# pyinotifyd A daemon to monitore filesystems events with inotify on Linux and execute tasks (Python methods or Shell commands) with an optional delay. It is also possible to cancel delayed tasks. ## Requirements * [pyinotify](https://github.com/seb-m/pyinotify) ## Installation ```sh pip install pyinotifyd ``` ## Configuration The config file **/etc/pyinotifyd/config.py** is written in Python syntax. pyinotifyd reads and executes its content, that means you can add your custom Python code to the config file. ### Tasks Tasks are Python methods that are called in case an event occurs. They can be bound directly to an event type in an event map. Although this is the easiest and quickest way, it is usually better to add a task to a scheduler and bind the scheduler to event types. #### Simple This is a very basic example task that just logs each event and task_id: ```python async def task(event, task_id): logging.info(f"{task_id}: execute example task: {event}") ``` #### FileManager FileManager moves, copy or deletes files and/or directories following a list of *rules*. A rule holds an *action* (move, copy or delete) and a regular expression *src_re*. The FileManager task will be executed if *src_re* matches the path of an event. If the action is copy or move, the destination path *dst_re* is mandatory and if *action* is delete and *rec* is set to True, non-empty directories will be deleted recursively. With *auto_create* set to True, possibly missing subdirectories in *dst_re* are created automatically. Regex subgroups or named-subgroups may be used in *src_re* and *dst_re*. Set the mode of moved/copied files/directories with *filemode* and *dirmode*. Ownership of moved/copied files/directories is set with *user* and *group*. Mode and ownership is also set to automatically created subdirectories. Use *logname* in log messages. ```python rule = Rule( action="move", src_re="^/src_path/(?P.*).to_move$", dst_re="/dst_path/\g", auto_create=False, rec=False, filemode=None, dirmode=None, user=None, group=None) fm = FileManager( rules=[rule], logname="FileManager") ``` FileManager provides a task **fm.task**. ### Schedulers pyinotifyd has different schedulers to schedule tasks with an optional delay. The advantages of using a scheduler are consistent logging and the possibility to cancel delayed tasks. Furthermore, schedulers have the ability to differentiate between files and directories. #### TaskScheduler TaskScheduler schedules *task* with an optional *delay* in seconds. Use the *files* and *dirs* arguments to schedule tasks only for files and/or directories. Use *logname* in log messages. All arguments except for *task* are optional. ```python s = TaskScheduler( task=task, files=True, dirs=False, delay=0, logname="TaskScheduler") ``` TaskScheduler provides two tasks which can be bound to an event in an event map. * **s.schedule** Schedule a task. If there is already a scheduled task, it will be canceled first. * **s.cancel** Cancel a scheduled task. #### ShellScheduler ShellScheduler schedules Shell command *cmd*. The placeholders **{maskname}**, **{pathname}** and **{src_pathname}** are replaced with the actual values of the event. ShellScheduler has the same optional arguments as TaskScheduler and provides the same tasks. ```python s1 = ShellScheduler(cmd="/usr/local/bin/task.sh {maskname} {pathname} {src_pathname}") ``` ### Event maps EventMap maps event types to tasks. It is possible to set a list of tasks to run multiple tasks on a single event. If the task of an event type is set to None, it is ignored. This is an example: ```python event_map = EventMap({ "IN_CLOSE_NOWRITE": [s.schedule, s1.schedule], "IN_CLOSE_WRITE": s.schedule}) ``` The following event types are available: * **IN_ACCESS**: a file was accessed * **IN_ATTRIB**: a metadata changed * **IN_CLOSE_NOWRITE**: an unwritable file was closed * **IN_CLOSE_WRITE**: a writable file was closed * **IN_CREATE**: a file/directory was created * **IN_DELETE**: a file/directory was