recorder integration is responsible for storing details in a database, which then are handled by the
This integration constantly saves data. If you use the default configuration, the data will be saved on the media Home Assistant is installed on. In case of Raspberry Pi with an SD card, it might affect your system’s reaction time and life expectancy of the storage medium (the SD card). It is therefore recommended to set the commit_interval to higher value, e.g. 30s, limit the amount of stored data (e.g., by excluding devices) or store the data elsewhere (e.g., another system).
Home Assistant uses SQLAlchemy, which is an Object Relational Mapper (ORM). This means that you can use any SQL backend for the recorder that is supported by SQLAlchemy, like MySQL, MariaDB, PostgreSQL, or MS SQL Server.
The default database engine is SQLite which does not require any configuration. The database is stored in your Home Assistant configuration directory (’/config/’) and is named
To change the defaults for the
recorder integration in your installation, add the following to your
# Example configuration.yaml entry recorder:
Enables the recorder integration. Only allowed once.
The max amount of times, the recorder retries to connect to the database.
The time in seconds, that the recorder sleeps when trying to connect to the database.
When using SQLite, if Home Assistant does not restart cleanly or the recorder fails to shut down, a
quick_check is performed on the database to ensure it is usable. If
db_integrity_check is disabled, the system will only perform necessary sanity checks and skip the
quick_check. Home Assistant may not be able to automatically recover and start the recorder in the event the database is malformed if
db_integrity_check is disabled.
Automatically purge the database every night at 04:12 local time. Purging keeps the database from growing indefinitely, which takes up disk space and can make Home Assistant slow. If you disable
auto_purge it is recommended that you create an automation to call the
Specify the number of history days to keep in recorder database after a purge.
How often (in seconds) the events and state changes are committed to the database. The default of
1 allows events to be committed almost right away without trashing the disk when an event storm happens. Increasing this will reduce disk I/O and may prolong disk (SD card) lifetime with the trade-off being that the logbook and history will lag. If this is set to
0 (zero), commit are made as soon as possible after an event is processed.
Configure which integrations should be excluded from recordings. (Configure Filter)
Exclude all entities matching a listed pattern from recordings (e.g.,
Configure which integrations should be included in recordings. If set, all other entities will not be recorded. (Configure Filter)
By default, no entity will be excluded. To limit which entities are being exposed to
recorder, you can use the
# Example filter to include specified domains and exclude specified entities recorder: include: domains: - alarm_control_panel - light entity_globs: - binary_sensor.*_occupancy exclude: entities: - light.kitchen_light
Filters are applied as follows:
- No includes or excludes - pass all entities
- Includes, no excludes - only include specified entities
- Excludes, no includes - only exclude specified entities
- Both includes and excludes:
- Include domain and/or glob patterns specified
- If domain is included, and entity not excluded or match exclude glob pattern, pass
- If entity matches include glob pattern, and entity does not match any exclude criteria (domain, glob pattern or listed), pass
- If domain is not included, glob pattern does not match, and entity not included, fail
- Exclude domain and/or glob patterns specified and include does not list domains or glob patterns
- If domain is excluded and entity not included, fail
- If entity matches exclude glob pattern and entity not included, fail
- If entity does not match any exclude criteria (domain, glob pattern or listed), pass
- Neither include or exclude specifies domains or glob patterns
- If entity is included, pass (as #2 above)
- If entity include and exclude, the entity exclude is ignored
- Include domain and/or glob patterns specified
If you only want to hide events from your history, take a look at the
history integration. The same goes for the logbook. But if you have privacy concerns about certain events or want them in neither the history or logbook, you should use the
include options of the
recorder integration. That way they aren’t even in your database, you can reduce storage and keep the database small by excluding certain often-logged events (like
Defining domains and entities to
exclude (i.e. blocklist) is convenient when you are basically happy with the information recorded, but just want to remove some entities or domains.
# Example configuration.yaml entry with exclude recorder: purge_keep_days: 5 db_url: sqlite:////home/user/.homeassistant/test exclude: domains: - automation - updater entity_globs: - sensor.weather_* entities: - sun.sun # Don't record sun data - sensor.last_boot # Comes from 'systemmonitor' sensor platform - sensor.date event_types: - call_service # Don't record service calls
Defining domains and entities to record by using the
include configuration (i.e. allowlist) is convenient if you have a lot of entities in your system and your
exclude lists possibly get very large, so it might be better just to define the entities or domains to record.
# Example configuration.yaml entry with include recorder: include: domains: - sensor - switch - media_player
You can also use the
include list to define the domains/entities to record, and exclude some of those within the
exclude list. This makes sense if you, for instance, include the
sensor domain, but want to exclude some specific sensors. Instead of adding every sensor entity to the
entities list just include the
sensor domain and exclude the sensor entities you are not interested in.
# Example configuration.yaml entry with include and exclude recorder: include: domains: - sensor - switch - media_player exclude: entities: - sensor.last_boot - sensor.date entity_globs: - sensor.weather_*
Call the service
recorder.purge to start a purge task which deletes events and states older than x days, according to
keep_days service data.
