lookyloo/bin/archiver.py

#!/usr/bin/env python3

import csv
import gzip
import logging
import logging.config
import os
import shutil

from collections import defaultdict
from collections.abc import Mapping
from datetime import datetime, timedelta, date
from pathlib import Path
from typing import Dict, List, Optional

from redis import Redis

from lookyloo.default import AbstractManager, get_config, get_homedir, get_socket_path
from lookyloo.helpers import get_captures_dir

logging.config.dictConfig(get_config('logging'))


class Archiver(AbstractManager):

    def __init__(self, loglevel: Optional[int]=None):
        super().__init__(loglevel)
        self.script_name = 'archiver'
        self.redis = Redis(unix_socket_path=get_socket_path('cache'))

        # make sure archived captures dir exists
        self.archived_captures_dir = get_homedir() / 'archived_captures'
        self.archived_captures_dir.mkdir(parents=True, exist_ok=True)

        self._load_indexes()

    def _to_run_forever(self):
        self._archive()
        self._update_all_capture_indexes()
        self._load_indexes()
        self._compress_hars()

    def _update_index(self, root_dir: Path) -> None:
        current_index: Dict[str, str] = {}
        if not os.listdir(root_dir):
            # the directory is empty, we can safely remove it
            root_dir.rmdir()
            return

        self.logger.debug(f'Updating index for {root_dir}')
        index_file = root_dir / 'index'
        existing_captures_names = {existing_capture.name for existing_capture in index_file.parent.iterdir()
                                   if existing_capture.name != 'index'}
        if index_file.exists():
            # Skip index if the directory has been archived.
            try:
                with index_file.open('r') as _f:
                    current_index = {uuid: dirname for uuid, dirname in csv.reader(_f)
                                     if uuid
                                     and dirname
                                     and dirname in existing_captures_names}
            except Exception as e:
                # the index file is broken, it will be recreated.
                self.logger.warning(f'Index for {root_dir} broken, recreating it: {e}')
                pass
            if not current_index:
                index_file.unlink()

        if set(current_index.values()) == existing_captures_names:
            # No new captures, quitting
            self.logger.debug(f'No new captures in {root_dir}.')
            return

        new_captures = sorted(existing_captures_names - set(current_index.values()), reverse=True)
        self.logger.info(f'{len(new_captures)} new captures in {root_dir}.')

        for capture_dir_name in new_captures:
            capture_dir = root_dir / capture_dir_name
            if not capture_dir.is_dir():
                self.logger.warning(f'{capture_dir} is not a directory')
                continue
            uuid_file = capture_dir / 'uuid'
            if not uuid_file.exists():
                self.logger.warning(f'No UUID file in {capture_dir}.')
                shutil.move(str(capture_dir), str(get_homedir() / 'discarded_captures'))
                continue
            with uuid_file.open() as _f:
                uuid = _f.read().strip()
                if not uuid:
                    self.logger.warning(f'{uuid_file} is empty')
                    shutil.move(str(capture_dir), str(get_homedir() / 'discarded_captures'))
                    continue
                if uuid in current_index:
                    self.logger.warning(f'Duplicate UUID ({uuid}) in {current_index[uuid]} and {uuid_file.parent.name}')
                    shutil.move(str(capture_dir), str(get_homedir() / 'discarded_captures'))
                    continue
                current_index[uuid] = uuid_file.parent.name

        if not current_index:
            # The directory has been archived. It is probably safe to unlink, but
            # if it's not, we will lose a whole buch of captures. Moving instead for safety.
            shutil.move(str(root_dir), str(get_homedir() / 'discarded_captures' / root_dir.parent / root_dir.name))
            return

        with index_file.open('w') as _f:
            index_writer = csv.writer(_f)
            for uuid, dirname in current_index.items():
                index_writer.writerow([uuid, dirname])

    def _make_dirs_list(self, root_dir: Path) -> List[Path]:
        directories = []
        year_now = date.today().year
        while True:
            year_dir = root_dir / str(year_now)
            if not year_dir.exists():
                # if we do not have a directory with this year, quit the loop
                break
            for month in range(12, 0, -1):
                month_dir = year_dir / f'{month:02}'
                if month_dir.exists():
                    directories.append(month_dir)
            year_now -= 1
        return directories

    def _update_all_capture_indexes(self):
        '''Run that after the captures are in the proper directories'''
        # Recent captures
        self.logger.info('Update recent indexes')
        # NOTE: the call below will check the existence of every path ending with `uuid`,
        #       it is extremely inneficient as we have many hundred of thusands of them
        #       and we only care about the rood directory (ex: 2023/06)
        # directories_to_index = {capture_dir.parent.parent
        #                        for capture_dir in get_captures_dir().glob('*/*/*/uuid')}
        for directory_to_index in self._make_dirs_list(get_captures_dir()):
            if self.shutdown_requested():
                self.logger.warning('Shutdown requested, breaking.')
                break
            self._update_index(directory_to_index)
        self.logger.info('Recent indexes updated')
        # Archived captures
        self.logger.info('Update archives indexes')
        for directory_to_index in self._make_dirs_list(self.archived_captures_dir):
            if self.shutdown_requested():
                self.logger.warning('Shutdown requested, breaking.')
                break
            self._update_index(directory_to_index)
        self.logger.info('Archived indexes updated')

    def _archive(self):
        archive_interval = timedelta(days=get_config('generic', 'archive'))
        cut_time = (datetime.now() - archive_interval).date()
        cut_time = cut_time.replace(day=1)

