Feature/refactor mongo

environments/template/group_vars/mongo_servers.yml

@@ -1,5 +1,5 @@
 ---
-replica_set_name: my_mongo_cluster
+mongo_replica_set_name: my_mongo_cluster
 
 mongo_cluster_members:
   - host: "mongo3.example.com:{{ mongo_port }}" # arbiter first or change mongo_arbiter_index

environments/template/secrets/secret_example.yml

@@ -13,7 +13,7 @@ mongo_passwords:
    oidcng: secret
    myconext: secret
 
-mongo_admin_password: secret
+mongo_admin_password: secret # this works for first time install, if you change it later you will have to do it manually
 mongo_ca_passphrase: secret
 
 engine_api_metadata_push_password: secret

roles/mongo/README.md

@@ -14,6 +14,31 @@ Set the mongo_cluster_private_key variable encrypted in host_vars
 
 Please review the official Mongo documentation for more information.
 
+# Mongo deployment
+
+To avoid surprisesyou can enable or disable cluster configuration with the boolean option mongo_configure_cluster. The role willonly initiate or reconfigure cluster if this is true (safest option is to use -e mongo_configure_cluster=true with your deployment when cluster configuration is necessary).
+Another issue is the serial value, it is safest to set it to 1 in your playbook, if it is higher multiple mongo nodes can will be restarted at once and it can break your cluster. However when you want to intialise a new cluster you need to run the tasks parallel and serial needs to be as high as the amount of nodes. We handled this with a variable serial with the name serial_number in our playbook with a default 1. If cluster initialisation or reconfiguration is necessary use -e "serial_number=<AMOUNT_OF_CLUSTERMEMBERS>"
+
+
+See also https://docs.ansible.com/projects/ansible/latest/playbook_guide/playbooks_strategies.html#setting-the-batch-size-with-serial
+
+# Cluster reconfiguration
+
+Warning: the cluster reconfiguration option in the mongodb_replicationset module is experimental. and you can only add or remove one node at a time.
+
 # Todo
-- [ ] Add the possibility for adding and removing cluster members
-- [ ] Add the possibility for a standalone mongo server
+- [x] Check mongo_replication_roles and give a clear fail message when not set
+- [ ] Add option to change the already existing admin user, for now change the password manually and change it in the ansible config accordingly
+- [x] Add the possibility for adding and removing cluster members
+- [x] Add the possibility for a standalone mongo server
+- [x] Cluster changes can be enabled or disabled
+- [ ] Reconfigure cluster always reports changed
+- [ ] Initialise cluster always reports changed
+- [ ] check mode for writeconcern change tasks does not report change () same for any other mongodb_shell task "remote module (community.mongodb.mongodb_shell) does not support check mode"}
+- [X] Clearer error messaging for even number of votes
+- [X] Role refuses to add users when a new cluster is built (3 nodes) (cannot add users on a broken cluster)
+- [X] it would be helpfull if role (for example primary) is not defined in host_vars but in the mongo_cluster_members array
+- [X] removing primary from the cluster will not work but the error is unclear, this is related to the todo above
+- [ ] is it necessary to make votes configurable?
+- [X] preflight check are cluster members in the inventory and monog_servers group
+- [ ] Standalone mongo also requires cluster certificates, not logical although it doens't hurt

roles/mongo/defaults/main.yml

@@ -13,35 +13,37 @@ mongo_servers: [] # Set this in group_vars
 # Not all mongo servers in the inventory are cluster members, so we use a separate list for this.
 # Set this in group_vars of your environment(s). The arbiter should go first, or change the mongo_arbiter_index.
 # mongo_cluster_members:
-#   - host: "mongoarbiter.example.com:27017"
+#   - host: "mongoarbiter.example.com"
 #     priority: 1 # can vote, cannot become primary
-#   - host: "mongo2.example.com:27017"
+#     port: 27017
+#   - host: "mongo2.example.com"
 #     priority: 2
-#   - host: "mongo1.example.com:27017"
+#     port: 27017
+#   - host: "mongo1.example.com"
 #     priority: 3
-# mongo_arbiter_index: 0
-
-# The replication role
-# mongo_replication_role: # Set this in host_vars, it can have the values: "primary", "secondary" or arbiter
+#     port: 27017
 
 # Todo: there is a link between mongo_replication_role and priority (arbiter is priority 1, primary the highest) so
 # setting them separately is not ideal.
 
