Skip to content

Storage Layout

The deployed host deliberately keeps raw mirrored inputs and derived products in separate trees.

/project/aurora

  • Function: raw mirrored source data
  • What lives there: synced instrument files coming from the remote source machines
  • Examples:
  • /project/aurora/raw/cl61
  • /project/aurora/raw/rpgfmcw94
  • /project/aurora/raw/vaisalamet
  • /project/aurora/raw/asfs/crd
  • /project/aurora/raw/power/level1
  • /project/aurora/raw/wxcam
  • Storage type: shared Ceph network filesystem
  • Current filesystem size on 2026-05-21: 4.0T
  • Current used on 2026-05-21: 57G
  • Current available on 2026-05-21: 3.9T

So /project/aurora is the raw landing and mirror area.

/data/aurora

  • Function: processed products and dashboard-serving outputs
  • What lives there:
  • Zarr stores
  • quicklook PNGs
  • WXcam catalog SQLite
  • WXcam daily videos and thumbnails
  • performance logs and other dashboard products
  • Examples:
  • /data/aurora/products/cl61/...zarr
  • /data/aurora/products/rpgfmcw94/cloud_radar.zarr
  • /data/aurora/products/quicklooks/...
  • /data/aurora/products/wxcam/...
  • Storage type: local disk on /dev/vdb
  • Current filesystem size on 2026-05-21: 983G
  • Current used on 2026-05-21: 262G
  • Current available on 2026-05-21: 672G

So /data/aurora is the product, work, and output area.

Why the split matters

  • raw files stay separate from regenerated products
  • products can be deleted and rebuilt without touching the source mirror
  • the dashboard reads smaller processed artifacts from local disk instead of always working directly from the raw mirror