manager: files: watchdog: added RPZ files

alesmrazek · alesmrazek · commit 7016be52619e · 2025-03-14T13:43:40.000+01:00
Separate timer for each command.
diff --git a/NEWS b/NEWS
@@ -5,6 +5,10 @@ Bugfixes
 --------
 - /management/unix-socket: revert to absolute path (#926, !1664)
 
+Improvements
+------------
+- /local-data/rpz/*/watchdog: new configuration to enable watchdog for RPZ files (!1665)
+
 
 Knot Resolver 6.0.11 (2025-02-26)
 =================================
diff --git a/python/knot_resolver/manager/files/watchdog.py b/python/knot_resolver/manager/files/watchdog.py
@@ -2,27 +2,29 @@
 from pathlib import Path
 from threading import Timer
 from typing import Any, Dict, List, Optional
+from urllib.parse import quote
 
 from knot_resolver.constants import WATCHDOG_LIB
 from knot_resolver.controller.registered_workers import command_registered_workers
 from knot_resolver.datamodel import KresConfig
 from knot_resolver.manager.config_store import ConfigStore, only_on_real_changes_update
 from knot_resolver.utils import compat
+from knot_resolver.utils.requests import SocketDesc, request
 
 logger = logging.getLogger(__name__)
 
+FilesToWatch = Dict[Path, Optional[str]]
 
-def tls_cert_files_config(config: KresConfig) -> List[Any]:
+
+def watched_files_config(config: KresConfig) -> List[Any]:
     return [
         config.network.tls.files_watchdog,
         config.network.tls.cert_file,
         config.network.tls.key_file,
+        config.local_data.rpz,
     ]
 
 
-FilesToWatch = Dict[Path, str]
-
-
 if WATCHDOG_LIB:
     from watchdog.events import (
         FileSystemEvent,
@@ -31,58 +33,96 @@ def tls_cert_files_config(config: KresConfig) -> List[Any]:
     from watchdog.observers import Observer
 
     class FilesWatchdogEventHandler(FileSystemEventHandler):
-        def __init__(self, files: FilesToWatch) -> None:
+        def __init__(self, files: FilesToWatch, config: KresConfig) -> None:
             self._files = files
-            self._timer: Optional[Timer] = None
+            self._config = config
+            self._policy_timer: Optional[Timer] = None
+            self._timers: Dict[str, Timer] = {}
+
+        def _trigger(self, cmd: Optional[str]) -> None:
+            def policy_reload() -> None:
+                management = self._config.management
+                socket = SocketDesc(
+                    f'http+unix://{quote(str(management.unix_socket), safe="")}/',
+                    'Key "/management/unix-socket" in validated configuration',
+                )
+                if management.interface:
+                    socket = SocketDesc(
+                        f"http://{management.interface.addr}:{management.interface.port}",
+                        'Key "/management/interface" in validated configuration',
+                    )
+
+                response = request(socket, "POST", "renew")
+                if response.status != 200:
+                    logger.error(f"Failed to reload policy rules: {response.body}")
+                logger.info("Reloading policy rules has finished")
+
+            if not cmd:
+                # skipping if reload was already triggered
+                if self._policy_timer and self._policy_timer.is_alive():
+                    logger.info("Skipping reloading policy rules, it was already triggered")
+                    return
+                # start a 5sec timer
+                logger.info("Delayed policy rules reload has started")
+                self._policy_timer = Timer(5, policy_reload)
+                self._policy_timer.start()
+                return
 
-        def _reload(self, cmd: str) -> None:
             def command() -> None:
                 if compat.asyncio.is_event_loop_running():
                     compat.asyncio.create_task(command_registered_workers(cmd))
                 else:
                     compat.asyncio.run(command_registered_workers(cmd))
-                logger.info("Reloading of TLS certificate files has finished")
+                logger.info(f"Sending '{cmd}' command to reload watched files has finished")
 
-            # skipping if reload was already triggered
-            if self._timer and self._timer.is_alive():
-                logger.info("Skipping TLS certificate files reloading, reload command was already triggered")
+            # skipping if command was already triggered
+            if cmd in self._timers and self._timers[cmd].is_alive():
+                logger.info(f"Skipping sending '{cmd}' command, it was already triggered")
                 return
             # start a 5sec timer
-            logger.info("Delayed reload of TLS certificate files has started")
-            self._timer = Timer(5, command)
-            self._timer.start()
+            logger.info(f"Delayed send of '{cmd}' command has started")
+            self._timers[cmd] = Timer(5, command)
+            self._timers[cmd].start()
 
