Wiki documenting the CBRG cluster setup
#
- only right option
key works - left will not give you #
|
and ~
might swap around - to correct log out of sessions and log back on. Slurm
See queues = sinfo
- for training use `teaching`
- for general use `batch`
You can check how much your requesting using
squeue --me -o "%.18i %.9P %.8j %.8u %.2t %.10M %.6D %R %C %d %D %A %H %J %m %S %z %Z"
qrsh
From Ewan: you can do:
$ srun --pty bash
which also takes all the usual parameters that squeue does for allocating memory
and so forth. However, while you can do it if absolutely necessary or just for teaching purposes
, in practice we don't generally recommend people do it 'for real' on our system
- our usual MO is to run stuff on the normal interactive login nodes (cbrglogin1 or cbrglogin2),
- if it's too big for them then run it on the large login node (cbrglogin3),
- and if it's too big for *that* then just deal with it and run it as a batch job.
- Lucy - this is what I have been doing: srun -p batch --cpus-per-task=5 --mem-per-cpu=10G --pty bash -i
sims-lab/shared
folder -> change permissions with chmod 770 ( 777 does not work)
Where to put them:
/t1-data/home/
= faster loading but limited to 20G
Guide to installing conda on CCB cluster
Exporting conda envs yml files:
conda env export --no-builds > environment_nobuild.yml
conda env export > environment.yml
= full env details, version and hash code for software build conda env export --from-history > env_history.yml
= gives you just the main things you asked for (not all the dependencies) conda list --revisions > env_revisions.txt
= the order and date you installed extra packagesconda list --explicit > proj095_list_explicit.txt
= really thourgh list of exactly what you had in your environment which conda
path changes on conda activate