deleted * **IN_DELETE_SELF**: a watched item itself was deleted * **IN_IGNORED**: raised when a watch is removed, probably useless for you * **IN_MODIFY**: a file was modified * **IN_MOVE_SELF**: a watched item was moved, currently its full pathname destination can only be known if its source and destination directories were both watched. Otherwise, the file is still being watched but you cannot rely anymore on the given path attribute *event.path* * **IN_MOVED_FROM**: a file/directory in a watched directory was moved from another specified watched directory. Can trace the full move of an item when IN_MOVED_TO is available too, in this case if the moved item is itself watched, its path will be updated (see IN_MOVE_SELF) * **IN_MOVED_TO**: a file/directory was moved to another specified watched directory (see IN_MOVE_FROM) * **IN_OPEN**: a file was opened * **IN_Q_OVERFLOW**: the event queue overflown. This event is not associated with any watch descriptor * **IN_UNMOUNT**: when backing filesystem was unmounted. Notified to each watch of this filesystem ### Watches Watch watches *path* for event types in *event_map* and execute the corresponding task(s). If *rec* is True, a watch will be added on each subdirectory in *path*. If *auto_add* is True, a watch will be added automatically on newly created subdirectories in *path*. ```python watch = Watch( path="/tmp", event_map=event_map, rec=False, auto_add=False) ``` ### PyinotifydConfig pyinotifyd expects an instance of PyinotifydConfig named **pyinotifyd_config** that holds its config options. The options are a list of *watches*, the *loglevel* (see https://docs.python.org/3/library/logging.html#levels) and the *shutdown_timeout*. pyinotifyd will wait *shutdown_timeout* seconds for pending tasks to complete during shutdown. ```python pyinotifyd_config = PyinotifydConfig( watches=[watch], loglevel=logging.INFO, shutdown_timeout=30) ``` ### Autostart pyinotifyd provides a systemd service file. ```sh # start pyinotifyd at boot time systemctl enable pyinotifyd.service # start the daemon immediately systemctl start pyinotifyd.service ``` ## Examples ### Schedule Python task for all events ```python async def task(event, task_id): logging.info(f"{task_id}: execute example task: {event}") s = TaskScheduler( task=task, files=True, dirs=True) event_map = EventMap( default_task=s.schedule) watch = Watch( path="/tmp", event_map=event_map, rec=True, auto_add=True) pyinotifyd_config = PyinotifydConfig( watches=[watch], loglevel=logging.INFO, shutdown_timeout=5) ``` ### Schedule Shell commands for specific events on files ```python s = ShellScheduler( cmd="/usr/local/sbin/task.sh {pathname}", files=True, dirs=False) event_map = EventMap( {"IN_WRITE_CLOSE": s.schedule}) watch = Watch( path="/tmp", event_map=event_map, rec=True, auto_add=True) pyinotifyd_config = PyinotifydConfig( watches=[watch], loglevel=logging.INFO, shutdown_timeout=5) ``` ### Move, copy or delete newly created files after a delay ```python move_rule = Rule( action="move", src_re="^/src_path/(?P.*)\.to_move$", dst_re="/dst_path/\g", auto_create=True, filemode=0o644, dirmode=0o755) copy_rule = Rule( action="copy", src_re="^/src_path/(?P.*)\.to_copy$", dst_re="/dst_path/\g", auto_create=True, filemode=0o644, dirmode=0o755) delete_rule = Rule( action="delete", src_re="^/src_path/(?P.*)\.to_delete$", rec=False) fm = FileManager( rules=[move_rule, copy_rule, delete_rule]) s = TaskScheduler( task=fm.task, delay=30, files=True, dirs=False) event_map = EventMap({ "IN_CLOSE_WRITE": s.schedule, "IN_DELETE": s.cancel, "IN_DELETE_SELF": s.cancel, "IN_MODIFY": s.cancel, "IN_MOVED_TO": s.schedule, "IN_UNMOUNT": s.cancel}) watch = Watch( path="/src_path", event_map=event_map, rec=True, auto_add=True) # note that shutdown_timeout should be greater than the greatest scheduler delay, # otherwise pending tasks may get cancelled during shutdown. pyinotifyd_config = PyinotifydConfig( watches=[watch], loglevel=logging.INFO, shutdown_timeout=35) ```