Fix templating

Author: jovial

defaults/main.yml

@@ -13,7 +13,9 @@ openhpc_retry_delay: 10
 openhpc_job_maxtime: '60-0' # quote this to avoid ansible converting some formats to seconds, which is interpreted as minutes by Slurm
 openhpc_config: "{{ openhpc_extra_config | default({}) }}"
 openhpc_gres_template: gres.conf.j2
-openhpc_gres_autodetect: 'disabled'
+# Autodetection mechanism to use: https://slurm.schedmd.com/gres.html#AutoDetect e.g nvml.
+# NOTE: 'off' is a special sentinal value to indicate that autodetection should not be used.
+openhpc_gres_autodetect: 'off'
 openhpc_slurm_configless: "{{ 'enable_configless' in openhpc_config.get('SlurmctldParameters', []) }}"
 
 openhpc_state_save_location: /var/spool/slurm

filter_plugins/slurm_conf.py

@@ -72,7 +72,7 @@ def group_by_gres_autodetect(context, hosts):
     """
     result = defaultdict(list)
     for host in hosts:
-        gres_autodetect = _get_hostvar(context, 'gres_autodetect', inventory_hostname=host)
+        gres_autodetect = _get_hostvar(context, 'openhpc_gres_autodetect', inventory_hostname=host) or 'off'
         result[gres_autodetect].append(host)
     return result
 

templates/gres.conf.j2

@@ -1,29 +1,28 @@
 AutoDetect=off
 {% for part in openhpc_slurm_partitions %}
 {%  set nodelist = [] %}
-{%  set donehosts = set() %}
+{%  set donehosts = [] | unique %}
 {%  for group in part.get('groups', [part]) %}
 {%      if 'gres' in group %}
 {%          set group_name = group.cluster_name|default(openhpc_cluster_name) ~ '_' ~ group.name %}
 {%          set inventory_group_hosts = groups.get(group_name, []) %}
 {%          set autodetect_mechanisms = inventory_group_hosts | group_by_gres_autodetect %}
 {%          for mechanism, _mechanism_hosts in autodetect_mechanisms.items() %}
-{%              set mechanism_hosts = _mechanism_hosts - donehosts %}
+{%              set mechanism_hosts = _mechanism_hosts | difference(donehosts) %}
 {%              if mechanism != 'disabled' %}
 {%                  for hostlist in (mechanism_hosts | hostlist_expression) %}
 NodeName={{ hostlist }} AutoDetect={{ mechanism }}
-
 {%                  endfor %}
 {%              else %}
 {%                  for gres in group.gres %}
 {%                      set gres_name, gres_type, _ = gres.conf.split(':') %}
 {%                      for hostlist in (mechanism_hosts | hostlist_expression) %}
-NodeName={{ hostlist }} Name={{ gres_name }} Type={{ gres_type }} File={{ gres.file }}
+NodeName={{ hostlist }} Name={{ gres_name }} Type={{ gres_type }} File={{ gres.file | mandatory('The gres configuration dictionary: ' ~  gres ~ ' is missing the file key, but openhpc_gres_autodetect is set to off. The error occured on partition: ' ~  part.name ~ '. Please add the file key or set openhpc_gres_autodetect.') }}
 
 {%                      endfor %}
 {%                  endfor %}
 {%              endif %}
-                {% set _ = donehosts.update(mechanism_hosts) %}
+{% set donehosts = donehosts | union(mechanism_hosts) %}
 {%          endfor %}
 {%      endif %}
 {%  endfor %}