[cluster_run] Add code watchdog

$ ./cluster_run -w

(modify erl files)

$ make

(files get reloaded automatically)

Change-Id: I10290692dd6c911fff9fd4b0e1d78ced60db1095
Reviewed-on: https://review.couchbase.org/c/ns_server/+/225572
Reviewed-by: Bryan McCoid <[email protected]>
Tested-by: Timofey Barmin <[email protected]>
Tested-by: Build Bot <[email protected]>
Well-Formed: Build Bot <[email protected]>

Author: timofey-barmin

cluster_run

@@ -166,6 +166,11 @@ def argument_parser():
         nargs="*",
         help=argparse.SUPPRESS,
     )
+    arg_parser.add_argument(
+        '--code-watchdog', '-w',
+        action='store_true',
+        default=False,
+        help="Reload ns_server beam files automatically when they are modified")
 
     args = arg_parser.parse_args()
 
@@ -189,8 +194,8 @@ def argument_parser():
               "run_serverless": args.serverless,
               "run_provisioned": args.provisioned,
               "num_vbuckets": args.num_vbuckets,
-              "args": args.additional_args
-              }
+              "args": args.additional_args,
+              "code_watchdog": args.code_watchdog}
 
     # Removes the keys when value is None, therefore start_cluster uses
     # the default values.
@@ -204,6 +209,8 @@ def main():
     params = argument_parser()
     nodes = []
     terminal_attrs = None
+    code_watchdog = params.get("code_watchdog", False)
+    del params["code_watchdog"]
 
     def kill_nodes():
         cluster_run_lib.kill_nodes(nodes, terminal_attrs)
@@ -218,9 +225,17 @@ def main():
 
     nodes = cluster_run_lib.start_cluster(**params)
 
-    for node in nodes:
-        node.wait()
+    if code_watchdog:
+        observer = cluster_run_lib.start_code_watchdog(
+                       params.get("num_nodes", 1),
+                       params.get("start_index", 0))
 
+    try:
+        for node in nodes:
+            node.wait()
+    finally:
+        if code_watchdog:
+            cluster_run_lib.stop_code_watchdog(observer)
 
 if __name__ == '__main__':
     main()

pylib/cluster_run_lib.py

@@ -513,6 +513,69 @@ def start_node(node_num):
     return processes
 
 
+def start_code_watchdog(num_nodes, start_index):
+    # This is a workaround to avoid importing watchdog if it's not installed.
+    # Otherwise, we will need to install watchdog on all test machines, where
+    # it will never be actually used.
+    from watchdog.observers import Observer
+    from watchdog.events import FileSystemEventHandler
+
+    class BeamFileHandler(FileSystemEventHandler):
+        def __init__(self, num_nodes, start_index):
+            self.num_nodes = num_nodes
+            self.start_index = start_index
+
+        def on_modified(self, event):
+            self.handle_file_change(event.src_path)
+
+        def on_created(self, event):
+            self.handle_file_change(event.src_path)
+
+        def on_moved(self, event):
+            self.handle_file_change(event.dest_path)
+
+        def handle_file_change(self, file_path):
+            if file_path.endswith('.beam'):
+                module_name = os.path.splitext(os.path.basename(file_path))[0]
+                self.reload_module(module_name)
+
+        def reload_module(self, module_name):
+            for i in range(self.num_nodes):
+                port = base_api_port + self.start_index + i
+                node = f"http://127.0.0.1:{port}"
+                url = f"{node}/diag/eval"
+
+                # Erlang code to reload the module
+                erlang_code = f"{{module, _}} = c:l({module_name})."
+
+                try:
+                    response = requests.post(url,
+                                            auth=(default_username, default_pass),
+                                            data=erlang_code)
+                    if response.status_code == 200:
+                        print(f"*** reloaded {module_name} on {node}")
+                    else:
+                        print(f"*** reload failed for {module_name} on {node}:" \
+                            f" {response.text}")
+                except Exception as e:
+                    print(f"*** reload failed for {module_name} on {node}: {e}")
+
+    code_path = os.path.join(ns_server_dir, '_build', 'default', 'lib')
+    print(f"*** Starting code watchdog for {code_path}")
+    # Create a watchdog observer
+    observer = Observer()
+    handler = BeamFileHandler(num_nodes, start_index)
+    observer.schedule(handler, code_path, recursive=True)
+    observer.start()
+    return observer
+
+
+def stop_code_watchdog(observer):
+    observer.stop()
+    observer.join()
+    print("*** Code watchdog stopped")
+
+
 def wait_nodes_up(num_nodes=1, start_index=0, master_passwords=None,
                   root_dir=None, timeout_s=node_start_timeout_s,
                   node_urls=None, verbose=True):

requirements.txt

@@ -1 +1,2 @@
 requests>=2.32.3
+watchdog>=6.0.0