Rewrite plugin for src cluster support

README.md

@@ -52,15 +52,14 @@ alt_home_location: /data/volume_3
 # Vagrant user is instructor
 # The students defined below can be used to login as a student
 students: 'student1:pw1,student2:pw2'
+worker_ip_addresses: []
 ```
 
 The token can be found in the eWaterCycle password manager.
 
 ```shell
 vagrant --version
 # Vagrant 2.4.1
-vagrant plugin install vagrant-vbguest
-# Installed the plugin 'vagrant-vbguest (0.32.0)'
 vagrant up
 ```
 

research-cloud-plugin.yml

@@ -1,13 +1,12 @@
 # code: language=ansible
-- name: Install and configure eWaterCycle Jupyter on jumphost
+- name: Configure workers
   hosts:
     - all
     - localhost
   gather_facts: false
-  vars:
-    # dCache token for mounting shared data
-    dcache_ro_token: null # Must be filled from command line
+  vars: {}
   tasks:
+    # Heavily inspired by https://github.com/RS-DAT/JupyterDaskOnSRC/blob/main/research-cloud-plugin.yml
     - name: Wait for system to become reachable
       wait_for_connection:
         timeout: 300
@@ -19,10 +18,42 @@
       debug:
         var: ansible_version
 
-    - name: Set up workers
-      debug:
-        var: worker_ip_addresses
+    - name: Alias name for jumphost
+      add_host:
+        name: jumphost
+        hostname: localhost
+
+    - name: Create group workers with workers' IPs
+      add_host:
+        name: "{{ item }}"
+        groups: workers
+        ansible_user: ubuntu
+        ansible_connection: ssh
+        ansible_ssh_private_key_file: ~/.ssh/id_rsa
+        ansible_become: yes
+      loop: '{{ worker_ip_addresses }}'
+
+    - name: Scan for worker keys.
+      shell:
+        cmd: ssh-keyscan {{ item }}
+      register: ssh_scan
+      loop: '{{ worker_ip_addresses }}'
+
+    - name: Write the worker keys to known hosts
+      known_hosts:
+        name: "{{ item.0.item }}"
+        key: "{{ item.1 }}"
+      loop:
+        "{{ ssh_scan.results | subelements('stdout_lines') | default([])}}"
 
+- name: Install and configure eWaterCycle Jupyter on jumphost
+  hosts:
+    - default
+    - jumphost
+  vars:
+    # dCache token for mounting shared data
+    dcache_ro_token: null # Must be filled from command line
+  tasks:
     - name: Common stuff
       include_role:
         name: common
@@ -73,15 +104,6 @@
       include_role:
         name: labstart
 
-    # https://explore.ewatercycle.org/ functionality
-    # - name: Experiment launcher
-    #   include_role:
-    #     name: launcher
-
-    # - name: Explorer
-    #   include_role:
-    #     name: terria
-
     # https://jupyter.ewatercycle.org/ functionality
     - name: Create eWaterCycle conda env
       include_role:
@@ -104,3 +126,23 @@
       debug:
         msg: The eWaterCycle Jupyter plugin has completed
 
+
+- name: Install Workers
+  hosts: workers
+  tasks:
+    - name: Hello
+      debug:
+        msg: "Hello from worker"
+
+
+# TODO for cluster setup
+# - Check if playbook works on vagrant, single src machine and src cluster
+# - Check sram users exist on workers with same uid
+# - Check network, workers should be able to reach jumphost but not accessable from internet
+# - Check storage, only jumphost should have home and cache storage items mounted
+# - Export homes and dcache and exchange dir over nfs to workers
+# - Install apptainer/conda/nbgrader/ewatercycle on workers
+# - use jupyterhub spawner to have each user end up on a different worker or jumphost
+#   - try ssh spawner with ssh keys+authorized_keys for each user on each machine
+#   - try slurm batch spawner with slurm installation
+#   - try dockerspawner.SystemUserSpawner with docker swarm installation