Before
- Back up application data and files. Allow time for contact-attribute and storage-URL data migrations.
- The attribute migration infers
text,numberordatefor each key. It counts only number/date values that match the conversion pattern, are nonempty and still lack a typed value. Below 1,000,000 eligible values, it fills those columns; at or above that threshold, it skips the value backfill. Use the migration logs to decide whether the backfill below is needed; the total attribute count is not the threshold. - Storage URLs in surveys, styling, branding and response data are converted to relative paths. Keep the existing S3 endpoint, bucket and proxy route; this is not a transfer to another storage provider.
During
Use the 4.x/5.x procedure for those targets, or the v6 procedure for a v6 target. Check the application/migration logs forSkipping value backfill. A successful schema migration does not
mean a skipped backfill has completed.
Complete a skipped backfill
Run the script from the installed release image, after its migrations succeed. Do not run multiple copies concurrently. Complete it before you upgrade to the next minor version.- Docker
- Kubernetes
Run from the directory containing your existing
docker-compose.yml. The app must be running on the
upgraded image; keep public traffic closed until the backfill finishes.0 and Backfill Complete!. The script fills only missing typed values and can resume after
interruption; batches commit independently. It avoids the migration transaction timeout but still uses
database resources. Stop and investigate any error before advancing.
After
- Check number/date segment filters against known contacts, including invalid or empty values.
- Open old survey images and previously uploaded response files through the public route.
- Compare surveys and responses with the baseline and complete the acceptance checks.