        # Format:
        # { 2020: { 12: [(directory, uuid)] } }
        to_archive: Dict[int, Dict[int, List[Path]]] = defaultdict(lambda: defaultdict(list))
        # In order to avoid scanning the complete directory on each run, we check if year and month are
        # older than the cut time.
        for index in get_captures_dir().glob('*/*/index'):
            if self.shutdown_requested():
                self.logger.warning('Shutdown requested, breaking.')
                break
            month = int(index.parent.name)
            year = int(index.parent.parent.name)
            if date(year, month, 1) >= cut_time:
                continue

            for capture_uuid in index.parent.glob('*/uuid'):
                try:
                    timestamp = datetime.strptime(capture_uuid.parent.name, '%Y-%m-%dT%H:%M:%S.%f')
                except ValueError:
                    timestamp = datetime.strptime(capture_uuid.parent.name, '%Y-%m-%dT%H:%M:%S')
                if timestamp.date() >= cut_time:
                    continue
                to_archive[timestamp.year][timestamp.month].append(capture_uuid.parent)
                self.logger.info(f'Archiving {capture_uuid.parent}.')

        if not to_archive:
            self.logger.info('Nothing to archive.')
            return

        p = self.redis.pipeline()
        for year, month_captures in to_archive.items():
            for month, captures in month_captures.items():
                dest_dir = self.archived_captures_dir / str(year) / f'{month:02}'
                dest_dir.mkdir(parents=True, exist_ok=True)
                for capture_path in captures:
                    p.delete(str(capture_path))
                    (capture_path / 'tree.pickle').unlink(missing_ok=True)
                    (capture_path / 'tree.pickle.gz').unlink(missing_ok=True)
                    shutil.move(str(capture_path), str(dest_dir))
        p.execute()

        self.logger.info('Archiving done.')

    def _compress_hars(self):
        self.logger.info('Compressing archived captures')
        for index in self.archived_captures_dir.glob('*/*/index'):
            if self.shutdown_requested():
                self.logger.warning('Shutdown requested, breaking.')
                break
            with index.open('r') as _f:
                for uuid, dirname in csv.reader(_f):
                    for har in (index.parent / dirname).rglob('*.har'):
                        if not har.exists():
                            continue
                        with har.open('rb') as f_in:
                            with gzip.open(f'{har}.gz', 'wb') as f_out:
                                shutil.copyfileobj(f_in, f_out)
                        har.unlink()
        self.logger.info('Archived captures compressed')

    def _load_indexes(self):
        # Initialize archives
        for index in get_captures_dir().glob('*/*/index'):
            if self.shutdown_requested():
                self.logger.warning('Shutdown requested, breaking.')
                break

            self.logger.info(f'Loading {index}')
            with index.open('r') as _f:
                recent_uuids: Mapping = {uuid: str(index.parent / dirname)
                                         for uuid, dirname in csv.reader(_f)
                                         if (index.parent / dirname).exists()}
            if recent_uuids:
                self.logger.info(f'{len(recent_uuids)} captures in directory.')
                self.redis.hset('lookup_dirs', mapping=recent_uuids)
            else:
                index.unlink()
        self.logger.info('Recent indexes loaded')

        already_archived_uuids = {k.decode() for k in self.redis.hkeys('lookup_dirs_archived')}
        self.logger.info(f'Already have {len(already_archived_uuids)} UUIDs archived')
        # Initialize archives
        for index in sorted(self.archived_captures_dir.glob('*/*/index'), reverse=True):
            if self.shutdown_requested():
                self.logger.warning('Shutdown requested, breaking.')
                break
            self.logger.debug(f'Loading {index}')
            with index.open('r') as _f:
                archived_uuids: Mapping = {uuid: index.parent / dirname
                                           for uuid, dirname in csv.reader(_f)}
            if archived_uuids:
                self.logger.debug(f'{len(archived_uuids)} captures in directory.')
                new_uuids = set(archived_uuids.keys()) - already_archived_uuids
                if not new_uuids:
                    self.logger.debug('No new archived UUID to check.')
                    continue

                self.logger.info(f'Loading {index}, {len(archived_uuids)} captures in directory, {len(new_uuids)} archived UUID to check.')
                # NOTE: Only check if the directory exists if the UUID isn't in the cache.
                self.redis.hset('lookup_dirs_archived',
                                mapping={uuid: str(dirname)
                                         for uuid, dirname in archived_uuids.items()
                                         if uuid in new_uuids and dirname.exists()})
                self.redis.hdel('lookup_dirs', *archived_uuids.keys())
            else:
                index.unlink()
        self.logger.info('Archived indexes loaded')


def main():
    a = Archiver()
    a.run(sleep_in_sec=3600)


if __name__ == '__main__':
    main()