2026 - 05 - 19 - New low-priority lazyqueue available on Mjolnir

Dear Mjolnir users,

I have now introduced a new low-priority queue on Mjolnir named:

using the QOS:

The purpose of the lazy queue is to improve overall cluster utilization and sustainability by allowing users to run non-time-sensitive jobs using otherwise idle compute resources.

Jobs submitted to the lazy queue:

- run at very low priority,

- only use currently available idle resources,

- and may automatically be requeued if higher-priority jobs require the resources.

This allows Mjolnir to make productive use of compute capacity that would otherwise remain unused, while still ensuring that normal and interactive workloads receive priority access to resources.

The lazy queue is particularly well suited for:

- exploratory analyses,

- large BLAST searches,

- metagenomic indexing,

- database construction,

- parameter sweeps,

- AI/embedding generation,

- short-running Nextflow or Snakemake workflows,

- and other non-urgent background jobs.

To use it, simply add:

to your Slurm submission script.

- Jobs in the lazy queue may be interrupted and automatically requeued.

- Users are therefore strongly encouraged to use checkpoint-capable workflows where possible.

- The queue is not recommended for time-sensitive analyses or jobs that cannot safely restart.

- The queue shares the same hardware resources as the normal cpuqueue, but runs at lower priority.

As part of ongoing sustainability and optimization efforts on Mjolnir, I am also working on a daily script to help users better understand CPU and memory efficiency for completed jobs only, in order to improve queue throughput and reduce unnecessary resource reservation.

Coming to your inbox soon. 

Best regards,

Bent