def bi_packs_sample_config(): bi_packs = BIAggregationPacks("") bi_packs._load_config(sample_config.bi_packs_config) bi_packs.load_config = lambda: None bi_packs.save_config = lambda: None yield bi_packs
def get_cached_bi_packs() -> BIAggregationPacks: bi_packs = BIAggregationPacks(BIManager.bi_configuration_file()) bi_packs.load_config() return bi_packs
class BICompiler: def __init__(self, bi_configuration_file, sites_callback: SitesCallback): self._sites_callback = sites_callback self._bi_configuration_file = bi_configuration_file self._logger = logger.getChild("bi.compiler") self._compiled_aggregations: Dict[str, BICompiledAggregation] = {} self._path_compilation_lock = Path(get_cache_dir(), "compilation.LOCK") self._path_compilation_timestamp = Path(get_cache_dir(), "last_compilation") self._path_compiled_aggregations = Path(get_cache_dir(), "compiled_aggregations") self._path_compiled_aggregations.mkdir(parents=True, exist_ok=True) self._redis_client: Optional["RedisDecoded"] = None self._setup() def _setup(self): self._bi_packs = BIAggregationPacks(self._bi_configuration_file) self._bi_structure_fetcher = BIStructureFetcher(self._sites_callback) self.bi_searcher = BISearcher() @property def compiled_aggregations(self) -> Dict[str, BICompiledAggregation]: return self._compiled_aggregations def cleanup(self) -> None: self._compiled_aggregations.clear() def load_compiled_aggregations(self) -> None: try: self._check_compilation_status() finally: self._load_compiled_aggregations() def _load_compiled_aggregations(self) -> None: for path_object in self._path_compiled_aggregations.iterdir(): if path_object.is_dir(): continue aggr_id = path_object.name if aggr_id.endswith(".new") or aggr_id in self._compiled_aggregations: continue self._logger.debug("Loading cached aggregation results %s" % aggr_id) aggr_data = self._load_data(str(path_object)) self._compiled_aggregations[aggr_id] = BIAggregation.create_trees_from_schema(aggr_data) def _check_compilation_status(self) -> None: current_configstatus = self.compute_current_configstatus() if not self._compilation_required(current_configstatus): self._logger.debug("No compilation required") return with store.locked(self._path_compilation_lock): # Re-check compilation required after lock has been required # Another apache might have done the job current_configstatus = self.compute_current_configstatus() if not self._compilation_required(current_configstatus): self._logger.debug("No compilation required. An other process already compiled it") return self.prepare_for_compilation(current_configstatus["online_sites"]) # Compile the raw tree for aggregation in self._bi_packs.get_all_aggregations(): start = time.time() self._compiled_aggregations[aggregation.id] = aggregation.compile(self.bi_searcher) self._logger.debug( "Compilation of %s took %f" % (aggregation.id, time.time() - start) ) self._verify_aggregation_title_uniqueness(self._compiled_aggregations) for aggr_id, aggr in self._compiled_aggregations.items(): start = time.time() result = aggr.serialize() self._logger.debug( "Schema dump %s took config took %f (%d branches)" % (aggr_id, time.time() - start, len(aggr.branches)) ) self._save_data(self._path_compiled_aggregations.joinpath(aggr_id), result) self._generate_part_of_aggregation_lookup(self._compiled_aggregations) known_sites = {kv[0]: kv[1] for kv in current_configstatus.get("known_sites", set())} self._cleanup_vanished_aggregations() self._bi_structure_fetcher.cleanup_orphaned_files(known_sites) store.save_text_to_file( str(self._path_compilation_timestamp), str(current_configstatus["configfile_timestamp"]) ) def _cleanup_vanished_aggregations(self): valid_aggregations = list(self._compiled_aggregations.keys()) for path_object in self._path_compiled_aggregations.iterdir(): if path_object.is_dir(): continue if path_object.name not in valid_aggregations: path_object.unlink(missing_ok=True) def _verify_aggregation_title_uniqueness( self, compiled_aggregations: Dict[str, BICompiledAggregation] ) -> None: used_titles: Dict[str, str] = {} for aggr_id, bi_aggregation in compiled_aggregations.items(): for bi_branch in bi_aggregation.branches: branch_title = bi_branch.properties.title if branch_title in used_titles: raise MKGeneralException( _( 'The aggregation titles are not unique. "%s" is created ' "by aggregation <b>%s</b> and <b>%s</b>" ) % (branch_title, aggr_id, used_titles[branch_title]) ) used_titles[branch_title] = aggr_id def prepare_for_compilation(self, online_sites: