👌 IMPROVE: Remove hostname setting

Instead dynamically set the slurm configuration with the original hostname.

README.md

@@ -9,10 +9,8 @@ An Ansible role that installs the [slurm](https://slurm.schedmd.com/) workload m
 The role:
 
 - Installs the slurm packages
-- Sets the hostname to be that defined in `slurm_hostname`
-- If `slurm_hostname_service: true` adds a service to set the hostname on VM start-up (required for cloud platforms)
-- Sets up the slurm configuration (`/etc/slurm-llnl/slurm.conf`) to dynamically use the correct platform resources (#CPUs, etc), configuring one node and one partition.
-- Adds a `slurm-resources` script and start-up service to automate the initiation of correct platform resources (required if creating a VM instance with different resources to the build VM image)
+- Sets up the slurm configuration (`/etc/slurm-llnl/slurm.conf`) to dynamically use the correct platform resources (hostname, #CPUs, etc), configuring one node (named `$HOSTNAME`) and one partition (named `slurm_partition_name`).
+- Adds a `slurm-resources` script and start-up service to automate the initiation of correct platform resources (required if creating a VM image where instances may have different resources)
 - Starts the slurm services.
 
 To check the services are running (assuming systemd in use):
@@ -48,10 +46,6 @@ $ slurm-resources -e restart_on_change=true -e slurm_max_cpus=2
 
 This will update the resources defined for the node, set the maximum CPUs for the partition to 2 (independent of the CPUs available on the node), and restart the slurm services with the updated configuration (if the configuration has changed).
 
-**NOTE!**
-It is important that the hostname is properly set in the machine
-(both with `hostname <HOSTNAME>` (which sets `/etc/hostname`) and in the `/etc/hosts` file, in the line with the IP address (e.g. in docker this line should read `172.x.x.x <HOSTNAME>`, where `<HOSTNAME>` should be replaced with the hostname, and should match the variable `slurm_hostname` (default value: `qmobile`).
-
 ## Installation
 
 `ansible-galaxy install marvel-nccr.slurm`

defaults/main.yml

@@ -1,15 +1,8 @@
-slurm_hostname: qmobile
 slurm_user: slurm
-slurm_cluster_name: "{{ slurm_hostname }}"
 slurm_partition_name: jobs
 
 slurm_test_folder: "/tmp/slurm-tests"
 
-# Adds a system service that forces the hostname on startup
-# Note: This is e.g. necessary on cloud platforms like AWS
-slurm_hostname_service: false
-slurm_set_hostname: false
-
 # Enables the slurm-resources system service that re-configures the slurm compute resources on startup
 # This is necessary when preparing image that can start on a different hardware than it was built on
 slurm_resources_service_enabled: true

molecule/default/converge.yml

@@ -1,9 +1,4 @@
 - hosts: all
 
-  vars:
-  - run_tests: true
-  - cloud_platform: docker
-  - slurm_hostname_service: true
-
   roles:
   - role: marvel-nccr.slurm

molecule/default/molecule.yml

@@ -33,3 +33,4 @@ provisioner:
       all:
         vars:
           ansible_python_interpreter: /usr/bin/python3
+          run_tests: true

tasks/main.yml

@@ -75,11 +75,6 @@
   args:
     creates: "/etc/munge/munge.key"
 
-- import_tasks: hostname.yml
-
-- include_tasks: hostname_service.yml
-  when: slurm_hostname_service
-
 # munge should start before slurm daemons
 # https://slurm.schedmd.com/quickstart_admin.html
 - name: start munge service

templates/config-playbook.yml.j2

@@ -2,7 +2,6 @@
   hosts: localhost
 
   vars:
-    slurm_hostname: "{{ slurm_hostname }}"
     slurm_partition_name: "{{ slurm_partition_name }}"
 {% raw %}
     slurm_conf_file: /etc/slurm-llnl/slurm.conf
@@ -15,14 +14,22 @@
     - debug:
         msg: "Run: {{ lookup('pipe', 'date +%Y-%m-%d-%H:%M:%S') }}"
 
-    - name: Update SLURM configuration
+    - name: "Update SLURM ControlMachine={{ ansible_hostname }}"
+      become: true
+      lineinfile:
+        dest: "{{ slurm_conf_file }}"
+        regexp: '^ControlMachine='
+        line: "ControlMachine={{ ansible_hostname }}"
+        state: present
+
+    - name: Update SLURM node configuration
       become: true
       blockinfile:
         path: "{{ slurm_conf_file }}"
         marker: "# {mark} ANSIBLE MANAGED NODES"
         block: |
-          NodeName={{ slurm_hostname }} Sockets={{ ansible_processor_count }} CoresPerSocket={{ ansible_processor_cores }} ThreadsPerCore={{ ansible_processor_threads_per_core }} State=UNKNOWN
-          PartitionName={{ slurm_partition_name }} Nodes={{ slurm_hostname }} Default=YES MaxTime=INFINITE State=UP MaxNodes=1 MaxCPUsPerNode={{ slurm_max_cpus }}
+          NodeName={{ ansible_hostname }} Sockets={{ ansible_processor_count }} CoresPerSocket={{ ansible_processor_cores }} ThreadsPerCore={{ ansible_processor_threads_per_core }} State=UNKNOWN
+          PartitionName={{ slurm_partition_name }} Nodes={{ ansible_hostname }} Default=YES MaxTime=INFINITE State=UP MaxNodes=1 MaxCPUsPerNode={{ slurm_max_cpus }}
         backup: yes
       register: update
 

templates/slurm.conf

@@ -2,7 +2,7 @@
 # Put this file on all nodes of your cluster.
 # See the slurm.conf man page for more information.
 #
-ControlMachine={{ slurm_hostname }}
+ControlMachine={{ ansible_hostname }}
 #ControlAddr=
 # 
 MailProg=/usr/sbin/sendmail
@@ -53,6 +53,6 @@ SlurmdLogFile=/var/log/slurm-llnl/slurmd.log
 # 
 # 
 # BEGIN ANSIBLE MANAGED NODES
-NodeName={{ slurm_hostname }} Sockets={{ ansible_processor_count }} CoresPerSocket={{ ansible_processor_cores }} ThreadsPerCore={{ ansible_processor_threads_per_core }} State=UNKNOWN
-PartitionName={{ slurm_partition_name }} Nodes={{ slurm_hostname }} Default=YES MaxTime=INFINITE State=UP MaxNodes=1 MaxCPUsPerNode={{ ansible_processor_vcpus }}
+NodeName={{ ansible_hostname }} Sockets={{ ansible_processor_count }} CoresPerSocket={{ ansible_processor_cores }} ThreadsPerCore={{ ansible_processor_threads_per_core }} State=UNKNOWN
+PartitionName={{ slurm_partition_name }} Nodes={{ ansible_hostname }} Default=YES MaxTime=INFINITE State=UP MaxNodes=1 MaxCPUsPerNode={{ ansible_processor_vcpus }}
 # END ANSIBLE MANAGED NODES