core/homeassistant/components/recorder/core.py

"""Support for recording details."""

from __future__ import annotations

import asyncio
from collections.abc import Callable, Iterable
from concurrent.futures import CancelledError
import contextlib
from datetime import datetime, timedelta
import logging
import queue
import sqlite3
import threading
import time
from typing import TYPE_CHECKING, Any, cast

from propcache.api import cached_property
import psutil_home_assistant as ha_psutil
from sqlalchemy import create_engine, event as sqlalchemy_event, exc, select, update
from sqlalchemy.engine import Engine
from sqlalchemy.engine.interfaces import DBAPIConnection
from sqlalchemy.exc import SQLAlchemyError
from sqlalchemy.orm import scoped_session, sessionmaker
from sqlalchemy.orm.session import Session

from homeassistant.components import persistent_notification
from homeassistant.const import (
    ATTR_ENTITY_ID,
    EVENT_HOMEASSISTANT_CLOSE,
    EVENT_HOMEASSISTANT_FINAL_WRITE,
    EVENT_STATE_CHANGED,
    MATCH_ALL,
)
from homeassistant.core import (
    CALLBACK_TYPE,
    Event,
    EventStateChangedData,
    HomeAssistant,
    callback,
)
from homeassistant.helpers.event import (
    async_track_time_change,
    async_track_time_interval,
    async_track_utc_time_change,
)
from homeassistant.helpers.start import async_at_started
from homeassistant.helpers.typing import UNDEFINED, UndefinedType
import homeassistant.util.dt as dt_util
from homeassistant.util.enum import try_parse_enum
from homeassistant.util.event_type import EventType

from . import migration, statistics
from .const import (
    DB_WORKER_PREFIX,
    DEFAULT_MAX_BIND_VARS,
    DOMAIN,
    KEEPALIVE_TIME,
    LAST_REPORTED_SCHEMA_VERSION,
    MARIADB_PYMYSQL_URL_PREFIX,
    MARIADB_URL_PREFIX,
    MAX_QUEUE_BACKLOG_MIN_VALUE,
    MIN_AVAILABLE_MEMORY_FOR_QUEUE_BACKLOG,
    MYSQLDB_PYMYSQL_URL_PREFIX,
    MYSQLDB_URL_PREFIX,
    SQLITE_URL_PREFIX,
    SupportedDialect,
)
from .db_schema import (
    SCHEMA_VERSION,
    Base,
    EventData,
    Events,
    EventTypes,
    StateAttributes,
    States,
    StatesMeta,
    Statistics,
    StatisticsShortTerm,
)
from .executor import DBInterruptibleThreadPoolExecutor
from .models import DatabaseEngine, StatisticData, StatisticMetaData, UnsupportedDialect
from .pool import POOL_SIZE, MutexPool, RecorderPool
from .table_managers.event_data import EventDataManager
from .table_managers.event_types import EventTypeManager
from .table_managers.recorder_runs import RecorderRunsManager
from .table_managers.state_attributes import StateAttributesManager
from .table_managers.states import StatesManager
from .table_managers.states_meta import StatesMetaManager
from .table_managers.statistics_meta import StatisticsMetaManager
from .tasks import (
    AdjustLRUSizeTask,
    AdjustStatisticsTask,
    ChangeStatisticsUnitTask,
    ClearStatisticsTask,
    CommitTask,
    CompileMissingStatisticsTask,
    DatabaseLockTask,
    ImportStatisticsTask,
    KeepAliveTask,
    PerodicCleanupTask,
    PurgeTask,
    RecorderTask,
    StatisticsTask,
    StopTask,
    SynchronizeTask,
    UpdateStatesMetadataTask,
    UpdateStatisticsMetadataTask,
    WaitTask,
)
from .util import (
    async_create_backup_failure_issue,
    build_mysqldb_conv,
    dburl_to_path,
    end_incomplete_runs,
    is_second_sunday,
    move_away_broken_database,
    session_scope,
    setup_connection_for_dialect,
    validate_or_move_away_sqlite_database,
    write_lock_db_sqlite,
)

_LOGGER = logging.getLogger(__name__)

DEFAULT_URL = "sqlite:///{hass_config_path}"

# Controls how often we clean up
# States and Events objects
EXPIRE_AFTER_COMMITS = 120

SHUTDOWN_TASK = object()

COMMIT_TASK = CommitTask()
KEEP_ALIVE_TASK = KeepAliveTask()
WAIT_TASK = WaitTask()
ADJUST_LRU_SIZE_TASK = AdjustLRUSizeTask()

DB_LOCK_TIMEOUT = 30
DB_LOCK_QUEUE_CHECK_TIMEOUT = 10  # check every 10 seconds

QUEUE_CHECK_INTERVAL = timedelta(minutes=5)

INVALIDATED_ERR = "Database connection invalidated"
CONNECTIVITY_ERR = "Error in database connectivity during commit"

# Pool size must accommodate Recorder thread + All db executors
MAX_DB_EXECUTOR_WORKERS = POOL_SIZE - 1


class Recorder(threading.Thread):
    """A threaded recorder class."""

    stop_requested: bool

    def __init__(
        self,
        hass: HomeAssistant,
        auto_purge: bool,
        auto_repack: bool,
        keep_days: int,
        commit_interval: int,
        uri: str,
        db_max_retries: int,
        db_retry_wait: int,
        entity_filter: Callable[[str], bool] | None,
        exclude_event_types: set[EventType[Any] | str],
    ) -> None:
        """Initialize the recorder."""
        threading.Thread.__init__(self, name="Recorder")

        self.hass = hass
        self.thread_id: int | None = None
        self.recorder_and_worker_thread_ids: set[int] = set()
        self.auto_purge = auto_purge
        self.auto_repack = auto_repack
        self.keep_days = keep_days
        self.is_running: bool = False
        self._hass_started: asyncio.Future[object] = hass.loop.create_future()
        self.commit_interval = commit_interval
        self._queue: queue.SimpleQueue[RecorderTask | Event] = queue.SimpleQueue()
        self.db_url = uri
        self.db_max_retries = db_max_retries
        self.db_retry_wait = db_retry_wait
        self.database_engine: DatabaseEngine | None = None
        # Database connection is ready, but non-live migration may be in progress
        db_connected: asyncio.Future[bool] = hass.data[DOMAIN].db_connected
        self.async_db_connected: asyncio.Future[bool] = db_connected
        # Database is ready to use but live migration may be in progress
        self.async_db_ready: asyncio.Future[bool] = hass.loop.create_future()
        # Database is ready to use and all migration steps completed (used by tests)
        self.async_recorder_ready = asyncio.Event()
        self._queue_watch = threading.Event()
        self.engine: Engine | None = None
        self.max_backlog: int = MAX_QUEUE_BACKLOG_MIN_VALUE
        self._psutil: ha_psutil.PsutilWrapper | None = None

