Please read carefully the instructions below for a considerate usage of blafis!
Usage of the blafis cluster should be made with care and consideration for other Gr@v group members. The blafis cluster is managed in a shared scheme with other research groups, where each group owns a fraction of the processors, but they are available to be used/borrowed among groups if not active.
Currently, our group (named astro) owns 84 cores, but there is a floating quota that goes up to 211 cores (i.e. +127 from other groups). If this is fully in use by a fraction of the group users, other group users will have their jobs blocked! Thus, in periods of high demand the following guidelines should be followed for a fair usage:
- Check how many users are running jobs (see panel below or run the following line in the terminal after logging into blafis: a='grep -e JOB';for i in $(members -p astro); do a=$a" -e "$i; done; showq | $a)
- Compute the number of processors for each active user in a democratic split, i.e. 211/(# active users). If you need all of these and they are not available, politely email the current users (or meet them) explaining the situation, and negotiate for some cores to be freed up.
- Finally, an effort should be made to leave some cores free in case someone who's not active decides to do some quick test jobs. The guideline is typically between 12 and 24 cores (1-2 nodes).
- Important! In periods of high demand, our group will only be allowed to use up to 84 cores if other groups are filling their quota. In that case you should divide 84/(# active users).