add openhpc_login_only_nodes group so can run without login nodes in --limit

Author: sjpb

README.md

@@ -25,17 +25,18 @@ package in the image.
 * `control`: whether to enable control host
 * `database`: whether to enable slurmdbd
 * `batch`: whether to enable compute nodes
-* `login`: whether to enable login-only nodes in configless mode (for combined control/login nodes this can be left false)
 * `runtime`: whether to enable OpenHPC runtime
 * `drain`: whether to drain compute nodes
 * `resume`: whether to resume compute nodes
 
 `openhpc_slurmdbd_host`: Optional. Where to deploy slurmdbd if are using this role to deploy slurmdbd, otherwise where an existing slurmdbd is running. This should be the name of a host in your inventory. Set this to `none` to prevent the role from managing slurmdbd. Defaults to `openhpc_slurm_control_host`.
 
-`openhpc_slurm_configless`: Optional, default false. If True then slurm's ["configless" mode](https://slurm.schedmd.com/configless_slurm.html) is used. **NB: Requires Centos8/OpenHPC v2.**
+`openhpc_slurm_configless`: Optional, default false. If true then slurm's ["configless" mode](https://slurm.schedmd.com/configless_slurm.html) is used. **NB: Requires Centos8/OpenHPC v2.**
 
 `openhpc_munge_key_path`: Optional, default ''. Define a path for a local file containing a munge key to use, otherwise one will be generated on the slurm control node.
 
+`openhpc_login_only_nodes`: Optional. If using "configless" mode specify the name of an ansible group containing nodes which are login-only nodes (i.e. not also control nodes), if required. These nodes will run `slurmd` to contact the control node for config.
+
 ### slurm.conf
 
 `openhpc_slurm_partitions`: list of one or more slurm partitions.  Each partition may contain the following values:

defaults/main.yml

@@ -38,7 +38,6 @@ openhpc_slurmdbd_mysql_username: slurm
 openhpc_enable:
   control: false
   batch: false
-  login: true
   database: false
   runtime: false
   drain: false
@@ -52,3 +51,4 @@ ohpc_release_repos:
   "8": "http://repos.openhpc.community/OpenHPC/2/CentOS_8/x86_64/ohpc-release-2-1.el8.x86_64.rpm" # ohpc v2 for Centos 8
 openhpc_slurm_configless: false
 openhpc_munge_key: ''
+openhpc_login_only_nodes: ''

molecule/README.md

@@ -15,6 +15,7 @@ test4  | 1            | N                       | 2x compute node, accounting en
 test5  | 1            | N                       | As for #1 but configless
 test6  | 1            | N                       | 0x compute nodes, configless
 test7  | 1            | N                       | 1x compute node, no login node, configless
+test8  | 1            | N                       | 2x compute node, 2x login-only nodes, configless
 
 # Local Installation & Running
 

molecule/test8/converge.yml

@@ -16,4 +16,5 @@
           - name: "compute"
         openhpc_cluster_name: testohpc
         openhpc_slurm_configless: true
+        openhpc_login_only_nodes: 'testohpc_login'
 

molecule/test8/molecule.yml

@@ -73,5 +73,7 @@ platforms:
       - name: net1
 provisioner:
   name: ansible
+  # ansible_args:
+  #   - --limit=testohpc-control,testohpc-compute-0
 verifier:
   name: ansible

tasks/main.yml

@@ -7,6 +7,11 @@
   when: "openhpc_enable.get(item, false)"
   tags: always
 
+- name: Set slurmd as service for openhpc_login_only_nodes
+  set_fact:
+    openhpc_slurm_service: "slurmd"
+  when: openhpc_login_only_nodes and (openhpc_login_only_nodes in group_names)
+
 - name: Install packages
   block:
     - include_tasks: install.yml

tasks/runtime.yml

@@ -43,11 +43,6 @@
   notify:
     - Restart Munge service
 
-- name: Set fact for login-only nodes
-  # needed b/c openhpc_enable doesn't end up in hostvars
-  set_fact:
-    openhpc_login_only: "{{ openhpc_enable.login | default(false) | bool }}"
-
 - name: Template slurmdbd.conf
   template:
     src: slurmdbd.conf.j2
@@ -80,7 +75,7 @@
     group: root
     mode: 0644
   when:
-    - (openhpc_enable.batch | default(false) | bool) or (openhpc_enable.login | default(false) | bool)
+    - openhpc_slurm_service == 'slurmd'
     - openhpc_slurm_configless
   notify:
     - Reload SLURM service

templates/slurm.conf.j2

@@ -139,11 +139,9 @@ PartitionName={{part.name}} \
 
 
 # Define slurmd nodes not in partitions for configless login-only nodes:
-{% for node in hostvars %}
-    {% if hostvars[node].openhpc_login_only %}
+{%if openhpc_login_only_nodes %}{% for node in groups[openhpc_login_only_nodes] %}
 NodeName={{ node }}
-    {% endif %}
-{% endfor %}{# configless login-only #}
+{% endfor %}{% endif %}
 
 # Want nodes that drop out of SLURM's configuration to be automatically
 # returned to service when they come back.