Install, activate and run your first pipeline
Bridge runs on a laptop for evaluation and on a single VM or multi-node cluster for production. This quickstart covers the node requirements, the activation model and the first pipeline.
Docs · Quickstart · Updated 2026-11-03 · Applies to current Bridge releases
Before you start
| Requirement | Evaluation | Production (Standard) | Production (Business) |
|---|---|---|---|
| Hardware | Laptop or small VM | 1 node, up to 4 vCPU | Up to 2 nodes, 16 vCPU total |
| Environments | 1 | 1 production | 1 production + 1 non-production |
| Source access | Read-only account | Read-only account plus log-read privileges | Same, per captured database |
| Network | Outbound to targets only | Outbound to targets only; no inbound required | Same, or file exchange for isolated networks |
| Licence | 14-day full-feature trial | Annual or monthly subscription key | Annual or monthly subscription key |
Steps
- Install the runtime. Deploy the Bridge engine on a supported Linux host, or start the container image. The same build runs on a laptop, a single VM or a cluster — the mode is a configuration choice, not a different product.
- Activate the licence. Apply your licence key locally. Activation is a key, not a live call to a vendor service, so the platform can run without outbound control-plane traffic.
- Register a source. Create a read-only database account and, for log-based capture, grant the log-read privileges listed in the CDC reference. Test the connection before configuring a pipeline.
- Register a target. Authenticate to the warehouse, lake or file system. Check identifier case rules — warehouses disagree about unquoted identifiers, and this is a common first-run surprise.
- Build the pipeline. Choose a source-to-target mapping on the canvas, or write it declaratively as configuration. Preview the mapping and the generated DDL before running.
- Run and verify. Start with one table. Confirm row counts, then a checksum on a sampled key range, then the change stream. Only then add the remaining tables.
Expected time: most teams reach a working single-table pipeline within minutes of installing, and complete a safe dual-run migration from a legacy tool within days rather than quarters.
What "activation" does not do
Activation does not open an inbound port, does not upload schema metadata, and does not require a vendor-hosted control plane. The platform's configuration, metadata and job logs stay in the deployment. If a firewall rule blocks all outbound traffic except your targets, the product still works.
Troubleshooting the first run
| Symptom | Likely cause | Fix |
|---|---|---|
| Connection test passes, capture reads nothing | Log retention or supplemental logging not configured | Enable logging and re-check the recorded position |
| Target rejects identifiers | Case-folding rules differ by warehouse | Set identifier quoting policy per target |
| Duplicate rows after restart | Offset not committed before failure | Confirm checkpointing is enabled on the pipeline |
| Snapshot runs long | Snapshot taken through a busy replica | Use a read replica or raise snapshot parallelism |
Install questions
Do I need Kubernetes?
No. The engine runs on a single host without Kubernetes. Cluster mode is optional and uses the same build.
Does activation require internet access?
No. Activation is a licence key applied locally; there is no mandatory call to a vendor service.
Can I run it in an isolated network?
Yes. Use file exchange across the security gateway; no inbound connection is required.
Is there a trial limit?
The 14-day trial is full-feature, including log-based capture and DDL sync.