 # The port for mongo server
-mongod_port: 27017
+mongo_port: 27017
 
 # The password for admin user
-mongo_admin_pass: "{{ mongo_admin_password }}" # Set this in secrets
+# mongo_admin_password: # set this in secrets
+
+# Are we using a cluster?
+mongo_mode: "cluster" # cluster or standalone
 
 # The name of the replication set
-replica_set_name: "{{ instance_name }}" # Set this in group_vars
+mongo_replica_set_name: "{{ instance_name }}" # Set this in group_vars
 
 # Add a database
 mongo:
   users:
-    - { name: managerw, db_name: metadata, password: "{{ mongo_passwords.manage }}" }
-    - { name: oidcsrw, db_name: oidc, password: "{{ mongo_passwords.oidcng }}" }
-    - { name: myconextrw, db_name: myconext, password: "{{ mongo_passwords.myconext }}" }
+    - { name: managerw, db_name: metadata, password: "{{ mongo_passwords.manage }}", role: "readWrite" }
+    - { name: oidcsrw, db_name: oidc, password: "{{ mongo_passwords.oidcng }}", role: "readWrite"}
+    - { name: myconextrw, db_name: myconext, password: "{{ mongo_passwords.myconext }}", role: "readWrite" }
 
 # Listen on all addresses by default
 mongo_bind_listen_address: "0.0.0.0"
@@ -53,3 +55,31 @@ mongo_pki_dir: "/etc/pki/mongo"
 
 # Users and groups
 mongo_group: "mongod"
+
+# Paths
+mongo_config_file: "/etc/mongod.conf"
+mongo_data_path: "/var/lib/mongo"
+mongo_pymongo_version: 4.16.0
+
+# cluster members
+# set in group_vars
+# mongo_cluster_members:
+#   - host: mongo1.example.com
+#     priority: 3
+#     votes: 1
+#     port: 27017
+#   - host: mongo2.example.com
+#     priority: 2
+#     votes: 1
+#     port: 27017
+#   - host: mongo3.example.com
+#     priority: 1
+#     votes: 1
+#     port: 27017
+#     arbiterOnly: true
+
+mongo_cluster_write_concern: "majority"
+mongo_cluster_write_timeout: 5000
+
+# to avoid surprises only initiate or reconfigure cluster if this is true (safest option is to use -e mongo_configure_cluster=true with your deployment when cluster configuration is necessary)
+mongo_configure_cluster: false

roles/mongo/tasks/clusterconfig.yml

@@ -1,60 +1,191 @@
 ---
-# todo this weorks only for new deployments
-# rewrite so mongo config can be changed and cluster members can be added or removed
-- name: Check if hosts are in clustered
-  ansible.builtin.command: mongosh --port {{ mongod_port }} --quiet --eval 'db.isMaster().hosts'
-  register: check_cluster
-  changed_when: false
-  check_mode: false
-
-- name: Debug check_cluster variable
+# In this task file the cluster is configured
+
+# priority moet matchen met replication role, of replication role uit cluster mebers halen?
+# todo write concern zetten
+
+# Do some preflight checks
+- name: Check some cluster related variables
+  when: mongo_mode == "cluster"
+  block:
+    - name: Fail on undefined mongo_replica_set_name
+      when: mongo_replica_set_name is not defined
+      ansible.builtin.fail:
+        msg: "Something is wrong, mongo_mode was set to cluster but mongo_replica_set_name is undefined."
+
+- name: Debug replica settings
   ansible.builtin.debug:
-    msg: "{{ check_cluster }}"
+    msg: "Replica set name {{ mongo_replica_set_name }}"
     verbosity: 2
 
