fix(replace): own service startup suppression

ansible-scylla-node/defaults/main.yml

@@ -167,6 +167,10 @@ cleanup_timeout_seconds: 36000
 # this should be set to false
 start_scylla_service: true
 
+# Let operation playbooks suppress automatic service startup even when reused
+# deployment parameters set start_scylla_service to true.
+skip_start_scylla_service: false
+
 # Only relevant for Debian/Ubuntu
 scylla_repo_keyserver: 'hkp://keyserver.ubuntu.com:80'
 scylla_repo_keys:

ansible-scylla-node/tasks/common.yml

@@ -395,7 +395,9 @@
         line: ""
       become: true
       when: token_distributor is defined
-  when: start_scylla_service is defined and start_scylla_service|bool
+  when:
+    - start_scylla_service is defined and start_scylla_service|bool
+    - not skip_start_scylla_service|bool
 
 - name: Copy system and table keys
   include_tasks: copy_system_and_table_keys.yml
@@ -424,7 +426,8 @@
     _keyspace_replication_strategy: "{{ system_auth_replication_strategy }}"
     _keyspace_rf: "{{ system_auth_rf }}"
   when:
-    - start_scylla_service|bool
+    - start_scylla_service is defined and start_scylla_service|bool
+    - not skip_start_scylla_service|bool
     - adjust_system_auth_replication is defined and adjust_system_auth_replication|bool
     - _authentication_enabled is defined and _authentication_enabled|bool
     - system_auth_rf is defined and system_auth_replication_strategy is defined
@@ -436,7 +439,8 @@
     _keyspace_replication_strategy: "{{ audit_replication_strategy }}"
     _keyspace_rf: "{{ audit_rf }}"
   when:
-    - start_scylla_service|bool
+    - start_scylla_service is defined and start_scylla_service|bool
+    - not skip_start_scylla_service|bool
     - adjust_audit_replication is defined and adjust_audit_replication|bool
     - _audit_enabled is defined and _audit_enabled|bool
     - audit_rf is defined and audit_replication_strategy is defined
@@ -448,7 +452,8 @@
     _keyspace_replication_strategy: "{{ system_distributed_replication_strategy }}"
     _keyspace_rf: "{{ system_distributed_rf }}"
   when:
-    - start_scylla_service|bool
+    - start_scylla_service is defined and start_scylla_service|bool
+    - not skip_start_scylla_service|bool
     - adjust_system_distributed_replication is defined and adjust_system_distributed_replication|bool
     - system_distributed_rf is defined and system_distributed_replication_strategy is defined
 
@@ -459,7 +464,8 @@
     _keyspace_replication_strategy: "{{ system_traces_replication_strategy }}"
     _keyspace_rf: "{{ system_traces_rf }}"
   when:
-    - start_scylla_service|bool
+    - start_scylla_service is defined and start_scylla_service|bool
+    - not skip_start_scylla_service|bool
     - adjust_system_traces_replication is defined and adjust_system_traces_replication|bool
     - system_traces_rf is defined and system_traces_replication_strategy is defined
 

ansible-scylla-node/tasks/manager_agents.yml

@@ -60,4 +60,7 @@
     state: restarted
     enabled: yes
   become: true
-  when: manager_agent_config_change.changed and start_scylla_service is defined and start_scylla_service|bool
+  when:
+    - manager_agent_config_change.changed
+    - start_scylla_service is defined and start_scylla_service|bool
+    - not skip_start_scylla_service|bool

example-playbooks/replace_node/README.md

@@ -7,7 +7,6 @@ This playbook will run the replace dead node procedure.
 * A scylla cluster with scylla-manager and scylla-monitoring installed.
 * The inventory file must be updated as described in the Usage section below, ie.: The new node should replace the dead node in the inventory.
 * It's necessary to have files with the same parameters used when the cluster was created, as described in the Parameters section.
-* It's very important to have `start_scylla_service` set to `false`. The replace will not work otherwise.
 
 ## Usage:
 
@@ -50,6 +49,8 @@ This playbook uses the node role to install and configure Scylla in the new node
 and use Scylla Manager to issue a repair if needed.
 So the same parameters for all three roles (node, monitoring and manager) that were
 used when the cluster was created should also be passed to the `replace_node.yml` playbook.
+The playbook suppresses automatic Scylla startup while applying the node role, regardless of the
+`start_scylla_service` value in those reused parameters.
 See the `Usage` section of this README for the way how these parameters are supposed to be passed.
 
 Besides the vars from the node role, this playbook has the following mandatory parameters: `replaced_node`, `replaced_node_broadcast_address` and `new_node`.

example-playbooks/replace_node/replace_node.yml

@@ -151,7 +151,9 @@
   vars:
     disable_firewall: false
   roles:
-    - ansible-scylla-node
+    - role: ansible-scylla-node
+      vars:
+        skip_start_scylla_service: true
 
 
 - name: Update scylla-monitoring

example-playbooks/replace_node/vars/main.yml

@@ -9,12 +9,6 @@
 # scylla_seeds:
 #   - "{{ groups['scylla'][0] }}"
 
-# This should always be set to False, since it'll be passed to the node role
-# and we don't want the node role to start the new node automatically.
-# If you're reusing (as expected) the parameters you used to deploy the other nodes,
-# make sure that you set this specific value to false
-start_scylla_service: false
-
 # By default, the playbook will repair the new node only if RBNO was not used
 # during the replacement.
 # If you set this to true, the repair will be skipped even in such case.