         def on_created(self, event: FileSystemEvent) -> None:
             src_path = Path(str(event.src_path))
             if src_path in self._files.keys():
                 logger.info(f"Watched file '{src_path}' has been created")
-                self._reload(self._files[src_path])
+                self._trigger(self._files[src_path])
 
         def on_deleted(self, event: FileSystemEvent) -> None:
             src_path = Path(str(event.src_path))
             if src_path in self._files.keys():
                 logger.warning(f"Watched file '{src_path}' has been deleted")
-                if self._timer:
-                    self._timer.cancel()
+                cmd = self._files[src_path]
+                if cmd in self._timers:
+                    self._timers[cmd].cancel()
             for file in self._files.keys():
                 if file.parent == src_path:
                     logger.warning(f"Watched directory '{src_path}' has been deleted")
-                    if self._timer:
-                        self._timer.cancel()
+                    cmd = self._files[file]
+                    if cmd in self._timers:
+                        self._timers[cmd].cancel()
+
+        def on_moved(self, event: FileSystemEvent) -> None:
+            src_path = Path(str(event.src_path))
+            if src_path in self._files.keys():
+                logger.info(f"Watched file '{src_path}' has been moved")
+                self._trigger(self._files[src_path])
 
         def on_modified(self, event: FileSystemEvent) -> None:
             src_path = Path(str(event.src_path))
             if src_path in self._files.keys():
                 logger.info(f"Watched file '{src_path}' has been modified")
-                self._reload(self._files[src_path])
+                self._trigger(self._files[src_path])
 
     _files_watchdog: Optional["FilesWatchdog"] = None
 
     class FilesWatchdog:
-        def __init__(self, files_to_watch: FilesToWatch) -> None:
+        def __init__(self, files_to_watch: FilesToWatch, config: KresConfig) -> None:
             self._observer = Observer()
 
-            event_handler = FilesWatchdogEventHandler(files_to_watch)
+            event_handler = FilesWatchdogEventHandler(files_to_watch, config)
             dirs_to_watch: List[Path] = []
             for file in files_to_watch.keys():
                 if file.parent not in dirs_to_watch:
@@ -104,7 +144,7 @@ def stop(self) -> None:
             self._observer.join()
 
 
-@only_on_real_changes_update(tls_cert_files_config)
+@only_on_real_changes_update(watched_files_config)
 async def _init_files_watchdog(config: KresConfig) -> None:
     if WATCHDOG_LIB:
         global _files_watchdog
@@ -119,9 +159,15 @@ async def _init_files_watchdog(config: KresConfig) -> None:
             files_to_watch[config.network.tls.cert_file.to_path()] = net_tls
             files_to_watch[config.network.tls.key_file.to_path()] = net_tls
 
+        # local-data.rpz
+        if config.local_data.rpz:
+            for rpz in config.local_data.rpz:
+                if rpz.watchdog:
+                    files_to_watch[rpz.file.to_path()] = None
+
         if files_to_watch:
             logger.info("Initializing files watchdog")
-            _files_watchdog = FilesWatchdog(files_to_watch)
+            _files_watchdog = FilesWatchdog(files_to_watch, config)
             _files_watchdog.start()
 
 
diff --git a/python/knot_resolver/manager/server.py b/python/knot_resolver/manager/server.py
@@ -138,6 +138,14 @@ async def _reload_config(self) -> None:
                 logger.error(f"Reloading of the configuration file failed: {e}")
                 logger.error("Configuration have NOT been changed.")
 
+    async def _renew_config(self) -> None:
+        try:
+            await self.config_store.renew()
+            logger.info("Configuration successfully renewed")
+        except KresManagerException as e:
+            logger.error(f"Renewing the configuration failed: {e}")
+            logger.error("Configuration have NOT been renewed.")
+
     async def sigint_handler(self) -> None:
         logger.info("Received SIGINT, triggering graceful shutdown")
         self.trigger_shutdown(0)
@@ -325,6 +333,15 @@ async def _handler_reload(self, _request: web.Request) -> web.Response:
         await self._reload_config()
         return web.Response(text="Reloading...")
 