        # The entity_filter is exposed on the recorder instance so that
        # it can be used to see if an entity is being recorded and is called
        # by is_entity_recorder and the sensor recorder.
        self.entity_filter = entity_filter
        self.exclude_event_types = exclude_event_types

        self.schema_version = 0
        self._commits_without_expire = 0
        self._event_session_has_pending_writes = False

        self.recorder_runs_manager = RecorderRunsManager()
        self.states_manager = StatesManager()
        self.event_data_manager = EventDataManager(self)
        self.event_type_manager = EventTypeManager(self)
        self.states_meta_manager = StatesMetaManager(self)
        self.state_attributes_manager = StateAttributesManager(self)
        self.statistics_meta_manager = StatisticsMetaManager(self)

        self.event_session: Session | None = None
        self._get_session: Callable[[], Session] | None = None
        self._completed_first_database_setup: bool | None = None
        self.migration_in_progress = False
        self.migration_is_live = False
        self.use_legacy_events_index = False
        self._database_lock_task: DatabaseLockTask | None = None
        self._db_executor: DBInterruptibleThreadPoolExecutor | None = None

        self._event_listener: CALLBACK_TYPE | None = None
        self._queue_watcher: CALLBACK_TYPE | None = None
        self._keep_alive_listener: CALLBACK_TYPE | None = None
        self._commit_listener: CALLBACK_TYPE | None = None
        self._periodic_listener: CALLBACK_TYPE | None = None
        self._nightly_listener: CALLBACK_TYPE | None = None
        self._dialect_name: SupportedDialect | None = None
        self.enabled = True

        # We update the value once we connect to the DB
        # and determine what is actually supported.
        self.max_bind_vars = DEFAULT_MAX_BIND_VARS

    @property
    def backlog(self) -> int:
        """Return the number of items in the recorder backlog."""
        return self._queue.qsize()

    @cached_property
    def dialect_name(self) -> SupportedDialect | None:
        """Return the dialect the recorder uses."""
        return self._dialect_name

    @property
    def _using_file_sqlite(self) -> bool:
        """Short version to check if we are using sqlite3 as a file."""
        return self.db_url != SQLITE_URL_PREFIX and self.db_url.startswith(
            SQLITE_URL_PREFIX
        )

    @property
    def recording(self) -> bool:
        """Return if the recorder is recording."""
        return self._event_listener is not None

    def get_session(self) -> Session:
        """Get a new sqlalchemy session."""
        if self._get_session is None:
            raise RuntimeError("The database connection has not been established")
        return self._get_session()

    def queue_task(self, task: RecorderTask | Event) -> None:
        """Add a task to the recorder queue."""
        self._queue.put(task)

    def set_enable(self, enable: bool) -> None:
        """Enable or disable recording events and states."""
        self.enabled = enable

    @callback
    def async_start_executor(self) -> None:
        """Start the executor."""
        self._db_executor = DBInterruptibleThreadPoolExecutor(
            self.recorder_and_worker_thread_ids,
            thread_name_prefix=DB_WORKER_PREFIX,
            max_workers=MAX_DB_EXECUTOR_WORKERS,
            shutdown_hook=self._shutdown_pool,
        )

    def _shutdown_pool(self) -> None:
        """Close the dbpool connections in the current thread."""
        if self.engine and hasattr(self.engine.pool, "shutdown"):
            self.engine.pool.shutdown()

    @callback
    def async_initialize(self) -> None:
        """Initialize the recorder."""
        entity_filter = self.entity_filter
        exclude_event_types = self.exclude_event_types
        queue_put = self._queue.put_nowait

        @callback
        def _event_listener(event: Event) -> None:
            """Listen for new events and put them in the process queue."""
            if event.event_type in exclude_event_types:
                return

            if entity_filter is None or not (
                entity_id := event.data.get(ATTR_ENTITY_ID)
            ):
                queue_put(event)
                return

            if isinstance(entity_id, str):
                if entity_filter(entity_id):
                    queue_put(event)
                return

            if isinstance(entity_id, list):
                for eid in entity_id:
                    if entity_filter(eid):
                        queue_put(event)
                        return
                return

            # Unknown what it is.
            queue_put(event)

        self._event_listener = self.hass.bus.async_listen(
            MATCH_ALL,
            _event_listener,
        )
        self._queue_watcher = async_track_time_interval(
            self.hass,
            self._async_check_queue,
            QUEUE_CHECK_INTERVAL,
            name="Recorder queue watcher",
        )

    @callback
    def _async_keep_alive(self, now: datetime) -> None:
        """Queue a keep alive."""
        if self._event_listener:
            self.queue_task(KEEP_ALIVE_TASK)

    @callback
    def _async_commit(self, now: datetime) -> None:
        """Queue a commit."""
        if (
            self._event_listener
            and not self._database_lock_task
            and self._event_session_has_pending_writes
        ):
            self.queue_task(COMMIT_TASK)

    @callback
    def async_add_executor_job[_T](
        self, target: Callable[..., _T], *args: Any
    ) -> asyncio.Future[_T]:
        """Add an executor job from within the event loop."""
        return self.hass.loop.run_in_executor(self._db_executor, target, *args)

    @callback
    def _async_check_queue(self, *_: Any) -> None:
        """Periodic check of the queue size to ensure we do not exhaust memory.

