Can InfluxDB schema changes be reverted without re-ingesting historical data?
0 reputation · 21 Feb 2022, 06:00 UTC
Goal
Determine whether a field or tag key modification in InfluxDB can be rolled back to its prior state without requiring a full historical data re-ingestion or backup restoration.
Constraints and uncertainty
InfluxDB treats schema as per-point dynamic field and tag keys; adding or removing keys does not rewrite existing points. The platform lacks a native transactional rollback for schema-level changes. Retention policies control visibility and compaction but cannot reverse metadata alterations once keys have been written. Flux schema export/import and UI-driven migration support manual re-creation of field/tag configurations, yet restoring the exact prior state appears to depend on re-ingesting data or restoring from backups. Version differences across 1.x, 2.x, and 3.x further complicate behavior, especially for continuous queries and task definitions that may not survive field key removals.
Questions
- Does the schema export subcommand capture enough metadata to reconstruct a pre-change state without the original raw points?
- Can a retention policy adjustment alone make removed keys inaccessible to queries while preserving them for potential recovery?
- Which InfluxDB versions, if any, provide a built-in mechanism to atomically revert a field key addition or type change?