This is the current queue setup for Zero.
|Queue/partition||Max walltime/hours||Default walltime/hours||Notes|
|test||0.5||0.5||High priority queue for testing|
|cluster||672||24||The default queue|
By default a job will be assigned one core. If you want more resources (cores, memory, whole nodes) you can ask for them with various options to the SLURM commands. See SLURM usage for details.
The scheduling configuration aims to achieve the following:
- A job in the test partition should go straight to the top of the queue
- Priority in the queue is influenced by fairshare, ie how much cpu time a user account has used recently. Every account starts off with equal fairshare. If two or more people want to redistribute their total fairshare allowance between their accounts please contact the computer officers as we can easily do this.
- Backfill is enabled. What this means is that if the top queued job can't run the scheduler looks at jobs further down the queue and runs any that would not affect the top job.
- Jobs in the 'external' quality of service group are restricted to occupy no more than 368 cores at a time in total.
When people leave the Chemistry department (physically leave, that is) they are automatically put into the 'external' quality of service group. This is done by checking the department database. Everyone else is in the 'normal' quality of service group which has no restrictions on number of cores.