Importing a Large SQL File Into Google Cloud SQL

Cloud SQL doesn't import from your local disk — it imports from a Google Cloud Storage bucket. Once you know the flow it's reliable even for big dumps, but the service-account permission step trips almost everyone up the first time.

Step 1: Upload the dump to a bucket

gsutil cp dump.sql.gz gs://your-bucket/dump.sql.gz

Cloud SQL imports gzipped SQL directly, so keep it compressed — uploads are faster and the import handles .gz natively.

Step 2: Grant the Cloud SQL service account read access

This is the step people miss. Each Cloud SQL instance has its own service account that must be allowed to read the bucket object:

# Find the instance's service account
SA=$(gcloud sql instances describe my-instance \
  --format='value(serviceAccountEmailAddress)')

# Grant it object read on the bucket
gsutil iam ch serviceAccount:$SA:objectViewer gs://your-bucket

If the import fails immediately with a permissions error, this grant is missing or hasn't propagated. Re-run the gsutil iam ch command and wait a minute.

Step 3: Run the import

gcloud sql import sql my-instance \
  gs://your-bucket/dump.sql.gz \
  --database=your_database

Handling multi-GB dumps

• Cloud SQL runs one import at a time per instance — large imports block other operations, so plan a window
• An import can't be cancelled cleanly once running; a failed huge import means restoring or starting fresh
• Temporarily scale the instance up (more vCPU/RAM) for the import, then scale back down
• Split the dump so a failure costs you one chunk, not the whole multi-hour run

Cloud SQL doesn't grant SUPER privilege, so dumps containing DEFINER=`user`@`host` on views/triggers will fail. Remove them before uploading.