What Causes Home Assistant Metadata and History Growth During Whole-Home Control?

Home Assistant “metadata growth” is easy to misdiagnose because several different data classes live around the same installation. Device and entity registries preserve identities and configuration relationships; the config tree stores UI-managed state; Recorder stores the much larger time series of state changes and events; long-term statistics keep selected aggregates beyond raw-history retention.

Whole-home control increases these layers in different ways. Adding devices grows registries slowly, while adding high-frequency sensors or attribute-heavy entities can grow the database rapidly. Before changing retention or deleting files, identify which layer is actually increasing.

Entity and Device Registries Grow With Managed Objects

Home Assistant keeps durable registry records so an entity can retain its identity, user customizations, device relationship, area assignment, and integration ownership across restarts. That metadata is not the same thing as every historical sensor sample.

The current device-registry model describes how devices persist relationships to config entries and the entities that represent their functions. As the house adds integrations, bridges, child devices, and entities, this registry naturally becomes more complex.

Registry growth is usually modest compared with Recorder. A thousand entity definitions are operationally important, but a thousand entities each producing hundreds or thousands of historical rows can dominate storage.

Recorder Growth Is Driven by Change Rate, Not Device Count Alone

Recorder writes state changes and selected events. A door contact that changes twice a day can be cheaper than one power sensor that reports every few seconds, even though both count as one entity on the dashboard.

A 2026 tuning case found that a Home Assistant database reached 963 MB in six days before excluding noisy entities cut daily growth from about 160 MB to under 50 MB. The exact numbers are installation-specific; the mechanism is not.

Measure database growth per day and rank the entities or domains creating the most rows before applying a global retention cut. Preserve the history the household actually uses.

Attributes Can Add More Storage Than the Visible State Suggests

An entity may show a short state such as on, 23.4, or home while carrying a much larger attribute set with device details, forecasts, lists, coordinates, or diagnostic metadata.

Current Home Assistant developer guidance explicitly warns that entities with frequent state changes can grow the database quickly when extra_state_attributes also change often. The recommended direction is to minimize non-critical attributes or expose independent sensor entities instead.

Do not estimate storage from the visible entity state alone. Inspect both state frequency and attribute churn, especially for integrations that expose large JSON-like structures.

Featured

ZimaCube 2 Personal Cloud Home NAS

ZimaCube2

$799.00 - $2,499.00

Buy Now

Statistics Create a Different Long-Term Retention Curve

Raw history is normally bounded by retention, but long-term statistics can preserve aggregates for selected sensors far longer. That is useful for energy, temperature, and utility trends because the system does not need every raw sample to answer a monthly question.

This means deleting old raw states does not necessarily remove all historical data, and that is intentional. Treat recent troubleshooting history and long-term analytical history as separate retention products.

ZimaSpace's sensor-retention model shows why sample frequency, indexes, rollups, and backup generations must be measured separately rather than reduced to bytes per sensor.

Backups Multiply Whatever the Live System Retains

A larger Recorder database increases backup size and restore time. Multiple retained backups can therefore consume more capacity than the current live database, especially when each archive contains a full copy.

A community Recorder guide notes that high-update entities and large attributes are common causes of an ever-growing Home Assistant database.

Set retention for both live history and backups. Reducing the active database does not reclaim space from old immutable backup archives until those backups expire.

Use a Data-Role Audit Before Deleting Anything

Ask four separate questions: are obsolete device or entity records accumulating; which entities dominate raw Recorder changes; which sensors legitimately need long-term statistics; and how many full backup generations are multiplying the live data footprint?

Growth is healthy when it corresponds to useful devices, history, or analytics and remains inside a planned maintenance and recovery window. It becomes a problem when a small number of noisy entities, stale registries, or unnecessary backup generations consume most of the capacity.

FAQ

Is Home Assistant metadata the same as Recorder history?

No. Registry and configuration metadata describe devices, entities, integrations, and UI-managed state. Recorder history is a time series of state changes and events and is usually the much larger storage layer.

Does adding more devices always make the database grow quickly?

No. Change rate matters more than device count alone. A few high-frequency or attribute-heavy entities can generate more history than many low-activity switches and contact sensors.