Produces an Odoo-compatible <db>.zip restorable via /web/database/manager
on the live VPS (or curl -F backup_file=@... /web/database/restore).
Why this script instead of POST /web/database/backup:
the Odoo HTTP backup endpoint shells out to pg_dump from $PATH and on
this host that returns "Command pg_dump not found" because pg_dump
lives in the conda env that Odoo wasn't launched from. This script
calls pg_dump directly with PGBIN, copies the filestore, builds a
correct manifest.json (odoo_dump, db_name, version, pg_version,
modules), and zips the three artefacts in the exact layout Odoo's
restore code expects.
Layout matches Odoo's own backup output:
<db>.zip
|- dump.sql (pg_dump --no-owner --format=p)
|- manifest.json (odoo_dump=1, version, pg_version, installed modules)
|- filestore/ (data_dir/filestore/<db>/ contents)
Defaults are encoach_v2 / yamenahmad / /tmp/odoo-data, all overridable
via env vars (DB, DB_USER, DATA_DIR, PGBIN, OUT_DIR). Output filename
is encoach_v2_YYYYMMDD_HHMMSS.zip in ./backups/.
Verified: latest run produced encoach_v2_20260425_031949.zip — 30 MB
compressed, 815 entries, 633 tables x 633 COPY data blocks, 575
filestore files, 94 modules in manifest.
Made-with: Cursor