Set[SiteProgramStart]): self._bi_packs.load_config() self._bi_structure_fetcher.update_data(online_sites) self.bi_searcher.set_hosts(self._bi_structure_fetcher.hosts) def compile_aggregation_result( self, aggr_id: str, title: str ) -> Optional[BICompiledAggregation]: """Allows to compile a single aggregation with a given title. Does not save any results to disk""" current_configstatus = self.compute_current_configstatus() self.prepare_for_compilation(current_configstatus["online_sites"]) aggregation = self._bi_packs.get_aggregation(aggr_id) if not aggregation: return None try: aggregation.node.restrict_rule_title = title return aggregation.compile(self.bi_searcher) finally: aggregation.node.restrict_rule_title = None def _compilation_required(self, current_configstatus: ConfigStatus) -> bool: # Check monitoring core changes if self._site_status_changed(current_configstatus["online_sites"]): return True # Check BI configuration changes if current_configstatus["configfile_timestamp"] > self._get_compilation_timestamp(): return True return False def _get_compilation_timestamp(self) -> float: compilation_timestamp = 0.0 try: # I prefer Path.read_text # The corresponding cmk.utils.store has some "this function needs to die!" comment if self._path_compilation_timestamp.exists(): compilation_timestamp = float(self._path_compilation_timestamp.read_text()) except (FileNotFoundError, ValueError) as e: self._logger.warning("Can not determine compilation timestamp %s" % str(e)) return compilation_timestamp def _site_status_changed(self, required_program_starts: Set[SiteProgramStart]) -> bool: # The cached data may include more data than the currently required_program_starts # Empty branches are simply not shown during computation cached_program_starts = self._bi_structure_fetcher.get_cached_program_starts() return len(required_program_starts - cached_program_starts) > 0 def compute_current_configstatus(self) -> ConfigStatus: current_configstatus: ConfigStatus = { "configfile_timestamp": self._get_last_configuration_change(), "online_sites": set(), "known_sites": set(), } # The get status message also checks if the remote site is still alive result = self._sites_callback.query("GET status\nColumns: program_start\nCache: reload") program_start_times = {row[0]: int(row[1]) for row in result} for site_id, values in self._sites_callback.states().items(): start_time = program_start_times.get(site_id, 0) current_configstatus["known_sites"].add((site_id, start_time)) if values.get("state") == "online": current_configstatus["online_sites"].add((site_id, start_time)) return current_configstatus def _get_last_configuration_change(self) -> float: conf_dir = cmk.utils.paths.default_config_dir + "/multisite.d" latest_timestamp = 0.0 wato_config = Path(conf_dir, "wato", self._bi_configuration_file) if wato_config.exists(): latest_timestamp = max(latest_timestamp, wato_config.stat().st_mtime) for path_object in Path(conf_dir).iterdir(): if path_object.is_dir(): continue latest_timestamp = max(latest_timestamp, path_object.stat().st_mtime) return latest_timestamp def _save_data(self, filepath: Path, data) -> None: store.save_bytes_to_file(filepath, pickle.dumps(data)) def _load_data(self, filepath) -> Dict: return pickle.loads(store.load_bytes_from_file(filepath)) def _get_redis_client(self) -> "RedisDecoded": if self._redis_client is None: self._redis_client = get_redis_client() return self._redis_client def is_part_of_aggregation(self, host_name, service_description) -> bool: self._check_redis_lookup_integrity() return bool( self._get_redis_client().exists( "bi:aggregation_lookup:%s:%s" % (host_name, service_description) ) ) def _check_redis_lookup_integrity(self): client = self._get_redis_client() if client.exists("bi:aggregation_lookup"): return True # The following scenario only happens if the redis daemon loses its data # In the normal workflow the lookup cache is updated after the compilation phase # What happens if multiple apache process want to read the cache at the same time: # - One apache gets the lock, updates the cache # - The other apache wait till the cache has been updated lookup_lock = client.lock("bi:aggregation_lookup_lock") try: lookup_lock.acquire() if not client.exists("bi:aggregation_lookup"): self.load_compiled_aggregations() self._generate_part_of_aggregation_lookup(self._compiled_aggregations) finally: if lookup_lock.owned(): lookup_lock.release() def _generate_part_of_aggregation_lookup(self, compiled_aggregations): part_of_aggregation_map: Dict[str, List[str]] = {} for