Open slurmdbd.conf to full customization; Enable slurmdbd purge by default

[elelaysh]

Builds on #209

[gemini-code-assist]

Code Review

This pull request refactors the generation of slurmdbd.conf to be fully customizable through Ansible variables and centralizes the logic for computing node properties for slurm.conf. It also enables database purging by default. However, the full customization of slurmdbd.conf introduces a configuration injection vulnerability because keys and values are not validated or escaped for newlines, and proper quoting for values containing special characters is missing. Furthermore, a potential issue exists with how a complex variable is constructed, which could lead to errors, and there's a minor issue with a missing newline in a template.

[gemini-code-assist]

The current implementation for constructing ohpc_nodegroups_computed builds a string that is supposed to be a YAML/JSON dictionary. This is fragile and can fail due to a trailing comma if the last item in the loop meets the condition, which would result in invalid YAML. A more robust approach is to build a list of dictionary-item strings and then join them. This avoids manual string formatting and ensures valid output without a trailing comma.

ohpc_nodegroups_computed: >
  {
  {% set items = [] %}
  {% for nodegroup in openhpc_nodegroups %}
  {%  set inventory_group_name = openhpc_cluster_name ~ '_' ~ nodegroup.name %}
  {%  set inventory_group_hosts = groups.get(inventory_group_name, []) %}
  {%      if inventory_group_hosts | length > 0 %}
  {%          set play_group_hosts = inventory_group_hosts | intersect (play_hosts) %}
  {%          set first_host = play_group_hosts | first | mandatory('Inventory group "' ~ inventory_group_name ~ '" contains no hosts in this play - was --limit used?') %}
  {%          set first_host_hv = hostvars[first_host] %}
  {%          set ram_mb = (first_host_hv['ansible_memory_mb']['real']['total'] * (nodegroup.ram_multiplier | default(openhpc_ram_multiplier))) | int %}
  {%          capture item_str %}
      {{ nodegroup.name | to_json }}: {
          "inventory_group_name": {{ inventory_group_name | to_json }},
          "first_host": {{ first_host | to_json }},
          "ram_mb": {{ ram_mb }},
          "def_mem_per_cpu": {{ (ram_mb / first_host_hv['ansible_processor_vcpus']) | int }}
      }
  {%          endcapture %}
  {%          set _ = items.append(item_str) %}
  {%      endif %}
  {% endfor %}
  {{ items | join(',\n') }}
  }

[gemini-code-assist]

The loop used to generate slurmdbd.conf is vulnerable to configuration injection. If any of the keys or values in openhpc_slurmdbd_merged_config contain newline characters, an attacker who can control these variables can inject arbitrary configuration directives into the slurmdbd.conf file. This could lead to unauthorized configuration changes, such as modifying authentication settings or log file locations. Additionally, values containing special characters like # or spaces are not properly quoted, which can lead to parsing errors or truncated values (e.g., in passwords).

[gemini-code-assist]

It's a good practice for text-based configuration files to end with a newline character. This file is missing one, which can cause issues with some tools or scripts. Please add a newline at the end of the file.

{% endfor %}