-- name: Debug mongo_cluster_members variable
+# Loop over cluster members and check their presence in mong_servers group and their mode (not standalone)
+
+- name: Check if mongo_cluster_members exist in inventory group
+  ansible.builtin.assert:
+    that:
+      - item.host in groups['mongo_servers']
+    fail_msg: "Server '{{ item.host }}' is not in the mongo_servers inventory group"
+    success_msg: "Server '{{ item.host }}' found in mongo_servers inventory group"
+  run_once: true
+  loop: "{{ mongo_cluster_members }}"
+
+# Loop over cluster members and check for primary
+
+- name: Set primary host fact
+  ansible.builtin.set_fact:
+    mongo_primary_host: "{{ (mongo_cluster_members | max(attribute='priority')).host }}"
+
+- name: Debug primary settings
   ansible.builtin.debug:
-    msg: "{{ mongo_cluster_members }}"
+    msg: "Primary is {{ mongo_primary_host }}"
     verbosity: 2
 
-- name: Debug mongo_replication_role variable
+# What is the replication role of the current host
+- name: Debug replication role settings
   ansible.builtin.debug:
-    msg: "{{ mongo_replication_role }}"
+    msg: "This nodes replication role is {{ mongo_replication_role }}"
     verbosity: 2
 
-- name: Initial cluster initialisation
-  community.mongodb.mongodb_replicaset:
-    login_host: localhost
-    login_user: admin
-    login_port: "{{ mongod_port }}"
-    login_password: "{{ mongo_admin_password }}"
-    replica_set: "{{ replica_set_name }}"
-    members: "{{ mongo_cluster_members }}"
-    arbiter_at_index: "{{ mongo_arbiter_index | default(0) }}"
-    validate: false
-  run_once: true
-  when: mongo_replication_role == 'primary'
+# Cannot initialise a cluster without starting.......
+- name: Enable and start mongod
+  ansible.builtin.service:
+    name: mongod.service
+    enabled: true
+    state: started
 
-- name: Wait until cluster health is ok
-  community.mongodb.mongodb_status:
-    login_user: admin
-    login_password: "{{ mongo_admin_password }}"
-    login_database: admin
-    login_port: "{{ mongod_port }}"
-    validate: default
-    poll: 5
-    interval: 12
-    replica_set: "{{ replica_set_name }}"
+# Initialise cluster block
+- name: Initialise or reconfigure cluster block
   when: mongo_replication_role == 'primary'
+  block:
+    - name: Check if replica set is already initialised
+      community.mongodb.mongodb_shell:
+        login_host: localhost
+        login_user: admin
+        login_port: "{{ mongo_port }}"
+        login_password: "{{ mongo_admin_password }}"
+        eval: "rs.status().ok"
+        db: admin
+      register: rs_already_init
+      ignore_errors: true
 