aggr_id, compiled_aggregation in compiled_aggregations.items(): for branch in compiled_aggregation.branches: for _site, host_name, service_description in branch.required_elements(): # This information can be used to selectively load the relevant compiled # aggregation for any host/service. Right now it is only an indicator if this # host/service is part of an aggregation key = "bi:aggregation_lookup:%s:%s" % (host_name, service_description) part_of_aggregation_map.setdefault(key, []).append( "%s\t%s" % (aggr_id, branch.properties.title) ) client = self._get_redis_client() # The main task here is to add/update/remove keys without causing other processes # to wait for the updated data. There is no tempfile -> live mechanism. # Updates are done on the live data via pipeline, using transactions. # Fetch existing keys existing_keys = set(client.scan_iter("bi:aggregation_lookup:*")) # Update keys pipeline = client.pipeline() for key, values in part_of_aggregation_map.items(): pipeline.sadd(key, *values) pipeline.set("bi:aggregation_lookup", "1") # Remove obsolete keys obsolete_keys = existing_keys - set(part_of_aggregation_map.keys()) if obsolete_keys: pipeline.delete(*obsolete_keys) pipeline.execute()
class BICompiler: def __init__(self, bi_configuration_file, sites_callback: SitesCallback): self._sites_callback = sites_callback self._bi_configuration_file = bi_configuration_file self._logger = logger.getChild("bi.compiler") self._compiled_aggregations: Dict[str, BICompiledAggregation] = {} self._path_compilation_lock = Path(get_cache_dir(), "compilation.LOCK") self._path_compilation_timestamp = Path(get_cache_dir(), "last_compilation") self._path_compiled_aggregations = Path(get_cache_dir(), "compiled_aggregations") self._path_compiled_aggregations.mkdir(parents=True, exist_ok=True) self._part_of_aggregation_map: Dict[Tuple[HostName, Optional[ServiceName]], List[Tuple[str, str]]] = {} self._setup() def _setup(self): self._bi_packs = BIAggregationPacks(self._bi_configuration_file) self._bi_structure_fetcher = BIStructureFetcher(self._sites_callback) self.bi_searcher = BISearcher() @property def compiled_aggregations(self) -> Dict[str, BICompiledAggregation]: return self._compiled_aggregations def cleanup(self) -> None: self._compiled_aggregations.clear() def load_compiled_aggregations(self) -> None: try: self._check_compilation_status() finally: self._load_compiled_aggregations() def _load_compiled_aggregations(self) -> None: for path_object in self._path_compiled_aggregations.iterdir(): aggr_id = path_object.name if aggr_id in self._compiled_aggregations: continue self._logger.debug("Loading cached aggregation results %s" % aggr_id) aggr_data = self._marshal_load_data(str(path_object)) self._compiled_aggregations[aggr_id] = BIAggregation.create_trees_from_schema(aggr_data) self._update_part_of_aggregation_map() def used_in_aggregation(self, host_name: HostName, service_description: ServiceName) -> bool: return (host_name, service_description) in self._part_of_aggregation_map def _update_part_of_aggregation_map(self) -> None: self._part_of_aggregation_map = {} for aggr_id, compiled_aggregation in self._compiled_aggregations.items(): for branch in compiled_aggregation.branches: for _site, host_name, service_description in branch.required_elements(): key = (host_name, service_description) self._part_of_aggregation_map.setdefault(key, []) self._part_of_aggregation_map[key].append((aggr_id, branch.properties.title)) def _check_compilation_status(self) -> None: current_configstatus = self.compute_current_configstatus() if not self._compilation_required(current_configstatus): self._logger.debug("No compilation required") return with store.locked(self._path_compilation_lock): # Re-check compilation required after lock has been required # Another apache might have done the job current_configstatus = self.compute_current_configstatus() if not self._compilation_required(current_configstatus): self._logger.debug("No compilation required. An other process already compiled it") return self.prepare_for_compilation(current_configstatus["online_sites"]) # Compile the raw tree for aggregation in self._bi_packs.get_all_aggregations(): start = time.time() self._compiled_aggregations[aggregation.id] = aggregation.compile(self.bi_searcher) self._logger.debug("Compilation of %s took %f" % (aggregation.id, time.time() - start)) self._verify_aggregation_title_uniqueness(self._compiled_aggregations) for aggr_id, aggr in self._compiled_aggregations.items(): start = time.time() result = BICompiledAggregationSchema().dump(aggr) self._logger.debug("Schema dump took config took %f (%d branches)" % (time.time() - start, len(aggr.branches))) start = time.time() self._marshal_save_data(self._path_compiled_aggregations.joinpath(aggr_id), result) self._logger.debug("Save dump to disk took %f" % (time.time() - start)) known_sites = {kv[0]: kv[1] for kv in current_configstatus.get("known_sites", set())} self._bi_structure_fetcher._cleanup_orphaned_files(known_sites) self._path_compilation_timestamp.write_text( str(current_configstatus["configfile_timestamp"])) def _verify_aggregation_title_uniqueness( self, compiled_aggregations: Dict[str, BICompiledAggregation]) -> None: used_titles: Dict[str, str] = {} for aggr_id, bi_aggregation in compiled_aggregations.items(): for bi_branch in bi_aggregation.branches: branch_title = bi_branch.properties.title if branch_title in used_titles: raise MKGeneralException( _("The aggregation titles are not unique. \"%s\" is created " "by aggregation <b>%s</b> and <b>%s</b>") % (branch_title, aggr_id, used_titles[branch_title])) used_titles[branch_title] = aggr_id def prepare_for_compilation(self, online_sites: Set[SiteProgramStart]): self._bi_packs.load_config() self._bi_structure_fetcher.update_data(online_sites) self.bi_searcher.set_hosts(self._bi_structure_fetcher.hosts) def compile_aggregation_result(self, aggr_id: str, title: str) -> Optional[BICompiledAggregation]: """ Allows to compile a single aggregation with a given title. Does not save any results to disk """ current_configstatus = self.compute_current_configstatus() self.prepare_for_compilation(current_configstatus["online_sites"]) aggregation = self._bi_packs.get_aggregation(aggr_id) if not aggregation: return None try: aggregation.node.restrict_rule_title = title return aggregation.compile(self.bi_searcher) finally: aggregation.node.restrict_rule_title = None def _compilation_required(self, current_configstatus: ConfigStatus) -> bool: # Check monitoring core changes if self._site_status_changed(current_configstatus["online_sites"]): return True # Check BI configuration changes if current_configstatus["configfile_timestamp"] > self._get_compilation_timestamp(): return True return False def _get_compilation_timestamp(self) -> float: compilation_timestamp = 0.0 try: # I prefer Path.read_text # The corresponding cmk.utils.store has some "this function needs to die!" comment if self._path_compilation_timestamp.exists(): compilation_timestamp = float(self._path_compilation_timestamp.read_text()) except (FileNotFoundError, ValueError) as e: self._logger.warning("Can not determine compilation timestamp %s" % str(e)) return compilation_timestamp def _site_status_changed(self, required_program_starts: Set[SiteProgramStart]) -> bool: # The cached data may include more data than the currently required_program_starts # Empty branches are simply not shown during computation cached_program_starts = self._bi_structure_fetcher.get_cached_program_starts() return len(required_program_starts - cached_program_starts) > 0 def compute_current_configstatus(self) -> ConfigStatus: current_configstatus: ConfigStatus = { "configfile_timestamp": self._get_last_configuration_change(), "online_sites": set(), "known_sites": set(), } # The get status message also checks if the remote site is still alive result = self._sites_callback.query("GET status\nColumns: program_start\nCache: reload") program_start_times = {row[0]: int(row[1]) for row in result} for site_id, values in self._sites_callback.states().items(): start_time = program_start_times.get(site_id, 0) current_configstatus["known_sites"].add((site_id, start_time)) if values.get("state") == "online": current_configstatus["online_sites"].add((site_id, start_time)) return current_configstatus def _get_last_configuration_change(self) -> float: conf_dir = cmk.utils.paths.default_config_dir + "/multisite.d" latest_timestamp = 0.0 wato_config = Path(conf_dir, "wato", self._bi_configuration_file) if wato_config.exists(): latest_timestamp = max(latest_timestamp, wato_config.stat().st_mtime) for path_object in Path(conf_dir).iterdir(): if path_object.is_dir(): continue latest_timestamp = max(latest_timestamp, path_object.stat().st_mtime) return latest_timestamp def _marshal_save_data(self, filepath, data) -> None: with open(filepath, "wb") as f: marshal.dump(data, f) os.fsync(f.fileno()) def _marshal_load_data(self, filepath) -> Dict: try: with open(filepath, "rb") as f: return marshal.load(f) except ValueError: return {}