Initialize cli earlier during agent startup (#21188)

stelfrag · web-flow · commit 8ca14bcc361c · 2025-10-22T21:24:17.000+03:00
Initialize the cli command in two phases.
Support the ping command while the agent is initializing
Enable the full cli once the agent has finished initialization
diff --git a/src/daemon/commands.c b/src/daemon/commands.c
@@ -14,7 +14,7 @@ char cmd_prefix_by_status[] = {
         CMD_PREFIX_ERROR
 };
 
-static int command_server_initialized = 0;
+static cmd_init_status_t command_server_initialized = CMD_INIT_STATUS_OFF;
 static int command_thread_error;
 static int command_thread_shutdown;
 static unsigned clients = 0;
@@ -52,24 +52,24 @@ static cmd_status_t cmd_mark_stale_nodes_ephemeral(char *args, char **message);
 static cmd_status_t cmd_update_node_info(char *args, char **message);
 
 static command_info_t command_info_array[] = {
-    {"help", "", "Show this help menu.", cmd_help_execute, CMD_TYPE_HIGH_PRIORITY},                // show help menu
-    {"reload-health", "", "Reload health configuration.", cmd_reload_health_execute, CMD_TYPE_ORTHOGONAL}, // reload health configuration
-    {"reopen-logs", "", "Close and reopen log files.", cmd_reopen_logs_execute, CMD_TYPE_ORTHOGONAL},     // Close and reopen log files
-    {"shutdown-agent", "", "Cleanup and exit the netdata agent.", cmd_exit_execute, CMD_TYPE_EXCLUSIVE},          // exit cleanly
-    {"fatal-agent", "", "Log the state and halt the netdata agent.", cmd_fatal_execute, CMD_TYPE_HIGH_PRIORITY},        // exit with fatal error
-    {"reload-claiming-state", "", "Reload agent claiming state from disk.", cmd_reload_claiming_state_execute, CMD_TYPE_ORTHOGONAL}, // reload claiming state
-    {"reload-labels", "", "Reload all localhost labels.", cmd_reload_labels_execute, CMD_TYPE_ORTHOGONAL},                 // reload the labels
-    {"read-config", "", "", cmd_read_config_execute, CMD_TYPE_CONCURRENT},
-    {"write-config", "", "", cmd_write_config_execute, CMD_TYPE_ORTHOGONAL},
-    {"ping", "", "Return with 'pong' if agent is alive.", cmd_ping_execute, CMD_TYPE_ORTHOGONAL},
-    {"aclk-state", "[json]",  "Returns current state of ACLK and Netdata Cloud connection. (optionally in json).", cmd_aclk_state, CMD_TYPE_ORTHOGONAL},
-    {"version", "", "Returns the netdata version.", cmd_version, CMD_TYPE_ORTHOGONAL},
-    {"dumpconfig", "", "Returns the current netdata.conf on stdout.", cmd_dumpconfig, CMD_TYPE_ORTHOGONAL},
+    {"help", "", "Show this help menu.", cmd_help_execute, CMD_TYPE_HIGH_PRIORITY, CMD_INIT_STATUS_INIT},                // show help menu
+    {"reload-health", "", "Reload health configuration.", cmd_reload_health_execute, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL}, // reload health configuration
+    {"reopen-logs", "", "Close and reopen log files.", cmd_reopen_logs_execute, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},     // Close and reopen log files
+    {"shutdown-agent", "", "Cleanup and exit the netdata agent.", cmd_exit_execute, CMD_TYPE_EXCLUSIVE, CMD_INIT_STATUS_FULL},          // exit cleanly
+    {"fatal-agent", "", "Log the state and halt the netdata agent.", cmd_fatal_execute, CMD_TYPE_HIGH_PRIORITY, CMD_INIT_STATUS_FULL},        // exit with fatal error
+    {"reload-claiming-state", "", "Reload agent claiming state from disk.", cmd_reload_claiming_state_execute, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL}, // reload claiming state
+    {"reload-labels", "", "Reload all localhost labels.", cmd_reload_labels_execute, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},                 // reload the labels
+    {"read-config", "", "", cmd_read_config_execute, CMD_TYPE_CONCURRENT, CMD_INIT_STATUS_FULL},
+    {"write-config", "", "", cmd_write_config_execute, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},
+    {"ping", "", "Return with 'pong' if agent is alive.", cmd_ping_execute, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_INIT},                       // ping command
+    {"aclk-state", "[json]",  "Returns current state of ACLK and Netdata Cloud connection. (optionally in json).", cmd_aclk_state, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},
+    {"version", "", "Returns the netdata version.", cmd_version, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_INIT},
+    {"dumpconfig", "", "Returns the current netdata.conf on stdout.", cmd_dumpconfig, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},
     {"mark-stale-nodes-ephemeral", "<node_id | machine_guid | hostname | ALL_NODES>",
-        "Marks one or all disconnected nodes as ephemeral, while keeping their retention\n      available for queries on both this Netdata Agent dashboard and Netdata Cloud", cmd_mark_stale_nodes_ephemeral, CMD_TYPE_ORTHOGONAL},
+        "Marks one or all disconnected nodes as ephemeral, while keeping their retention\n      available for queries on both this Netdata Agent dashboard and Netdata Cloud", cmd_mark_stale_nodes_ephemeral, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},
     {"remove-stale-node", "<node_id | machine_guid | hostname | ALL_NODES>",
-     "Marks one or all disconnected nodes as ephemeral, and removes them\n      so that they are no longer available for queries, from both this\n      Netdata Agent dashboard and Netdata Cloud.", cmd_remove_stale_node, CMD_TYPE_ORTHOGONAL},
-    {"update-node-info", "", "Schedules an node update message for localhost to Netdata Cloud.", cmd_update_node_info, CMD_TYPE_ORTHOGONAL},
+     "Marks one or all disconnected nodes as ephemeral, and removes them\n      so that they are no longer available for queries, from both this\n      Netdata Agent dashboard and Netdata Cloud.", cmd_remove_stale_node, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},
+    {"update-node-info", "", "Schedules an node update message for localhost to Netdata Cloud.", cmd_update_node_info, CMD_TYPE_ORTHOGONAL, CMD_INIT_STATUS_FULL},
 };
 