        The queue grows during migration or if something really goes wrong.
        """
        _LOGGER.debug("Recorder queue size is: %s", self.backlog)
        if not self._reached_max_backlog():
            return
        _LOGGER.error(
            (
                "The recorder backlog queue reached the maximum size of %s events; "
                "usually, the system is CPU bound, I/O bound, or the database "
                "is corrupt due to a disk problem; The recorder will stop "
                "recording events to avoid running out of memory"
            ),
            self.backlog,
        )
        self._async_stop_queue_watcher_and_event_listener()

    def _available_memory(self) -> int:
        """Return the available memory in bytes."""
        if not self._psutil:
            self._psutil = ha_psutil.PsutilWrapper()
        return cast(int, self._psutil.psutil.virtual_memory().available)

    def _reached_max_backlog(self) -> bool:
        """Check if the system has reached the max queue backlog and return True if it has."""
        # First check the minimum value since its cheap
        if self.backlog < MAX_QUEUE_BACKLOG_MIN_VALUE:
            return False
        # If they have more RAM available, keep filling the backlog
        # since we do not want to stop recording events or give the
        # user a bad backup when they have plenty of RAM available.
        return self._available_memory() < MIN_AVAILABLE_MEMORY_FOR_QUEUE_BACKLOG

    @callback
    def _async_stop_queue_watcher_and_event_listener(self) -> None:
        """Stop watching the queue and listening for events."""
        if self._queue_watcher:
            self._queue_watcher()
            self._queue_watcher = None
        if self._event_listener:
            self._event_listener()
            self._event_listener = None

    @callback
    def _async_stop_listeners(self) -> None:
        """Stop listeners."""
        self._async_stop_queue_watcher_and_event_listener()
        if self._keep_alive_listener:
            self._keep_alive_listener()
            self._keep_alive_listener = None
        if self._commit_listener:
            self._commit_listener()
            self._commit_listener = None
        if self._nightly_listener:
            self._nightly_listener()
            self._nightly_listener = None
        if self._periodic_listener:
            self._periodic_listener()
            self._periodic_listener = None

    async def _async_close(self, event: Event) -> None:
        """Empty the queue if its still present at close."""

        # If the queue is full of events to be processed because
        # the database is so broken that every event results in a retry
        # we will never be able to get though the events to shutdown in time.
        #
        # We drain all the events in the queue and then insert
        # an empty one to ensure the next thing the recorder sees
        # is a request to shutdown.
        while True:
            try:
                self._queue.get_nowait()
            except queue.Empty:
                break
        self.queue_task(StopTask())
        await self.hass.async_add_executor_job(self.join)

    async def _async_shutdown(self, event: Event) -> None:
        """Shut down the Recorder at final write."""
        if not self._hass_started.done():
            self._hass_started.set_result(SHUTDOWN_TASK)
        self.queue_task(StopTask())
        self._async_stop_listeners()
        await self.hass.async_add_executor_job(self.join)

    @callback
    def _async_hass_started(self, hass: HomeAssistant) -> None:
        """Notify that hass has started."""
        self._hass_started.set_result(None)

    @callback
    def async_register(self) -> None:
        """Post connection initialize."""
        bus = self.hass.bus
        bus.async_listen_once(EVENT_HOMEASSISTANT_CLOSE, self._async_close)
        bus.async_listen_once(EVENT_HOMEASSISTANT_FINAL_WRITE, self._async_shutdown)
        async_at_started(self.hass, self._async_hass_started)

    @callback
    def _async_startup_done(self, startup_failed: bool) -> None:
        """Report startup failure."""
        # If a live migration failed, we were able to connect (async_db_connected
        # marked True), the database was marked ready (async_db_ready marked
        # True), the data in the queue cannot be written to the database because
        # the schema not in the correct format so we must stop listeners and report
        # failure.
        if not self.async_db_connected.done():
            self.async_db_connected.set_result(False)
        if not self.async_db_ready.done():
            self.async_db_ready.set_result(False)
        if startup_failed:
            persistent_notification.async_create(
                self.hass,
                "The recorder could not start, check [the logs](/config/logs)",
                "Recorder",
            )
        self._async_stop_listeners()

    @callback
    def async_connection_success(self) -> None:
        """Connect to the database succeeded, schema version and migration need known.

        The database may not yet be ready for use in case of a non-live migration.
        """
        self.async_db_connected.set_result(True)

    @callback
    def async_set_db_ready(self) -> None:
        """Database live and ready for use.

        Called after non-live migration steps are finished.
        """
        if self.async_db_ready.done():
            return
        self.async_db_ready.set_result(True)
        self.async_start_executor()

    @callback
    def _async_set_recorder_ready_migration_done(self) -> None:
        """Finish start and mark recorder ready.

        Called after all migration steps are finished.
        """
        self._async_setup_periodic_tasks()
        self.async_recorder_ready.set()

    @callback
    def async_nightly_tasks(self, now: datetime) -> None:
        """Trigger the purge."""
        if self.auto_purge:
            # Purge will schedule the periodic cleanups
            # after it completes to ensure it does not happen
            # until after the database is vacuumed
            repack = self.auto_repack and is_second_sunday(now)
            purge_before = dt_util.utcnow() - timedelta(days=self.keep_days)
            self.queue_task(PurgeTask(purge_before, repack=repack, apply_filter=False))
        else:
            self.queue_task(PerodicCleanupTask())

    @callback
    def _async_five_minute_tasks(self, now: datetime) -> None:
        """Run tasks every five minutes."""
        self.queue_task(ADJUST_LRU_SIZE_TASK)
        self.async_periodic_statistics()

    def _adjust_lru_size(self) -> None:
        """Trigger the LRU adjustment.

        If the number of entities has increased, increase the size of the LRU
        cache to avoid thrashing.
        """
        if new_size := self.hass.states.async_entity_ids_count() * 2:
            self.state_attributes_manager.adjust_lru_size(new_size)
            self.states_meta_manager.adjust_lru_size(new_size)
            self.statistics_meta_manager.adjust_lru_size(new_size)

    @callback
    def async_periodic_statistics(self) -> None:
        """Trigger the statistics run.

