Data retention

The Settings → Data retention page controls the one retention knob a tenant owns: the purge grace window. Viewing requires access to Settings; changing the value requires the admin role.

The purge grace window

When a selector leaves your demand set — you delete a dashboard panel, an alert, or an ingestion rule — the telemetry it demanded stops being forwarded immediately, but its already-stored data is not deleted on the spot. It is purged after the grace window, unless another live filter still demands it (filters are unioned across demands — data is only ever purged when nothing asks for it anymore).

The default is 24 hours; the minimum is 1 minute.

Longer window — more time to notice and undo a removal before its history is gone: restore the dashboard or alert inside the window and the data was never touched. Shorter window — storage is reclaimed sooner, which matters when you are close to a storage limit.

You can watch each removal’s outcome on the Demands → Purge tab — pending while inside the grace window, then deleted or kept because another demand still covers it. See Purge: what happens to stored data.

What this setting does not change

• Demanded data is never deleted. The grace window applies only to data that has fallen out of demand. As long as any live filter matches a series or stream, it stays — regardless of how long ago it last received a data point.
• The 30-day central retention is fixed. Demanded data ages out of the central stores after 30 days (metrics, logs, and traces alike); the grace window does not extend or shorten that. See Limits.
• Edge buffers are fixed too — about a day of metrics and an hour of logs/traces at the agent, by design.