-- name: Add the admin user
-  community.mongodb.mongodb_user:
-    database: admin
-    name: admin
-    password: "{{ mongo_admin_password }}"
-    login_port: "{{ mongod_port }}"
-    roles: root
-    state: present
-  when: check_cluster.stdout == ""
-  no_log: true
-  run_once: true
+    - name: Debug cluster initialization check
+      ansible.builtin.debug:
+        msg: "{{ rs_already_init }}"
+        verbosity: 2
+
+    # This should be possible with community.mongodb.mongodb_replicaset
+    # But we keep getting authenticatione error so leave it like this for now
+    - name: Initialise replica set if necessary
+      community.mongodb.mongodb_shell:
+        login_host: localhost
+        login_user: admin
+        login_port: "{{ mongo_port }}"
+        login_password: "{{ mongo_admin_password }}"
+        eval: |
+          rs.initiate({
+            _id: "{{ mongo_replica_set_name }}",
+            members: [
+              {% for m in mongo_cluster_members %}
+              { _id: {{ loop.index0 }}, host: "{{ m.host }}:{{ m.port }}", priority: {{ m.priority }}, votes: {{ m.votes }}{% if m.arbiterOnly is defined and m.arbiterOnly and m.arbiterOnly == true %}, arbiterOnly: true {% endif %} }{{ "," if not loop.last else "" }}
+              {% endfor %}
+            ]
+          })
+        db: admin
+      when: rs_already_init.failed
+      register: rs_init
+
+    - name: Debug cluster initialization
+      ansible.builtin.debug:
+        msg: "{{ rs_init }}"
+        verbosity: 2
+
+    - name: Format members list
+      ansible.builtin.set_fact:
+        mongo_cluster_members_formatted: "{{ mongo_cluster_members_formatted | default([]) + [m | combine({'host': m.host ~ ':' ~ (m.port | string)}) | dict2items | rejectattr('key', 'eq', 'port') | list | items2dict] }}"
+      loop: "{{ mongo_cluster_members }}"
+      loop_control:
+        loop_var: m
+
+    - name: Debug members list
+      ansible.builtin.debug:
+        msg: "{{ mongo_cluster_members }}"
+        verbosity: 2
+
+    - name: Debug formatted members list
+      ansible.builtin.debug:
+        msg: "{{ mongo_cluster_members_formatted }}"
+        verbosity: 2
+
+    # Reconfigure cluster
+    # todo: this always returns changed even when nothing changes
+    - name: Reconfigure cluster if necessary
+      community.mongodb.mongodb_replicaset:
+        login_host: localhost
+        login_user: admin
+        login_password: "{{ mongo_admin_password }}"
+        login_port: "{{ mongo_port }}"
+        reconfigure: true
+        replica_set: "{{ mongo_replica_set_name }}"
+        members: "{{ mongo_cluster_members_formatted }}"
+      register: rs_reconfigure
+
+    - name: Debug cluster reconfiguration
+      ansible.builtin.debug:
+        msg: "{{ rs_reconfigure }}"
+        verbosity: 2
+
+    - name: Wait for the replicaset to stabilise
+      community.mongodb.mongodb_status:
+        replica_set: "{{ mongo_replica_set_name }}"
+        login_host: localhost
+        login_user: admin
+        login_password: "{{ mongo_admin_password }}"
+        login_port: "{{ mongo_port }}"
+        poll: 5
+        interval: 30
+        validate: minimal # default fails on even number of servers and although this is not a great situation, it is sometimes the temporary situation because we can onlye add or remove 1 node at a time
+
+    # Cluster settings that cannot be changed with mongodb_replicaset
+
+    - name: Get current default write concern
+      community.mongodb.mongodb_shell:
+        login_host: localhost
+        login_port: 27017
+        login_user: admin
+        login_password: "{{ mongo_admin_password }}"
+        eval: "db.adminCommand({ getDefaultRWConcern: 1 })"
+      register: current_write_concern
+      changed_when: false
+
+    - name: Debug write concern check
+      ansible.builtin.debug:
+        msg: "{{ current_write_concern.transformed_output.defaultWriteConcern }}"
+        verbosity: 2
+      when: current_write_concern.transformed_output.defaultWriteConcern is defined
+
+    - name: Set default write concern
+      when: >
+          current_write_concern.transformed_output.defaultWriteConcern is defined
+          and
+          (current_write_concern.transformed_output.defaultWriteConcern.w | string != mongo_cluster_write_concern | default('majority') | string
+          or
+          current_write_concern.transformed_output.defaultWriteConcern.wtimeout | int != mongo_cluster_write_timeout | default(5000) | int)
+          or current_write_concern.transformed_output.defaultWriteConcern is not defined
+      block:
+        - name: "set write concern majority"
+          when: mongo_cluster_write_concern == "majority"
+          community.mongodb.mongodb_shell:
+            login_host: localhost
+            login_user: admin
+            login_password: "{{ mongo_admin_password }}"
+            login_port: "{{ mongo_port }}"
+            eval: "db.adminCommand({ setDefaultRWConcern: 1, defaultWriteConcern: { w: \"{{ mongo_cluster_write_concern | default('majority') }}\", wtimeout: {{ mongo_cluster_write_timeout | default(5000) }} } })"
+        # could not get this to work with either majority with quotes or number without quotes so for now an ugly fix
+        - name: "set write concern numeric"
+          when: mongo_cluster_write_concern != "majority"
+          community.mongodb.mongodb_shell:
+            login_host: localhost
+            login_user: admin
+            login_password: "{{ mongo_admin_password }}"
+            login_port: "{{ mongo_port }}"
+            eval: "db.adminCommand({ setDefaultRWConcern: 1, defaultWriteConcern: { w: {{ mongo_cluster_write_concern | default('majority') }}, wtimeout: {{ mongo_cluster_write_timeout | default(5000) }} } })"