Note that purging will not immediately decrease disk space usage but it will significantly slow down further growth.
|Service data attribute||Optional||Description|
||yes||The number of history days to keep in recorder database (defaults to the integration
||yes||When using SQLite or PostgreSQL this will rewrite the entire database. When using MySQL or MariaDB it will optimize or recreate the events and states tables. This is a heavy operation that can cause slowdowns and increased disk space usage while it runs. Only supported by SQLite, PostgreSQL, MySQL and MariaDB.|
||yes||Apply entity_id and event_type filter in addition to time based purge. Useful in combination with
Call the service
recorder.disable to stop saving events and states to the database.
Call the service
recorder.enable to start again saving events and states to the database. This is the opposite of
|MariaDB (omit pymysql)||
|MariaDB (omit pymysql, Socket)||
|PostgreSQL (Custom socket dir)||
|MS SQL Server||
Some installations of MariaDB/MySQL may require an ALTERNATE_PORT (3rd-party hosting providers or parallel installations) to be added to the SERVER_IP, e.g.,
When using a MariaDB or MySQL server, adding
+pymysql to the URL will use the pure Python MySQL library, which is slower but may be required if the C MySQL library is not available.
When using the official Docker image, the C MySQL library will always be available.
pymysql is most commonly used with
venv where the C MySQL library is not installed.
Unix Socket connections always bring performance advantages over TCP, if the database is on the same host as the
recorder instance (i.e.,
If you want to use Unix Sockets for PostgreSQL you need to modify the
pg_hba.conf. See PostgreSQL
If you are using the default
FULL recovery model for MS SQL Server you will need to manually backup your log file to prevent your transaction log from growing too large. It is recommended you change the recovery model to
SIMPLE unless you are worried about data loss between backups.
If you are running a database server instance on the same server as Home Assistant then you must ensure that this service starts before Home Assistant. For a Linux instance running Systemd (Raspberry Pi, Debian, Ubuntu and others) you should edit the service file. To help facilitate this, db_max_retry and db_retry_wait variables have been added to ensure the recorder retries the connection to your database enough times, for your database to start up.
sudo nano /etc/systemd/system/[email protected]
and add the service for the database, for example, PostgreSQL:
[Unit] Description=Home Assistant After=network.target postgresql.service
Save the file then reload
sudo systemctl daemon-reload
Not all Python bindings for the chosen database engine can be installed directly. This section contains additional details that should help you to get it working.
If you are in a virtual environment, don’t forget to activate it before installing the
mysqlclient Python package described below.
[email protected]:~ $ sudo -u homeassistant -H -s [email protected]:~$ source /srv/homeassistant/bin/activate (homeassistant) [email protected]:~$ pip3 install mysqlclient
For MariaDB you may have to install a few dependencies. If you’re using MariaDB version 10.2,
libmariadbclient-dev was renamed to
libmariadb-dev. If you’re using MariaDB 10.3, the package
libmariadb-dev-compat must also be installed. For MariaDB v10.0.34 only
libmariadb-dev-compat is needed. Please install the correct packages based on your MariaDB version.
On the Python side we use the
sudo apt-get install libmariadbclient-dev libssl-dev pip3 install mysqlclient
For MySQL you may have to install a few dependencies. You can choose between
sudo apt-get install default-libmysqlclient-dev libssl-dev pip3 install mysqlclient
After installing the dependencies, it is required to create the database manually. During the startup, Home Assistant will look for the database specified in the
db_url. If the database doesn’t exist, it will not automatically create it for you.
Once Home Assistant finds the database, with the right level of permissions, all the required tables will then be automatically created and the data will be populated accordingly.
Create the PostgreSQL database with
utf8 encoding. The PostgreSQL default encoding is
SQL_ASCII. From the
postgres user account;
createdb -E utf8 DB_NAME
DB_NAME is the name of your database
If the Database in use is not
?client_encoding=utf8 to the
db_url may solve any issue.
For PostgreSQL you may have to install a few dependencies:
sudo apt-get install postgresql-server-dev-X.Y pip3 install psycopg2
For using Unix Sockets, add the following line to your
local DB_NAME USER_NAME peer
DB_NAME is the name of your database and
USER_NAME is the name of the user running the Home Assistant instance (see securing your installation).
Reload the PostgreSQL configuration after that:
$ sudo -i -u postgres psql -c "SELECT pg_reload_conf();" pg_reload_conf ---------------- t (1 row)
A service restart will work as well.
For MS SQL Server you will have to install a few dependencies:
sudo apt-get install unixodbc-dev pip3 install pyodbc
If you are in a virtual environment, don’t forget to activate it before installing the pyodbc package.
sudo -u homeassistant -H -s source /srv/homeassistant/bin/activate pip3 install pyodbc
You will also need to install an ODBC Driver. Microsoft ODBC drivers are recommended, however FreeTDS is available for systems that are not supported by Microsoft. Instructions for installing the Microsoft ODBC drivers can be found here.
If you are using Hass.io, FreeTDS is already installed for you. The db_url you need to use is
Suggest an edit to this page, or provide/view feedback for this page.