feat: idempotency for node deployment

The introduction of the node manager broke the idempotent nature of the deployment process.

Now, when the genesis node is being created, it will only be created if there's not already an
existing genesis node. With respect to the remaining nodes, they use a count value to determine
whether any more need to be added. If the count is satisfied, the task to add the nodes will be
skipped.

The commit also introduced an argument to supply a custom build of the node manager using a URL.
This was useful for testing out experimental features of the node manager.

Cargo.toml

@@ -35,6 +35,7 @@ tar = "0.4"
 tempfile = "3.8.0"
 tokio = { version = "1.26", features = ["full"] }
 tokio-stream = "0.1.14"
+url = "2.5.0"
 
 [dev-dependencies]
 httpmock = "0.6"

resources/ansible/roles/node/tasks/main.yml

@@ -23,8 +23,27 @@
     dest: "{{ node_manager_archive_dest_path }}"
     remote_src: True
 
+#
+# Setup genesis node on genesis run
+#
+- name: check if genesis node is already set up
+  become: True
+  ansible.builtin.command: safenode-manager status --json
+  register: genesis_status
+  when: is_genesis
+
+- name: parse genesis node status
+  set_fact:
+    genesis_exists: "{{ (genesis_status.stdout | from_json | selectattr('genesis', 'equalto', true) | list | length) > 0 }}"
+  when: is_genesis and genesis_status.stdout != ""
+
 - name: add genesis node service
+  become: True
   ansible.builtin.command:
+    # The `omit` filter is used to remove the `--version` argument if the
+    # `node_version` variable is not defined. If the argv would be otherwise
+    # defined as '', the command would fail as it doesn't expect an empty
+    # argument, so it's omitted instead.
     argv: "{{ command_args | reject('equalto', omit) | list }}"
   vars:
     command_args:
@@ -34,23 +53,38 @@
       - --port={{ genesis_port }}
       - --rpc-port={{ genesis_rpc_port }}
       - "{{ ('--url=' + node_archive_url) if node_archive_url is defined else omit }}"
-  when: is_genesis
+  when: is_genesis and not genesis_exists | default(false)
+
+#
+# Setup remaining nodes on non-genesis run
+#
+- name: check current number of node services
+  become: True
+  ansible.builtin.command: safenode-manager status --json
+  register: node_status
+
+- name: calculate current non-genesis node count
+  set_fact:
+    current_node_count: "{{ (node_status.stdout | from_json | selectattr('genesis', 'equalto', false) | list | length) }}"
+  when: node_status.stdout != ""
+
+- name: calculate number of nodes to add
+  set_fact:
+    nodes_to_add: "{{ node_instance_count | int - current_node_count | int }}"
+  when: current_node_count is defined
 
 - name: add node services
+  become: True
   ansible.builtin.command:
-    # The `omit` filter is used to remove the `--version` argument if the
-    # `node_version` variable is not defined. If the argv would be otherwise
-    # defined as '', the command would fail as it doesn't expect an empty
-    # argument, so it's omitted instead.
     argv: "{{ command_args | reject('equalto', omit) | list }}"
   vars:
     command_args:
       - "{{ node_manager_archive_dest_path }}/safenode-manager"
       - add
-      - "--count={{ node_instance_count }}"
+      - "--count={{ nodes_to_add }}"
       - "--peer={{ genesis_multiaddr }}"
       - "{{ ('--url=' + node_archive_url) if node_archive_url is defined else omit }}"
-  when: not is_genesis
+  when: not is_genesis and nodes_to_add | default(0) | int > 0
 
 - name: start the node services
   become: True

src/deploy.rs

@@ -9,6 +9,7 @@ use crate::{
     print_duration, SnCodebaseType, TestnetDeploy,
 };
 use std::{net::SocketAddr, path::PathBuf, time::Instant};
+use url::Url;
 
 pub struct DeployCmd {
     testnet_deploy: TestnetDeploy,
@@ -17,6 +18,7 @@ pub struct DeployCmd {
     vm_count: u16,
     logstash_details: (String, Vec<SocketAddr>),
     sn_codebase_type: SnCodebaseType,
+    node_manager_url: Option<Url>,
 }
 
 impl DeployCmd {
@@ -28,6 +30,7 @@ impl DeployCmd {
         vm_count: u16,
         logstash_details: (String, Vec<SocketAddr>),
         sn_codebase_type: SnCodebaseType,
+        node_manager_url: Option<Url>,
     ) -> Self {
         Self {
             testnet_deploy,
@@ -36,6 +39,7 @@ impl DeployCmd {
             vm_count,
             logstash_details,
             sn_codebase_type,
+            node_manager_url,
         }
     }
 
@@ -366,6 +370,14 @@ impl DeployCmd {
         };
         Self::add_value(&mut extra_vars, "node_archive_url", &node_archive_url);
 
+        if let Some(node_manager_url) = self.node_manager_url.as_ref() {
+            Self::add_value(
+                &mut extra_vars,
+                "node_manager_archive_url",
+                node_manager_url.as_ref(),
+            );
+        }
+
         let (logstash_stack_name, logstash_hosts) = &self.logstash_details;
         Self::add_value(&mut extra_vars, "logstash_stack_name", logstash_stack_name);
         extra_vars.push_str("\"logstash_hosts\": [");

src/main.rs

@@ -14,6 +14,7 @@ use sn_testnet_deploy::{
     setup::setup_dotenv_file, CloudProvider, DeploymentInventory, SnCodebaseType,
     TestnetDeployBuilder,
 };
+use url::Url;
 
 pub fn parse_provider(val: &str) -> Result<CloudProvider> {
     match val {
@@ -109,6 +110,11 @@ enum Commands {
         /// --branch and --repo-owner arguments. You can only supply version numbers or a custom
         /// branch, not both.
         safenode_version: Option<String>,
+        /// Optionally supply the URL of a gzipped tar archive containing a node manager binary.
+        ///
+        /// Useful for testing experimental versions of the node manager.
+        #[arg(long)]
+        node_manager_url: Option<String>,
     },
     Inventory {
         /// The name of the environment
@@ -303,7 +309,7 @@ async fn main() -> Result<()> {
             logstash_stack_name,
             safe_version,
             safenode_version,
-            ..
+            node_manager_url,
         }) => {
             let sn_codebase_type = get_sn_codebase_type(
                 branch,
@@ -343,13 +349,19 @@ async fn main() -> Result<()> {
             let stack_hosts = logstash_deploy
                 .get_stack_hosts(&logstash_stack_name)
                 .await?;
+            let node_manager_url = if let Some(node_manager_url) = node_manager_url {
+                Some(Url::parse(&node_manager_url)?)
+            } else {
+                None
+            };
             let deploy_cmd = DeployCmd::new(
                 testnet_deploy,
                 name,
                 node_count,
                 vm_count,
                 (logstash_stack_name, stack_hosts),
                 sn_codebase_type,
+                node_manager_url,
             );
             deploy_cmd.execute().await?;
             Ok(())