 /* Mutexes for commands of type CMD_TYPE_ORTHOGONAL */
@@ -599,7 +599,12 @@ cmd_status_t execute_command(cmd_t idx, char *args, char **message)
     cmd_type_t type = command_info_array[idx].type;
 
     cmd_lock_by_type[type](idx);
-    status = command_info_array[idx].func(args, message);
+    if (command_server_initialized >= command_info_array[idx].init_status)
+        status = command_info_array[idx].func(args, message);
+    else {
+        *message = strdupz("Agent is initializing");
+        status = CMD_STATUS_SUCCESS;
+    }
     cmd_unlock_by_type[type](idx);
 
     return status;
@@ -851,10 +856,19 @@ void commands_init(void)
     int error;
 
     sanity_check();
-    if (command_server_initialized)
+    if (command_server_initialized == CMD_INIT_STATUS_FULL)
         return;
 
-    netdata_log_info("Initializing command server.");
+    if (command_server_initialized == CMD_INIT_STATUS_OFF) {
+        netdata_log_info("Initializing command server for liveness CHECK");
+        command_server_initialized = CMD_INIT_STATUS_INIT;
+    }
+    else {
+        netdata_log_info("Initializing full command server.");
+        command_server_initialized = CMD_INIT_STATUS_FULL;
+        return;
+    }
+
     for (i = 0 ; i < CMD_TOTAL_COMMANDS ; ++i) {
         fatal_assert(0 == netdata_mutex_init(&command_lock_array[i]));
     }
@@ -878,7 +892,6 @@ void commands_init(void)
         goto after_error;
     }
 
-    command_server_initialized = 1;
     return;
 
 after_error:
@@ -889,7 +902,7 @@ void commands_exit(void)
 {
     cmd_t i;
 
-    if (!command_server_initialized)
+    if (command_server_initialized == CMD_INIT_STATUS_OFF)
         return;
 
     command_thread_shutdown = 1;
@@ -903,5 +916,5 @@ void commands_exit(void)
     }
     netdata_rwlock_destroy(&exclusive_rwlock);
     netdata_log_info("Command server has stopped.");
-    command_server_initialized = 0;
+    command_server_initialized = CMD_INIT_STATUS_OFF;
 }
diff --git a/src/daemon/commands.h b/src/daemon/commands.h
@@ -65,12 +65,19 @@ typedef enum cmd_type {
  */
 typedef cmd_status_t (command_action_t) (char *args, char **message);
 
+typedef enum cmd_init_status {
+    CMD_INIT_STATUS_OFF,
+    CMD_INIT_STATUS_INIT,
+    CMD_INIT_STATUS_FULL,
+} cmd_init_status_t;
+
 typedef struct command_info {
     char *cmd_str;              // the command string
     char *params;
     char *help;
     command_action_t *func;     // the function that executes the command
     cmd_type_t type;            // Concurrency control information for the command
+    cmd_init_status_t init_status;     // command availability during start
 } command_info_t;
 
 typedef void (command_lock_t) (unsigned index);
diff --git a/src/daemon/main.c b/src/daemon/main.c
@@ -1055,6 +1055,10 @@ int netdata_main(int argc, char **argv) {
     // ----------------------------------------------------------------------------------------------------------------
     delta_startup_time("RRD structures");
 
+    delta_startup_time("commands liveness support");
+
+    commands_init();
+
     abort_on_fatal_disable();
     if (rrd_init(netdata_configured_hostname, system_info, false))
         fatal("Cannot initialize localhost instance with name '%s'.", netdata_configured_hostname);
@@ -1098,7 +1102,7 @@ int netdata_main(int argc, char **argv) {
     ml_start_threads();
 
     // ----------------------------------------------------------------------------------------------------------------
-    delta_startup_time("commands API");
+    delta_startup_time("commands full API");
 
     commands_init();
 
