Merge pull request #472 from pneerincx/feature/nfs

Feature/nfs

cluster.yml

@@ -51,6 +51,8 @@
 
 - import_playbook: single_group_playbooks/ldap_server.yml
 
+- import_playbook: single_group_playbooks/nfs_server.yml
+
 - import_playbook: single_group_playbooks/cluster_part2.yml
 
 - import_playbook: single_group_playbooks/sys_admin_interface.yml

group_vars/compute_vm.yml

@@ -0,0 +1,10 @@
+---
+volumes:
+  - mount_point: '/local'
+    device: '/dev/vdb'
+    mounted_owner: root
+    mounted_group: root
+    mounted_mode: '0755'
+    mount_options: 'rw,relatime'
+    type: ext4
+...

group_vars/wingedhelix_cluster/vars.yml

@@ -125,10 +125,11 @@ regular_users:
 #
 # Shared storage related variables
 #
-lustre_client_networks: 'tcp11(vlan???.lustre),tcp12(vlan???.lustre)'
+#lustre_client_networks: 'tcp11(vlan???.lustre),tcp12(vlan???.lustre)'
 pfs_mounts: [
   { pfs: 'umcgst12',
-    source: '',      # this is going to be filled when we have shared NFS server
+    device: '/dev/vdb',
+    source: 'lookup nfs_server ip address in storage VLAN',
     type: 'nfs4',    # checked with cat /proc/filesystem
     rw_options: 'defaults,_netdev,flock',
     ro_options: 'defaults,_netdev,ro' },

roles/lustre_client/defaults/main.yml

@@ -3,5 +3,5 @@
 # We cannot define a default for lustre_client_networks here;
 # Always configure lustre_client_networks in group_vars/[group]/vars.yml
 #
-#lustre_client_networks: 'tcp11(vlanMISSING.lustre),tcp12(vlanMISSING.lustre)'
-...
+# lustre_client_networks: 'tcp11(vlanMISSING.lustre),tcp12(vlanMISSING.lustre)'
+...

roles/lustre_client/tasks/install.yml

@@ -71,7 +71,8 @@
   become: true
   notify: reboot
 
-- meta: flush_handlers
+- name: Flush handlers
+  meta: flush_handlers
 
 - name: Install the Lustre client.
   yum:

roles/mount_volume/defaults/main.yml

@@ -1,10 +1,14 @@
----
-volumes:
-  - mount_point: '/local'
-    device: '/dev/vdb'
-    mounted_owner: root
-    mounted_group: root
-    mounted_mode: '0755'
-    mount_options: 'rw,relatime'
-    type: ext4
-...
+#
+# Example data structure used by the mount_volume role.
+#
+# ---
+# volumes:
+#   - mount_point: '/local'
+#     device: '/dev/vdb'
+#     mounted_owner: root
+#     mounted_group: root
+#     mounted_mode: '0755'
+#     mount_options: 'rw,relatime'
+#     type: ext4
+# ...
+#

roles/mount_volume/tasks/main.yml

@@ -1,4 +1,19 @@
 ---
+- name: 'Compile list of volumes to mount for NFS servers.'
+  ansible.builtin.set_fact:
+    volumes: "{{ volumes | default([]) + [ volume ] }}"
+  vars:
+    volume:
+      mount_point: "/mnt/{{ item.pfs }}"
+      device: "{{ item.device }}"
+      mounted_owner: root
+      mounted_group: root
+      mounted_mode: '0755'
+      mount_options: 'rw,relatime'
+      type: ext4
+  when: inventory_hostname in groups['nfs_server'] | default([])
+  with_items: "{{ pfs_mounts | selectattr('type', 'search', 'nfs') | selectattr('device', 'defined') | list }}"
+
 - name: 'Check the mount point.'
   command: "mountpoint {{ item.mount_point }}"
   register: mount_point_status

roles/nfs_client/tasks/main.yml

@@ -0,0 +1,7 @@
+---
+- name: Install NFS utils.
+  yum:
+    name: nfs-utils
+  when: pfs_mounts | selectattr('type', 'search', 'nfs') | list | length >= 1
+  become: true
+...