        Short term statistics run every 5 minutes
        """
        start = statistics.get_start_time()
        self.queue_task(StatisticsTask(start, True))

    @callback
    def async_adjust_statistics(
        self,
        statistic_id: str,
        start_time: datetime,
        sum_adjustment: float,
        adjustment_unit: str,
    ) -> None:
        """Adjust statistics."""
        self.queue_task(
            AdjustStatisticsTask(
                statistic_id, start_time, sum_adjustment, adjustment_unit
            )
        )

    @callback
    def async_clear_statistics(
        self, statistic_ids: list[str], *, on_done: Callable[[], None] | None = None
    ) -> None:
        """Clear statistics for a list of statistic_ids."""
        self.queue_task(ClearStatisticsTask(on_done, statistic_ids))

    @callback
    def async_update_statistics_metadata(
        self,
        statistic_id: str,
        *,
        new_statistic_id: str | UndefinedType = UNDEFINED,
        new_unit_of_measurement: str | None | UndefinedType = UNDEFINED,
        on_done: Callable[[], None] | None = None,
    ) -> None:
        """Update statistics metadata for a statistic_id."""
        self.queue_task(
            UpdateStatisticsMetadataTask(
                on_done, statistic_id, new_statistic_id, new_unit_of_measurement
            )
        )

    @callback
    def async_update_states_metadata(
        self,
        entity_id: str,
        new_entity_id: str,
    ) -> None:
        """Update states metadata for an entity_id."""
        self.queue_task(UpdateStatesMetadataTask(entity_id, new_entity_id))

    @callback
    def async_change_statistics_unit(
        self,
        statistic_id: str,
        *,
        new_unit_of_measurement: str,
        old_unit_of_measurement: str,
    ) -> None:
        """Change statistics unit for a statistic_id."""
        self.queue_task(
            ChangeStatisticsUnitTask(
                statistic_id, new_unit_of_measurement, old_unit_of_measurement
            )
        )

    @callback
    def async_import_statistics(
        self,
        metadata: StatisticMetaData,
        stats: Iterable[StatisticData],
        table: type[Statistics | StatisticsShortTerm],
    ) -> None:
        """Schedule import of statistics."""
        self.queue_task(ImportStatisticsTask(metadata, stats, table))

    @callback
    def _async_setup_periodic_tasks(self) -> None:
        """Prepare periodic tasks."""
        if self.hass.is_stopping or not self._get_session:
            # Home Assistant is shutting down
            return

        # If the db is using a socket connection, we need to keep alive
        # to prevent errors from unexpected disconnects
        if self.dialect_name != SupportedDialect.SQLITE:
            self._keep_alive_listener = async_track_time_interval(
                self.hass,
                self._async_keep_alive,
                timedelta(seconds=KEEPALIVE_TIME),
                name="Recorder keep alive",
            )

        # If the commit interval is not 0, we need to commit periodically
        if self.commit_interval:
            self._commit_listener = async_track_time_interval(
                self.hass,
                self._async_commit,
                timedelta(seconds=self.commit_interval),
                name="Recorder commit",
            )

        # Run nightly tasks at 4:12am
        self._nightly_listener = async_track_time_change(
            self.hass, self.async_nightly_tasks, hour=4, minute=12, second=0
        )

        # Compile short term statistics every 5 minutes
        self._periodic_listener = async_track_utc_time_change(
            self.hass, self._async_five_minute_tasks, minute=range(0, 60, 5), second=10
        )

    async def _async_wait_for_started(self) -> object | None:
        """Wait for the hass started future."""
        return await self._hass_started

    def _wait_startup_or_shutdown(self) -> object | None:
        """Wait for startup or shutdown before starting."""
        try:
            return asyncio.run_coroutine_threadsafe(
                self._async_wait_for_started(), self.hass.loop
            ).result()
        except CancelledError as ex:
            _LOGGER.warning(
                "Recorder startup was externally canceled before it could complete: %s",
                ex,
            )
            return SHUTDOWN_TASK

    def run(self) -> None:
        """Run the recorder thread."""
        self.is_running = True
        try:
            self._run()
        except Exception:
            _LOGGER.exception(
                "Recorder._run threw unexpected exception, recorder shutting down"
            )
        finally:
            # Ensure shutdown happens cleanly if
            # anything goes wrong in the run loop
            self.is_running = False
            self._shutdown()

    def _add_to_session(self, session: Session, obj: object) -> None:
        """Add an object to the session."""
        self._event_session_has_pending_writes = True
        session.add(obj)

    def _notify_migration_failed(self) -> None:
        """Notify the user schema migration failed."""
        persistent_notification.create(
            self.hass,
            "The database migration failed, check [the logs](/config/logs).",
            "Database Migration Failed",
            "recorder_database_migration",
        )

    def _dismiss_migration_in_progress(self) -> None:
        """Dismiss notification about migration in progress."""
        persistent_notification.dismiss(self.hass, "recorder_database_migration")

    def _run(self) -> None:
        """Start processing events to save."""
        thread_id = threading.get_ident()
        self.thread_id = thread_id
        self.recorder_and_worker_thread_ids.add(thread_id)

        setup_result = self._setup_recorder()

        if not setup_result:
            _LOGGER.error("Recorder setup failed, recorder shutting down")
            # Give up if we could not connect
            return

        schema_status = migration.validate_db_schema(self.hass, self, self.get_session)
        if schema_status is None:
            # Give up if we could not validate the schema
            _LOGGER.error("Failed to validate schema, recorder shutting down")
            return
        if schema_status.current_version > SCHEMA_VERSION:
            _LOGGER.error(
                "The database schema version %s is newer than %s which is the maximum "
                "database schema version supported by the installed version of "
                "Home Assistant Core, either upgrade Home Assistant Core or restore "
                "the database from a backup compatible with this version",
                schema_status.current_version,
                SCHEMA_VERSION,
            )
            return
        self.schema_version = schema_status.current_version

        if not schema_status.migration_needed and not schema_status.schema_errors:
            self._setup_run()
        else:
            self.migration_in_progress = True
            self.migration_is_live = migration.live_migration(schema_status)

        self.hass.add_job(self.async_connection_success)

        # First do non-live migration steps, if needed
        if schema_status.migration_needed:
            # Do non-live schema migration
            result, schema_status = self._migrate_schema_offline(schema_status)
            if not result:
                self._notify_migration_failed()
                self.migration_in_progress = False
                return
            self.schema_version = schema_status.current_version

            # Do non-live data migration
            self._migrate_data_offline(schema_status)

            # Non-live migration is now completed, remaining steps are live
            self.migration_is_live = True

        # After non-live migration, activate the recorder
        self._activate_and_set_db_ready(schema_status)
        # We wait to start a live migration until startup has finished
        # since it can be cpu intensive and we do not want it to compete
        # with startup which is also cpu intensive
        if self._wait_startup_or_shutdown() is SHUTDOWN_TASK:
            # Shutdown happened before Home Assistant finished starting
            self.migration_in_progress = False
            # Make sure we cleanly close the run if
            # we restart before startup finishes
            return

