Data retention in InfluxDB

The InfluxDB retention enforcement service checks for and removes data with timestamps beyond the defined retention period of the bucket the data is stored in. This service is designed to automatically delete “expired” data and optimize disk usage without any user intervention.

By default, the retention enforcement service runs every 30 minutes. You can configure this interval with the storage-retention-check-interval configuration option.

• Bucket retention period
• Shard group duration
• When does data actually get deleted?

Bucket retention period

A bucket retention period is the duration of time that a bucket retains data. Retention periods can be as short as an hour or infinite. Points in a bucket with timestamps beyond the defined retention period (relative to now) are eligible for deletion.

Shard group duration

InfluxDB stores data on disk in shards. Each shard belongs to a shard group and each shard group has a shard group duration. The shard group duration defines the duration of time that each shard in the shard group covers. Each shard contains only points with timestamps in a specific time range defined by the shard group duration.

By default, shard group durations are set automatically based on the bucket retention period, but can also be explicitly defined when creating or updating a bucket.

For more information, see InfluxDB shard group duration.

When does data actually get deleted?

The InfluxDB retention enforcement service runs at regular intervals and deletes shard groups, not individual points. The service will only delete a shard group when the entire time range covered by the shard group is beyond the bucket retention period.

To calculate the possible time data will persist before being deleted:

• minimum: bucket-retention-period
• maximum bucket-retention-period + shard-group-duration

For example, if your bucket retention period is three days (3d) and your shard group duration is one day (1d), the retention enforcement service deletes all shard groups with data that is three to four days old the next time the service runs.