Highlights
- Reduces memory usage during backups of large DOCSight databases by writing archives through disk-backed temporary files instead of buffering complete archives in memory.
- Prevents long-running scheduled backups from being started repeatedly when they exceed the previous 120-second observation window.
Special thanks
- Thanks to @saschabrockel for reporting the problem and providing detailed diagnostics in #738, which helped confirm the cause.
Changelog
Fixed
- Published scheduled backups atomically and removed partial temporary archives after failures.
- Streamed manual backup downloads from disk and cleaned their temporary files when responses close.
- Kept long-running collector jobs tracked until their actual result is processed.
- Prevented collectors with the same logical name from overlapping during configuration changes.
Large backups now use temporary disk space on the DOCSight data volume rather than holding an equivalent archive in RAM. Ensure the data volume has sufficient free space while a backup is running.
Docker image
ghcr.io/itsdnns/docsight:v2026-07-28.1ghcr.io/itsdnns/docsight:latest
Full diff: v2026-07-25.1...v2026-07-28.1