        # Do live migration steps and repairs, if needed
        if schema_status.migration_needed or schema_status.schema_errors:
            result, schema_status = self._migrate_schema_live(schema_status)
            if result:
                self.schema_version = SCHEMA_VERSION
                if not self._event_listener:
                    # If the schema migration takes so long that the end
                    # queue watcher safety kicks in because _reached_max_backlog
                    # was True, we need to reinitialize the listener.
                    self.hass.add_job(self.async_initialize)
            else:
                self.migration_in_progress = False
                self._dismiss_migration_in_progress()
                self._notify_migration_failed()
                return

        # Schema migration and repair is now completed
        if self.migration_in_progress:
            self.migration_in_progress = False
            self._dismiss_migration_in_progress()
            self._setup_run()

        # Catch up with missed statistics
        self._schedule_compile_missing_statistics()
        _LOGGER.debug("Recorder processing the queue")
        self._adjust_lru_size()
        self.hass.add_job(self._async_set_recorder_ready_migration_done)
        self._run_event_loop()

    def _activate_and_set_db_ready(
        self, schema_status: migration.SchemaValidationStatus
    ) -> None:
        """Activate the table managers or schedule migrations and mark the db as ready."""
        with session_scope(session=self.get_session()) as session:
            # Prime the statistics meta manager as soon as possible
            # since we want the frontend queries to avoid a thundering
            # herd of queries to find the statistics meta data if
            # there are a lot of statistics graphs on the frontend.
            self.statistics_meta_manager.load(session)

        migration.migrate_data_live(self, self.get_session, schema_status)

        # We must only set the db ready after we have set the table managers
        # to active if there is no data to migrate.
        #
        # This ensures that the history queries will use the new tables
        # and not the old ones as soon as the API is available.
        self.hass.add_job(self.async_set_db_ready)

    def _run_event_loop(self) -> None:
        """Run the event loop for the recorder."""
        # Use a session for the event read loop
        # with a commit every time the event time
        # has changed. This reduces the disk io.
        queue_ = self._queue
        startup_task_or_events: list[RecorderTask | Event] = []
        while not queue_.empty() and (task_or_event := queue_.get_nowait()):
            startup_task_or_events.append(task_or_event)
        self._pre_process_startup_events(startup_task_or_events)
        for task in startup_task_or_events:
            self._guarded_process_one_task_or_event_or_recover(task)

        # Clear startup tasks since this thread runs forever
        # and we don't want to hold them in memory
        del startup_task_or_events

        self.stop_requested = False
        while not self.stop_requested:
            self._guarded_process_one_task_or_event_or_recover(queue_.get())

    def _pre_process_startup_events(
        self, startup_task_or_events: list[RecorderTask | Event[Any]]
    ) -> None:
        """Pre process startup events."""
        # Prime all the state_attributes and event_data caches
        # before we start processing events
        state_change_events: list[Event[EventStateChangedData]] = []
        non_state_change_events: list[Event] = []

        for task_or_event in startup_task_or_events:
            # Event is never subclassed so we can
            # use a fast type check
            if type(task_or_event) is Event:
                event_ = task_or_event
                if event_.event_type == EVENT_STATE_CHANGED:
                    state_change_events.append(event_)
                else:
                    non_state_change_events.append(event_)

        assert self.event_session is not None
        session = self.event_session
        self.event_data_manager.load(non_state_change_events, session)
        self.event_type_manager.load(non_state_change_events, session)
        self.states_meta_manager.load(state_change_events, session)
        self.state_attributes_manager.load(state_change_events, session)

    def _guarded_process_one_task_or_event_or_recover(
        self, task: RecorderTask | Event
    ) -> None:
        """Process a task, guarding against exceptions to ensure the loop does not collapse."""
        _LOGGER.debug("Processing task: %s", task)
        try:
            self._process_one_task_or_event_or_recover(task)
        except Exception:
            _LOGGER.exception("Error while processing event %s", task)

    def _process_one_task_or_event_or_recover(self, task: RecorderTask | Event) -> None:
        """Process a task or event, reconnect, or recover a malformed database."""
        try:
            # Almost everything coming in via the queue
            # is an Event so we can process it directly
            # and since its never subclassed, we can
            # use a fast type check
            if type(task) is Event:
                self._process_one_event(task)
                return
            # If its not an event, commit everything
            # that is pending before running the task
            if TYPE_CHECKING:
                assert isinstance(task, RecorderTask)
            if task.commit_before:
                self._commit_event_session_or_retry()
            task.run(self)
        except exc.DatabaseError as err:
            if self._handle_database_error(err, setup_run=True):
                return
            _LOGGER.exception("Unhandled database error while processing task %s", task)
        except SQLAlchemyError:
            _LOGGER.exception("SQLAlchemyError error processing task %s", task)
        else:
            return

        # Reset the session if an SQLAlchemyError (including DatabaseError)
        # happens to rollback and recover
        self._reopen_event_session()

    def _setup_recorder(self) -> bool:
        """Create a connection to the database."""
        tries = 1

        while tries <= self.db_max_retries:
            try:
                self._setup_connection()
                return migration.initialize_database(self.get_session)
            except UnsupportedDialect:
                break
            except Exception:
                _LOGGER.exception(
                    "Error during connection setup: (retrying in %s seconds)",
                    self.db_retry_wait,
                )
            tries += 1

            if tries <= self.db_max_retries:
                self._close_connection()
                time.sleep(self.db_retry_wait)

        return False

    def _migrate_data_offline(
        self, schema_status: migration.SchemaValidationStatus
    ) -> None:
        """Migrate data."""
        with self.hass.timeout.freeze(DOMAIN):
            migration.migrate_data_non_live(self, self.get_session, schema_status)

    def _migrate_schema_offline(
        self, schema_status: migration.SchemaValidationStatus
    ) -> tuple[bool, migration.SchemaValidationStatus]:
        """Migrate schema to the latest version."""
        with self.hass.timeout.freeze(DOMAIN):
            return self._migrate_schema(schema_status, False)