roles/nfs_server/handlers/main.yml

@@ -0,0 +1,30 @@
+---
+#
+# Important: maintain correct handler order.
+# Handlers are executed in the order in which they are defined
+# and not in the order in which they are listed in a "notify: handler_name" statement!
+#
+# Restart before reload: an reload after a restart may be redundant but should not fail,
+# but the other way around may fail when the impact of changes was too large for a reload.
+#
+- name: 'Restart nfs-server service.'
+  systemd:
+    name: 'nfs-server.service'
+    state: restarted
+    daemon_reload: true
+  become: true
+  listen: restart_nfs-server
+
+- name: 'Reload nfs-server service.'
+  systemd:
+    name: 'nfs-server.service'
+    state: reloaded
+    daemon_reload: true
+  become: true
+  listen: reload_nfs-server
+
+- name: '(Re)export all NFS shares.'
+  command: 'exportfs -a'
+  become: true
+  listen: export_nfs_shares
+...

roles/nfs_server/tasks/main.yml

@@ -0,0 +1,34 @@
+#
+# This role may depend on the mount_volume role to mount the filesystems that must be exported over NFS.
+#
+---
+- name: 'Install NFS utils.'
+  yum:
+    name: nfs-utils
+  notify:
+    - restart_nfs-server
+    - export_nfs_shares
+  become: true
+
+- name: 'Enable and start nfs-server service.'
+  systemd:
+    name: nfs-server.service
+    state: 'started'
+    enabled: true
+    daemon_reload: true
+  notify:
+    - export_nfs_shares
+  become: true
+
+- name: 'Add NFS share to /etc/exports.'
+  lineinfile:
+    path: /etc/exports
+    line: "/mnt/{{ item.pfs }} {{ network_private_storage_cidr }}(rw,sync,root_squash,no_subtree_check)"
+  with_items: "{{ pfs_mounts | selectattr('type', 'search', 'nfs') | selectattr('device', 'defined') | list }}"
+  notify:
+    - export_nfs_shares
+  become: true
+
+- name: Flush handlers
+  meta: flush_handlers
+...

roles/shared_storage/tasks/main.yml

@@ -28,7 +28,9 @@
     opts: "{{ item.rw_options }}"
     state: 'mounted'
   with_items: "{{ pfs_mounts }}"
-  when: inventory_hostname in groups['sys_admin_interface']
+  when:
+    - inventory_hostname in groups['sys_admin_interface']
+    - inventory_hostname not in groups ['nfs_server'] | default([]) or 'nfs' not in item.type
   become: true
 
 - name: 'Create "home" Logical File System (LFS) on Physical File Systems (PFSs) mounted on SAIs.'

single_group_playbooks/cluster_part2.yml

@@ -8,5 +8,7 @@
     - sshd
     - ldap  # client
     - regular_users
+    - lustre_client
+    - nfs_client
     - shared_storage
 ...

single_group_playbooks/nfs_server.yml

@@ -0,0 +1,14 @@
+#
+# This play depends on:
+#   * cluster_part1.yml
+#
+---
+- import_playbook: pre_deploy_checks.yml
+
+- name: '###==-> Extra roles for NFS server. <-==###'
+  hosts:
+    - nfs_server
+  roles:
+    - mount_volume
+    - nfs_server
+...

single_group_playbooks/sys_admin_interface.yml

@@ -10,7 +10,6 @@
   hosts:
     - sys_admin_interface
   roles:
-    - mount_volume
     - slurm_management
     - prom_server
     - grafana

single_role_playbooks/mount_volume.yml

@@ -1,7 +1,7 @@
 - name: Mount the ceph volume
   hosts:
+     - nfs_server
      - compute_vm
-     - sys_admin_interface
      - deploy_admin_interface
      - data_transfer
   roles:

static_inventories/wingedhelix_hosts.ini

@@ -10,6 +10,9 @@ wh-repo
 [docs]
 docs
 
+[nfs_server]
+wh-sai
+
 [sys_admin_interface]
 wh-sai