search

stackhpc / ansible-role-openhpc

Ansible role for OpenHPC
Apache License 2.0
44 stars 15 forks source link

Enable use of custom Slurm builds #163

Open sjpb opened 10 months ago

sjpb commented 10 months ago

  1. Provides an install-generic.yml task book which can use existing Slurm binaries/libraries rather than installing OpenHPC packages.

    Note that the ./configure command for the Slurm build used with this must:

    • Set --sysconfdir to match the directory of openhpc_slurm_conf_path (default: /etc/slurm/slurm.conf)
    • Set --prefix to match the directory the slurm daemon binaries are deployed at (see openhpc_sbin_dir). This is because PluginDir defaults to being based on the build prefix. Although it can be overriden in slurm.conf, the slurmds do not appear to get this parameter when running configless, so they won't start saying the (default) plugin dir doesn't exist

  2. Adds openhpc_config_files to allow arbitrary Slurm configuration files to be templated.

  3. Adds openhpc_slurmd_spool_dir to allow SlurmdSpoolDir parameter to be modified.

TODO: needs to be updated to support RL9, see cgroups change in https://github.com/stackhpc/ansible-role-openhpc/pull/164, this should be done by default.

m-bull commented 9 months ago

How come all the tests fail here?

sjpb commented 9 months ago

Tested on https://github.com/stackhpc/ansible-slurm-appliance/pull/312 slurm works, ood works, monitoring works.