1. SLURM Job Manager, smanage
Slurm Manage, for submitting and reporting on job arrays run on slurm
Practices for Reproducible Research
Kerberos is an authentication method that will give you access to groups of resources based on tickets that you generate. And yes, Kerberos is also a three headed dog from Greek Mythology.
You will use this method via the tools kinit
when you need to log into a kerberos
realm. For example, the cluster at Stanford University use Kerberos.
The general authentication flow looks like this:
[1. client] --> [2. key distribution center] --> [3. credential] --> [4. login]
Briefly, you will use kinit
from your computer to identify yourself with your email and request
a ticket from the key distribution center. The ticket will have a life and expiration, and be
used to authenticate you to different realm resources. This is a way oversimplification, and
we point the interested reader to learn more about this workflow by reading
this nice post.
If you are on Windows or Mac, you can first download a client called Kerberos Commander. If you are on Linux, then you can follow the general instructions provided for the Sherlock cluster. We will review these steps here!
We first install system dependencies, which include the local client and openssh:
$ sudo apt-get install krb5-user openssh-client
The configuration for the realm can be obtained by using wget. If you aren’t using a Stanford realm, there might
be a different configuration file for you to download, and you would equally want to get the *.conf file that
describes the realm and output it to /etc/krb5.conf
sudo curl -o /etc/krb5.conf https://web.stanford.edu/dept/its/support/kerberos/dist/krb5.conf
Take a look at this file, it’s kind of neat! You can see many different realms defined and various configuration variables.
cat /etc/krb5.conf
Remember the step to get your ticket? You will do that by typing kinit email@stanford.edu
. Then you will see the credential with klist
.
$ kinit yourSUnet@stanford.edu
# See your kerberos ticket:
$ klist
Ticket cache: FILE:/tmp/krb5cc_1000
Default principal: yourSUnet@stanford.edu
Valid starting Expires Service principal
05/08/2018 07:31:27 05/09/2018 07:31:23 krbtgt/stanford.edu@stanford.edu
We would want to configure ssh (a secure shell) to access resources of interest, for example,
the Sherlock cluster at Stanford! Your computer has a file at ~/.ssh/config
where we can
add variables that will be used for this workflow. Here is how to edit that file so that you can
log in to Sherlock:
echo "Host *
GSSAPIDelegateCredentials yes
GSSAPIAuthentication yes" >> ~/.ssh/config
Once that is done, you should be able to login to Sherlock via Kerberos like this:
$ ssh yourSUnet@login.sherlock.stanford.edu
Have a question, need a clarification, or anything else? Ask away!
This series guides you through getting started with HPC cluster computing.
Slurm Manage, for submitting and reporting on job arrays run on slurm
A Quick Start to using Singularity on the Sherlock Cluster
Use the Containershare templates and containers on the Stanford Clusters
A custom built pytorch and Singularity image on the Sherlock cluster
Use Jupyter Notebooks via Singularity Containers on Sherlock with Port Forwarding
Use R via a Jupyter Notebook on Sherlock
Use Jupyter Notebooks (optionally with GPU) on Sherlock with Port Forwarding
Use Jupyter Notebooks on Sherlock with Port Forwarding
A native and container-based approach to using Keras for Machine learning with R
How to create and extract rar archives with Python and containers
Getting started with SLURM
Getting started with the Sherlock Cluster at Stanford University
Getting started with the Sherlock Cluster at Stanford University
Using Kerberos to authenticate to a set of resources