+    async def _handler_renew(self, _request: web.Request) -> web.Response:
+        """
+        Route handler for renewing the configuration
+        """
+
+        logger.info("Renewing configuration event triggered...")
+        await self._renew_config()
+        return web.Response(text="Renewing configuration...")
+
     async def _handler_processes(self, request: web.Request) -> web.Response:
         """
         Route handler for listing PIDs of subprocesses
@@ -359,6 +376,7 @@ def _setup_routes(self) -> None:
                 web.patch(r"/v1/config{path:.*}", self._handler_config_query),
                 web.post("/stop", self._handler_stop),
                 web.post("/reload", self._handler_reload),
+                web.post("/renew", self._handler_renew),
                 web.get("/schema", self._handler_schema),
                 web.get("/schema/ui", self._handle_view_schema),
                 web.get("/metrics", self._handler_metrics),
diff --git a/tests/packaging/interactive/rpz_watchdog.sh b/tests/packaging/interactive/rpz_watchdog.sh
@@ -0,0 +1,121 @@
+#!/usr/bin/env bash
+
+set -e
+
+gitroot=$(git rev-parse --show-toplevel)
+rpz_file=$gitroot/example.rpz
+
+rpz_example=$(cat <<EOF
+\$ORIGIN RPZ.EXAMPLE.ORG.
+ok.example.com                    CNAME rpz-passthru.
+EOF
+)
+# create example RPZ
+echo "$rpz_example" >> $rpz_file
+
+rpz_conf=$(cat <<EOF
+local-data:
+  rpz:
+    - file: $rpz_file
+      watchdog: false
+EOF
+)
+# add RPZ to config
+echo "$rpz_conf" >> /etc/knot-resolver/config.yaml
+
+function count_errors(){
+    echo "$(journalctl -u knot-resolver.service | grep -c error)"
+}
+
+function count_reloads(){
+    echo "$(journalctl -u knot-resolver.service | grep -c "Reloading policy rules has finished")"
+}
+
+# test that RPZ watchdog
+# {{
+
+err_count=$(count_errors)
+rel_count=$(count_reloads)
+
+# reload config with RPZ configured without watchdog turned on
+kresctl reload
+sleep 1
+if [ $(count_errors) -ne $err_count ] || [ $(count_reloads) -ne $rel_count ]; then
+    echo "RPZ file watchdog is running (should not) or other errors occurred."
+    exit 1
+fi
+
+# configure RPZ file and turn on watchdog
+kresctl config set -p /local-data/rpz/0/watchdog true
+sleep 1
+if [ "$?" -ne "0" ]; then
+    echo "Could not turn on RPZ file watchdog."
+    exit 1
+fi
+
+# }}
+
+# test RPZ modification
+# {{
+
+# modify RPZ file, it will trigger reload
+rel_count=$(count_reloads)
+echo "32.1.2.0.192.rpz-client-ip        CNAME rpz-passthru." >> $rpz_file
+
+# wait for files reload to finish
+sleep 10
+
+if [ $(count_errors) -ne $err_count ] || [ $(count_reloads) -eq $rel_count ]; then
+    echo "Could not reload modified RPZ file."
+    exit 1
+fi
+
+# }}
+
+# test replacement
+# {{
+
+rel_count=$(count_reloads)
+
+# copy RPZ file
+cp $rpz_file $rpz_file.new
+
+# edit new files
+echo "48.zz.101.db8.2001.rpz-client-ip  CNAME rpz-passthru." >> $rpz_file.new
+
+# replace files
+cp -f $rpz_file.new $rpz_file
+
+# wait for files reload to finish
+sleep 10
+
+if [ $(count_errors) -ne $err_count ] || [ $(count_reloads) -eq $rel_count ]; then
+    echo "Could not reload replaced RPZ file."
+    exit 1
+fi
+
+# }}
+
+# test recovery from deletion and creation
+# {{
+
+rel_count=$(count_reloads)
+
+# backup rpz file
+cp $rpz_file $rpz_file.backup
+
+# delete RPZ file
+rm $rpz_file
+
+# create cert files
+cp -f $rpz_file.backup $rpz_file
+
+# wait for files reload to finish
+sleep 10
+
+if [ $(count_errors) -ne $err_count ] || [ $(count_reloads) -eq $rel_count ]; then
+    echo "Could not reload created RPZ file."
+    exit 1
+fi
+
+# }}
diff --git a/tests/packaging/interactive/watchdog.sh b/tests/packaging/interactive/watchdog.sh
@@ -26,7 +26,7 @@ function count_errors(){
 }
 
 function count_reloads(){
-    echo "$(journalctl -u knot-resolver.service | grep -c "Reloading of TLS certificate files has finished")"
+    echo "$(journalctl -u knot-resolver.service | grep -c "to reload watched files has finished")"
 }
 
 # test that files watchdog is turned off

Original file line number	Diff line number	Diff line change
`@@ -26,7 +26,7 @@ function count_errors(){`
`26`	`26`	`}`
`27`	`27`
`28`	`28`	`function count_reloads(){`
`29`		`- echo "$(journalctl -u knot-resolver.service \| grep -c "Reloading of TLS certificate files has finished")"`
	`29`	`+ echo "$(journalctl -u knot-resolver.service \| grep -c "to reload watched files has finished")"`
`30`	`30`	`}`
`31`	`31`
`32`	`32`	`# test that files watchdog is turned off`