Migrating from HPC
What is the same ?
- your userid
- your shell (bash)
- your storage (/gpfs/home/... and /gpfs/scratch/...)
- the module system
What is different ?
- the address - you need to use ada.uea.ac.uk
- the base operating system is newer - which means code you have compiled may need to be recompiled
- the modules - we have installed the latest versions of software
- the queues - there are fewer queues, and you need to set a time limit
- the default queue "compute" should be your starting point - talk to us before you use anything else
- the queueing system - on HPC it is the licenced LSF system, one ADA it is the opensource SLURM system. This has save a lot of money on licences, which has been used instead on increasing the number of nodes
You can find detailed information about the new settings on the relevant pages on this site.
What do I need to do ?
- if you haven't already, apply for an account on ada.uea.ac.uk
- check which software modules are available and change your .bashrc accordingly if you automatically add modules. Please ask for advice if you are unsure.
- Do NOT set modules in .bashrc. Slurm passes the environment (eg modules, python, conda). If you set these in your .bashrc then it can cause unexpected and unwanted variable assignments when you run slurm jobs (both interactive and sbatch).
Modules should be set explicitly in sbatch scripts
frequently used combinations in interactive node can be set as an alias in .bashrc
eg alias stata16="module add stata/16"
alias mpi-intel=module add mpi/openmpi/4.0.3/intel/ib"
- recompile any code you have built yourself
- look at the web pages about Queues, Memory, and Jobs - there are example submission scripts
- rewrite your submissions scrips using SLURM rather than LSF commands
- learn which commands to use to submit, manage and monitor jobs