    def _migrate_schema_live(
        self, schema_status: migration.SchemaValidationStatus
    ) -> tuple[bool, migration.SchemaValidationStatus]:
        """Migrate schema to the latest version."""
        persistent_notification.create(
            self.hass,
            (
                "System performance will temporarily degrade during the database"
                " upgrade. Do not power down or restart the system until the upgrade"
                " completes. Integrations that read the database, such as logbook,"
                " history, and statistics may return inconsistent results until the"
                " upgrade completes. This notification will be automatically dismissed"
                " when the upgrade completes."
            ),
            "Database upgrade in progress",
            "recorder_database_migration",
        )
        return self._migrate_schema(schema_status, True)

    def _migrate_schema(
        self,
        schema_status: migration.SchemaValidationStatus,
        live: bool,
    ) -> tuple[bool, migration.SchemaValidationStatus]:
        """Migrate schema to the latest version."""
        assert self.engine is not None
        try:
            if live:
                migrator = migration.migrate_schema_live
            else:
                migrator = migration.migrate_schema_non_live
            new_schema_status = migrator(
                self, self.hass, self.engine, self.get_session, schema_status
            )
        except exc.DatabaseError as err:
            if self._handle_database_error(err, setup_run=False):
                # If _handle_database_error returns True, we have a new database
                # which does not need migration or repair.
                new_schema_status = migration.SchemaValidationStatus(
                    current_version=SCHEMA_VERSION,
                    initial_version=SCHEMA_VERSION,
                    migration_needed=False,
                    non_live_data_migration_needed=False,
                    schema_errors=set(),
                    start_version=SCHEMA_VERSION,
                )
                return (True, new_schema_status)
            _LOGGER.exception("Database error during schema migration")
            return (False, schema_status)
        except Exception:
            _LOGGER.exception("Error during schema migration")
            return (False, schema_status)
        else:
            return (True, new_schema_status)

    def _lock_database(self, task: DatabaseLockTask) -> None:
        @callback
        def _async_set_database_locked(task: DatabaseLockTask) -> None:
            task.database_locked.set()

        local_start_time = dt_util.now()
        hass = self.hass
        with write_lock_db_sqlite(self):
            # Notify that lock is being held, wait until database can be used again.
            hass.add_job(_async_set_database_locked, task)
            while not task.database_unlock.wait(timeout=DB_LOCK_QUEUE_CHECK_TIMEOUT):
                if self._reached_max_backlog():
                    _LOGGER.warning(
                        "Database queue backlog reached more than %s events "
                        "while waiting for backup to finish; recorder will now "
                        "resume writing to database. The backup cannot be trusted and "
                        "must be restarted",
                        self.backlog,
                    )
                    task.queue_overflow = True
                    hass.add_job(
                        async_create_backup_failure_issue, self.hass, local_start_time
                    )
                    break
        _LOGGER.info(
            "Database queue backlog reached %d entries during backup",
            self.backlog,
        )

    def _process_one_event(self, event: Event[Any]) -> None:
        if not self.enabled:
            return
        if event.event_type == EVENT_STATE_CHANGED:
            self._process_state_changed_event_into_session(event)
        else:
            self._process_non_state_changed_event_into_session(event)
        # Commit if the commit interval is zero
        if not self.commit_interval:
            self._commit_event_session_or_retry()

    def _process_non_state_changed_event_into_session(self, event: Event) -> None:
        """Process any event into the session except state changed."""
        session = self.event_session
        assert session is not None
        dbevent = Events.from_event(event)

        # Map the event_type to the EventTypes table
        event_type_manager = self.event_type_manager
        if pending_event_types := event_type_manager.get_pending(event.event_type):
            dbevent.event_type_rel = pending_event_types
        elif event_type_id := event_type_manager.get(event.event_type, session, True):
            dbevent.event_type_id = event_type_id
        else:
            event_types = EventTypes(event_type=event.event_type)
            event_type_manager.add_pending(event_types)
            self._add_to_session(session, event_types)
            dbevent.event_type_rel = event_types

        if not event.data:
            self._add_to_session(session, dbevent)
            return

        event_data_manager = self.event_data_manager
        if not (shared_data_bytes := event_data_manager.serialize_from_event(event)):
            return

        # Map the event data to the EventData table
        shared_data = shared_data_bytes.decode("utf-8")
        # Matching attributes found in the pending commit
        if pending_event_data := event_data_manager.get_pending(shared_data):
            dbevent.event_data_rel = pending_event_data
        # Matching attributes id found in the cache
        elif (data_id := event_data_manager.get_from_cache(shared_data)) or (
            (hash_ := EventData.hash_shared_data_bytes(shared_data_bytes))
            and (data_id := event_data_manager.get(shared_data, hash_, session))
        ):
            dbevent.data_id = data_id
        else:
            # No matching attributes found, save them in the DB
            dbevent_data = EventData(shared_data=shared_data, hash=hash_)
            event_data_manager.add_pending(dbevent_data)
            self._add_to_session(session, dbevent_data)
            dbevent.event_data_rel = dbevent_data

        self._add_to_session(session, dbevent)

    def _process_state_changed_event_into_session(
        self, event: Event[EventStateChangedData]
    ) -> None:
        """Process a state_changed event into the session."""
        state_attributes_manager = self.state_attributes_manager
        states_meta_manager = self.states_meta_manager
        entity_removed = not event.data.get("new_state")
        entity_id = event.data["entity_id"]

        dbstate = States.from_event(event)
        old_state = event.data["old_state"]

        assert self.event_session is not None
        session = self.event_session

        states_manager = self.states_manager
        if pending_state := states_manager.pop_pending(entity_id):
            dbstate.old_state = pending_state
            if old_state:
                pending_state.last_reported_ts = old_state.last_reported_timestamp
        elif old_state_id := states_manager.pop_committed(entity_id):
            dbstate.old_state_id = old_state_id
            if old_state:
                states_manager.update_pending_last_reported(
                    old_state_id, old_state.last_reported_timestamp
                )
        if entity_removed:
            dbstate.state = None
        else:
            states_manager.add_pending(entity_id, dbstate)

        if states_meta_manager.active:
            dbstate.entity_id = None

        if entity_id is None or not (
            shared_attrs_bytes := state_attributes_manager.serialize_from_event(event)
        ):
            return

