disk_to_sds

(formerly sdcard_to_sds; renamed 2026-05-23 as the scope generalises beyond Gecko SD cards to multiple recorder types and on-disk layouts. The Gecko SD-card path is one adapter.)

Architecture: see docs/ARCHITECTURE.md for the canonical pipeline (Pull on Mac → Plan/Apply on VM → Cleanup), the gap-fill decision logic, and the durable manifests in manifests/.

Processing a card: the step-by-step runbook — identify → pull → apply → cleanup → wipe, with the exact commands and the safety gates — is the Card upload runbook in CLAUDE.md. Just tell Claude "upload" (or "process this card") and it will drive those stages, pausing for your OK before the two irreversible steps (writing to the long-term archive, and wiping the card).

Seismic Upload Workflow (SD Cards → VM → Long-Term SDS)

This repository does not store seismic data.
It only tracks:

• The commands used to ingest data from SD cards
• A log of uploads (what was ingested, when, from where)
• Simple helper scripts (in scripts/)

All actual waveform data live in:

• inbox/ (temporary, shared mount; not tracked by git)
• A local SDS archive on the VM
• The long-term SDS repository (rsync target)

---

Directory layout

• inbox/
  Temporary location for raw data copied from SD cards.
  This is on the shared/mounted disk, visible to both Mac and VM, and ignored by git.

• scripts/
  Helper scripts, e.g. process_inbox_example.sh, showing how to:

  • run scart into a local SDS archive
  • rsync that SDS into the long-term repository

• README.md
  This file. Contains:

  • Workflow notes
  • A log of each upload session

---

Example usage

Gecko SD card → SDS in one local step (recommended)

For Gecko-format SD cards (data/YYYY/MM/DD/HH/*.ms) the unit already writes correct NET.STA.LOC.CHA codes in every MiniSEED record, so no remapping is needed. scripts/gecko_sdcard_to_sds.py streams 512-byte records straight from the SD card and routes each into its SDS day file — no decoding, no intermediate concat, no VM round-trip.

./scripts/gecko_sdcard_to_sds.py "/Volumes/NO NAME/data" ./local_sds

Then rsync the resulting local_sds/ into the staging mount (Mac writes to staging only, never to the long-term archive):

rsync -avh --progress ./local_sds/ /Volumes/proj-6700_seiscomp_staging-1128.4.1649/seiscomp_archive/

The script:

• walks YYYY/MM/DD/HH/*.ms on the SD card
• reads each record's SEED header (net/sta/loc/cha) and appends it to local_sds/YYYY/NET/STA/CHA.D/NET.STA.LOC.CHA.D.YYYY.DOY
• works with any Python that has no external deps (uses stdlib only)
• typical throughput: ~150 MB/s reading the SD card on a Mac

If the Gecko was misconfigured and the location code needs fixing, do it once at the end with scmssort / scart on the VM, or just rename files in the SDS tree.

Legacy workflow (concat → rsync → scart on VM)

Kept for reference; only useful if the Gecko codes are wrong and you need scart's --rename to fix them.

./scripts/merge_days_gecko.sh /Volumes/BGT2/data ./inbox/BGT2/
rsync -avh --progress inbox/BGT2 /Volumes/proj-6700_seiscomp_staging-1128.4.1649/inbox

# on VM:
for f in /mnt/seiscomp_staging/inbox/BGT2/*.mseed; do
    cat "$f" | scmssort -u -E 2>/dev/null \
      | scart -I - --rename "Z1.BGT2.*.*:Z1.BGT2.00.-" \
            /mnt/seiscomp_staging/seiscomp_archive
done