SSD resource configuration issues on cluster compute nodes

Install, Configure and Update
1

Because the local SSD hard drive size of the compute nodes on the cluster is limited, and there are multiple instances, it will be very slow without SSD acceleration.

I added the SSD to the Slurm resource pool, and requesting the corresponding size allows for normal scheduling. There’s only one question: what variable can I use to get the size of the job file to be cached? I’ve written it in the script.

2

Please answer this question as soon as possible; the corresponding examples and parameters cannot be found in the documentation.

3

who can answer me question？

4

@ltf Please review this guide section about an estimate.
If the estimate changes from job to job, you may find configuring a job-level cluster custom variable inside the script template useful.