        # Map the entity_id to the StatesMeta table
        if pending_states_meta := states_meta_manager.get_pending(entity_id):
            dbstate.states_meta_rel = pending_states_meta
        elif metadata_id := states_meta_manager.get(entity_id, session, True):
            dbstate.metadata_id = metadata_id
        elif states_meta_manager.active and entity_removed:
            # If the entity was removed, we don't need to add it to the
            # StatesMeta table or record it in the pending commit
            # if it does not have a metadata_id allocated to it as
            # it either never existed or was just renamed.
            return
        else:
            states_meta = StatesMeta(entity_id=entity_id)
            states_meta_manager.add_pending(states_meta)
            self._add_to_session(session, states_meta)
            dbstate.states_meta_rel = states_meta

        # Map the event data to the StateAttributes table
        shared_attrs = shared_attrs_bytes.decode("utf-8")
        # Matching attributes found in the pending commit
        if pending_event_data := state_attributes_manager.get_pending(shared_attrs):
            dbstate.state_attributes = pending_event_data
        # Matching attributes id found in the cache
        elif (
            attributes_id := state_attributes_manager.get_from_cache(shared_attrs)
        ) or (
            (hash_ := StateAttributes.hash_shared_attrs_bytes(shared_attrs_bytes))
            and (
                attributes_id := state_attributes_manager.get(
                    shared_attrs, hash_, session
                )
            )
        ):
            dbstate.attributes_id = attributes_id
        else:
            # No matching attributes found, save them in the DB
            dbstate_attributes = StateAttributes(shared_attrs=shared_attrs, hash=hash_)
            state_attributes_manager.add_pending(dbstate_attributes)
            self._add_to_session(session, dbstate_attributes)
            dbstate.state_attributes = dbstate_attributes

        self._add_to_session(session, dbstate)

    def _handle_database_error(self, err: Exception, *, setup_run: bool) -> bool:
        """Handle a database error that may result in moving away the corrupt db."""
        if (
            (cause := err.__cause__)
            and isinstance(cause, sqlite3.DatabaseError)
            and (cause_str := str(cause))
            # Make sure we do not move away a database when its only locked
            # externally by another process. sqlite does not give us a named
            # exception for this so we have to check the error message.
            and ("malformed" in cause_str or "not a database" in cause_str)
        ):
            _LOGGER.exception(
                "Unrecoverable sqlite3 database corruption detected: %s", err
            )
            self._handle_sqlite_corruption(setup_run)
            return True
        return False

    def _commit_event_session_or_retry(self) -> None:
        """Commit the event session if there is work to do."""
        if not self._event_session_has_pending_writes:
            return
        tries = 1
        while tries <= self.db_max_retries:
            try:
                self._commit_event_session()
            except (exc.InternalError, exc.OperationalError) as err:
                _LOGGER.error(
                    "%s: Error executing query: %s. (retrying in %s seconds)",
                    INVALIDATED_ERR if err.connection_invalidated else CONNECTIVITY_ERR,
                    err,
                    self.db_retry_wait,
                )
                if tries == self.db_max_retries:
                    raise

                tries += 1
                time.sleep(self.db_retry_wait)
            else:
                return

    def _commit_event_session(self) -> None:
        assert self.event_session is not None
        session = self.event_session
        self._commits_without_expire += 1

        if (
            pending_last_reported
            := self.states_manager.get_pending_last_reported_timestamp()
        ) and self.schema_version >= LAST_REPORTED_SCHEMA_VERSION:
            with session.no_autoflush:
                session.execute(
                    update(States),
                    [
                        {
                            "state_id": state_id,
                            "last_reported_ts": last_reported_timestamp,
                        }
                        for state_id, last_reported_timestamp in pending_last_reported.items()
                    ],
                )
        session.commit()

        self._event_session_has_pending_writes = False
        # We just committed the state attributes to the database
        # and we now know the attributes_ids.  We can save
        # many selects for matching attributes by loading them
        # into the LRU or committed now.
        self.states_manager.post_commit_pending()
        self.state_attributes_manager.post_commit_pending()
        self.event_data_manager.post_commit_pending()
        self.event_type_manager.post_commit_pending()
        self.states_meta_manager.post_commit_pending()

        # Expire is an expensive operation (frequently more expensive
        # than the flush and commit itself) so we only
        # do it after EXPIRE_AFTER_COMMITS commits
        if self._commits_without_expire >= EXPIRE_AFTER_COMMITS:
            self._commits_without_expire = 0
            session.expire_all()

    def _handle_sqlite_corruption(self, setup_run: bool) -> None:
        """Handle the sqlite3 database being corrupt."""
        try:
            self._close_event_session()
        finally:
            self._close_connection()
        move_away_broken_database(dburl_to_path(self.db_url))
        self.recorder_runs_manager.reset()
        self._setup_recorder()
        if setup_run:
            self._setup_run()

    def _close_event_session(self) -> None:
        """Close the event session."""
        self.states_manager.reset()
        self.state_attributes_manager.reset()
        self.event_data_manager.reset()
        self.event_type_manager.reset()
        self.states_meta_manager.reset()
        self.statistics_meta_manager.reset()

        if not self.event_session:
            return

        try:
            self.event_session.rollback()
            self.event_session.close()
        except SQLAlchemyError:
            _LOGGER.exception("Error while rolling back and closing the event session")

    def _reopen_event_session(self) -> None:
        """Rollback the event session and reopen it after a failure."""
        self._close_event_session()
        self._open_event_session()

    def _open_event_session(self) -> None:
        """Open the event session."""
        self.event_session = self.get_session()
        self.event_session.expire_on_commit = False

    def _send_keep_alive(self) -> None:
        """Send a keep alive to keep the db connection open."""
        assert self.event_session is not None
        _LOGGER.debug("Sending keepalive")
        self.event_session.connection().scalar(select(1))

    async def async_block_till_done(self) -> None:
        """Async version of block_till_done."""
        if self._queue.empty() and not self._event_session_has_pending_writes:
            return
        event = asyncio.Event()
        self.queue_task(SynchronizeTask(event))
        await event.wait()

    def block_till_done(self) -> None:
        """Block till all events processed.

        This is only called in tests.

        This only blocks until the queue is empty
        which does not mean the recorder is done.

