# Data loss in Influxdb3 database, but files still available **URL:** \ **Category:** InfluxDB 3\ **Created:** [November 4, 2026, 9:44am UTC](https://community.influxdata.com/t/data-loss-in-influxdb3-database-but-files-still-available/58823 "2026-11-04T09:44:16Z")\ **Posts on this page:** 1\ **Page:** 1
**Author:** ![Frank\_Walburg](https://sea1.discourse-cdn.com/flex023/user_avatar/community.influxdata.com/frank_walburg/32/20564_2.png) [@Frank\_Walburg](https://community.influxdata.com/u/Frank_Walburg)\ **Post date:** [November 4, 2026, 9:44am UTC](https://community.influxdata.com/t/data-loss-in-influxdb3-database-but-files-still-available/58823/1 "2026-11-04T09:44:16Z")
**InfluxDB 3 Enterprise 3.12.2 (Home license), single node, PachaTree: lost V2 checkpoints, need to rebuild compactor lineage** A crash-consistent Proxmox VM backup (no guest agent, so no fs-freeze) left 0-byte files from the moment of the snapshot: all three `cluster0/cv2/checkpoints/*.ptv2chk.lz4`, several `pt_snapshots/*.ptsnap`, `gen0/*.pt`, `pt_wal/*.pt` and a few `cv2/windows/.../run_sets/*.idx` / `files/*.pt`. The same pattern exists in every nightly backup. Startup failed with `Failed to decompress V2 checkpoint: expected another byte, found none`. After moving all 0-byte files aside: - `--mode=all` starts with “no checkpoint found, starting empty” and queries return `num_rows=0`. The orphan sweep (`mode=armed`) then ran with `checkpoint_sequence=0, reachable_files=0, live_run_sets=0` and listed almost all `cv2/windows` files (data back to Feb 2026) as candidates. Nothing has been deleted; I stopped the server and moved that sweep\_audit pass aside. - `--mode=ingest,query,process` runs and new writes are queryable, but all historical data is invisible. All ~4,400 compacted files (`PACH` magic, ~26 GB) are still on disk. 1. How can I rebuild the V2 checkpoint/lineage from the existing `cv2/windows` files? 2. How can I safely disable the orphan sweep until then? 3. Is `INFLUXDB3_UPGRADE_PACHA_TREE=true` still needed after migration?