# How to Migrate from Google Cloud storage to Quake AI

Source: https://docs.quake.ai/docs/object/migration/migrate-from-gcs
Markdown: https://docs.quake.ai/docs/object/migration/migrate-from-gcs.md

---

# How to migrate from Google Cloud storage to Quake AI

This guide walks through migrating objects from a Google Cloud Storage (GCS) bucket to Quake AI object storage using rclone for direct cloud-to-cloud transfer.

## Service mapping

<MigrationTable provider="gcp" service="storage/object" />


## Prerequisites

- A Quake AI account with [S3 credentials](/docs/object/how-to/create-s3-credentials)
- A GCS service account with the **Storage Object Viewer** role on the source bucket (or **Storage Admin** if you plan to delete after migration)
- A service account JSON key file downloaded from the Google Cloud Console
- [rclone](https://rclone.org/install/) installed (v1.65+)

## Before you start

Read the [migration planning guide](/docs/object/migration/plan-your-migration) to review S3/Swift feature compatibility and build a validation checklist.

**GCS-specific considerations:**

- **Storage classes**: GCS objects may be stored in Nearline, Coldline, or Archive classes. These classes have minimum storage duration charges (30, 90, and 365 days respectively). If you delete objects from GCS before the minimum duration, you are charged an early deletion fee. Factor this into your migration timeline.
- **IAM permissions**: GCS uses IAM roles (Storage Admin, Object Viewer). These do not translate to Swift ACLs. Re-apply access controls on Quake AI after migration using [grant access control](/docs/object/how-to/grant-access-control).
- **Object Lifecycle rules**: GCS lifecycle rules (SetStorageClass transitions, Delete actions) are not supported on Quake AI. Implement retention with external tooling. See [retention without lifecycle policies](/docs/object/migration/plan-your-migration#retention-without-lifecycle-policies).



GCS Archive-class objects carry a 365-day minimum storage duration. If you delete Archive objects from GCS before they reach 365 days, including deleting the source bucket after cut over, Google charges an early deletion fee for the remaining days. Nearline objects carry a 30-day minimum and Coldline objects a 90-day minimum. Check the age of your objects before you delete the source.



## Configure rclone remotes

Add both GCS and Quake AI to `~/.config/rclone/rclone.conf`:

```ini
[gcs]
type = google cloud storage
service_account_file = /path/to/service-account-key.json
project_number = YOUR_GCP_PROJECT_NUMBER
object_acl = private
bucket_acl = private

[quakeai]
type = s3
provider = Ceph
access_key_id = YOUR_RUMBLE_ACCESS_KEY
secret_access_key = YOUR_RUMBLE_SECRET_KEY
endpoint = object.us-east-2.rumble.cloud
acl = private
```

Replace `object.us-east-2.rumble.cloud` with your Quake AI region endpoint.

Verify both remotes:

```bash
rclone lsd gcs:
rclone lsd quakeai:
```

## Create the destination bucket

```bash
rclone mkdir quakeai:my-bucket
```

## Run the migration

### Initial copy

```bash
rclone copy gcs:my-source-bucket quakeai:my-bucket \
  --transfers 16 \
  --checkers 32 \
  --s3-chunk-size 64M \
  --progress \
  --log-file migration.log \
  --log-level INFO
```

### Delta syncs

Run incremental syncs to catch changes after the initial copy:

```bash
rclone sync gcs:my-source-bucket quakeai:my-bucket \
  --transfers 16 \
  --checkers 32 \
  --progress
```

Schedule this daily until you are ready to cut over.

## Estimate egress cost

GCS charges per-GB for egress, with rates that vary by region and destination. Google runs a free egress program for customers who close their Google Cloud account and complete the move within 60 days. A migration that keeps your GCP account active pays standard egress rates, so the egress meter dominates the bill for large data sets. Egress and exit terms change, so confirm the current policy on the [Google Cloud Storage pricing page](https://cloud.google.com/storage/pricing) before you plan your timeline.

For large data sets, consider a phased migration to spread the cost across billing periods.



For large migrations (hundreds of TB or more), evaluate Google's [Storage Transfer Service](https://cloud.google.com/storage-transfer-service) for the initial bulk copy. It can sustain higher throughput than a single rclone client, though transfer pricing and egress still apply. Confirm current pricing before you commit to a tool.



## Verify the migration

```bash
rclone size gcs:my-source-bucket
rclone size quakeai:my-bucket
```

Check file integrity on a sample:

```bash
rclone check gcs:my-source-bucket quakeai:my-bucket \
  --one-way \
  --log-file check.log
```

Review `check.log` for any mismatches. See the [validation checklist](/docs/object/migration/plan-your-migration#validation-checklist) for the full procedure.

## Cut over

1. Run a final `rclone sync` to capture last-minute changes.
2. Update your application's storage endpoint. See [update application endpoint](/docs/object/migration/update-application-endpoint).
3. Keep the GCS source bucket for 30 days as a rollback target.
4. Account for minimum storage duration charges before deleting Nearline/Coldline/Archive objects.

## See also

- [Plan your migration](/docs/object/migration/plan-your-migration): compatibility matrix, cost estimation, validation checklist
- [Migration tools comparison](/docs/object/migration/tools-comparison): rclone configuration and alternatives
- [Update application endpoint](/docs/object/migration/update-application-endpoint): swap SDK config to Quake AI