared #SBATCH -t 0-00:30 #SBATCH -c 1 #SBATCH --mem=4000 # Singularity command line options singularity exec hello-world.sif cat /etc/os-release
If the above batch-job script is named singularity.sbatch, for instance, the jobs is submitted as usual with sbatch:
[user@holylogin01 ~]$ sbatch singularity.sbatch
Upon the job completion, the STD output is located in the file singularity_test.out.
[user@holylogin01 ~]$ cat singularity_test.out NAME="Ubuntu" VERSION="14.04.5 LTS, Trusty Tahr" ID=ubuntu ID_LIKE=debian PRETTY_NAME="Ubuntu 14.04.5 LTS" VERSION_ID="14.04" HOME_URL="http://www.ubuntu.com/" SUPPORT_URL="http://help.ubuntu.com/" BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
Building Singularity images
To build Singularity containers, you need root access to the build system. Therefore, you cannot build a Singularity container on the cluster. Depending on whether or not you have an access to a Linux machine, possible options are:
- If you have a Linux system to which you have root (admin) access, you can install Singularity and build your Singularity containers there. See Install Singularity on Linux for more information.
- If you don’t have a Linux system you could easily install one in a virtual machine using software like VirtualBox, Vagrant, or VMware. See Install on Windows or Mac for specific Windows or MacOS instructions.
- You can build a Singularity image from a local Docker image on Windows, Mac, or Linux with the docker2singularity Docker image.
In addition to your own Linux environment, you will also need a definition file to build a Singularity container from scratch. You can find some simple definition files for a variety of Linux distributions in the /example directory of the source code. Detailed documentation about building Singularity container images is available at the Singularity website.