        Call tests.common's wait_recording_done
        after calling this to ensure the data
        is in the database.
        """
        self._queue_watch.clear()
        self.queue_task(WAIT_TASK)
        self._queue_watch.wait()

    async def lock_database(self) -> bool:
        """Lock database so it can be backed up safely."""
        if self.dialect_name != SupportedDialect.SQLITE:
            _LOGGER.debug(
                "Not a SQLite database or not connected, locking not necessary"
            )
            return True

        if self._database_lock_task:
            _LOGGER.warning("Database already locked")
            return False

        database_locked = asyncio.Event()
        task = DatabaseLockTask(database_locked, threading.Event(), False)
        self.queue_task(task)
        try:
            async with asyncio.timeout(DB_LOCK_TIMEOUT):
                await database_locked.wait()
        except TimeoutError as err:
            task.database_unlock.set()
            raise TimeoutError(
                f"Could not lock database within {DB_LOCK_TIMEOUT} seconds."
            ) from err
        self._database_lock_task = task
        return True

    @callback
    def unlock_database(self) -> bool:
        """Unlock database.

        Returns true if database lock has been held throughout the process.
        """
        if self.dialect_name != SupportedDialect.SQLITE:
            _LOGGER.debug(
                "Not a SQLite database or not connected, unlocking not necessary"
            )
            return True

        if not self._database_lock_task:
            _LOGGER.warning("Database currently not locked")
            return False

        self._database_lock_task.database_unlock.set()
        success = not self._database_lock_task.queue_overflow

        self._database_lock_task = None

        return success

    def _setup_recorder_connection(
        self, dbapi_connection: DBAPIConnection, connection_record: Any
    ) -> None:
        """Dbapi specific connection settings."""
        assert self.engine is not None
        if database_engine := setup_connection_for_dialect(
            self,
            self.engine.dialect.name,
            dbapi_connection,
            not self._completed_first_database_setup,
        ):
            self.database_engine = database_engine
            self.max_bind_vars = database_engine.max_bind_vars
        self._completed_first_database_setup = True

    def _setup_connection(self) -> None:
        """Ensure database is ready to fly."""
        kwargs: dict[str, Any] = {}
        self._completed_first_database_setup = False

        if self.db_url == SQLITE_URL_PREFIX or ":memory:" in self.db_url:
            kwargs["connect_args"] = {"check_same_thread": False}
            kwargs["poolclass"] = MutexPool
            MutexPool.pool_lock = threading.RLock()
            kwargs["pool_reset_on_return"] = None
        elif self.db_url.startswith(SQLITE_URL_PREFIX):
            kwargs["poolclass"] = RecorderPool
            kwargs["recorder_and_worker_thread_ids"] = (
                self.recorder_and_worker_thread_ids
            )
        elif self.db_url.startswith(
            (
                MARIADB_URL_PREFIX,
                MARIADB_PYMYSQL_URL_PREFIX,
                MYSQLDB_URL_PREFIX,
                MYSQLDB_PYMYSQL_URL_PREFIX,
            )
        ):
            kwargs["connect_args"] = {"charset": "utf8mb4"}
            if self.db_url.startswith((MARIADB_URL_PREFIX, MYSQLDB_URL_PREFIX)):
                # If they have configured MySQLDB but don't have
                # the MySQLDB module installed this will throw
                # an ImportError which we suppress here since
                # sqlalchemy will give them a better error when
                # it tried to import it below.
                with contextlib.suppress(ImportError):
                    kwargs["connect_args"]["conv"] = build_mysqldb_conv()

        # Disable extended logging for non SQLite databases
        if not self.db_url.startswith(SQLITE_URL_PREFIX):
            kwargs["echo"] = False

        if self._using_file_sqlite:
            validate_or_move_away_sqlite_database(self.db_url)

        assert not self.engine
        self.engine = create_engine(self.db_url, **kwargs, future=True)
        self._dialect_name = try_parse_enum(SupportedDialect, self.engine.dialect.name)
        self.__dict__.pop("dialect_name", None)
        sqlalchemy_event.listen(self.engine, "connect", self._setup_recorder_connection)

        migration.pre_migrate_schema(self.engine)
        Base.metadata.create_all(self.engine)
        self._get_session = scoped_session(sessionmaker(bind=self.engine, future=True))
        _LOGGER.debug("Connected to recorder database")

    def _close_connection(self) -> None:
        """Close the connection."""
        if self.engine:
            self.engine.dispose()
            self.engine = None
        self._get_session = None

    def _setup_run(self) -> None:
        """Log the start of the current run and schedule any needed jobs."""
        with session_scope(session=self.get_session()) as session:
            end_incomplete_runs(session, self.recorder_runs_manager.recording_start)
            self.recorder_runs_manager.start(session)
            self.states_manager.load_from_db(session)

        self._open_event_session()

    def _schedule_compile_missing_statistics(self) -> None:
        """Add tasks for missing statistics runs."""
        self.queue_task(CompileMissingStatisticsTask())

    def _end_session(self) -> None:
        """End the recorder session."""
        if self.event_session is None:
            return
        if self.recorder_runs_manager.active:
            # .end will add to the event session
            self._event_session_has_pending_writes = True
            self.recorder_runs_manager.end(self.event_session)
        try:
            self._commit_event_session_or_retry()
        except Exception:
            _LOGGER.exception("Error saving the event session during shutdown")

        self.event_session.close()
        self.recorder_runs_manager.clear()

    def _shutdown(self) -> None:
        """Save end time for current run."""
        _LOGGER.debug("Shutting down recorder")

        # If the schema version is not set, we never had a working
        # connection to the database or the schema never reached a
        # good state.
        # In either case, we want to mark startup as failed.
        startup_failed = (
            not self.schema_version or self.schema_version != SCHEMA_VERSION
        )
        self.hass.add_job(self._async_startup_done, startup_failed)

        try:
            self._end_session()
        finally:
            if self._db_executor:
                # We shutdown the executor without forcefully
                # joining the threads until after we have tried
                # to cleanly close the connection.
                self._db_executor.shutdown(join_threads_or_timeout=False)
            self._close_connection()
            if self._db_executor:
                # After the connection is closed, we can join the threads
                # or forcefully shutdown the threads if they take too long.
                self._db_executor.join_threads_or_timeout()