Upgrades and maintenance
Braintrust upgrades deployments to newer, reviewed module versions. Available features depend on the version deployed in your account, so confirm it with your account representative. See the data plane changelog for changes in each release.- Coordination. Braintrust reviews and confirms the Terraform plan for each upgrade. Maintenance notice and scheduling follow your agreement with Braintrust. Share scheduling preferences and blackout windows with your account representative.
- Validation and recovery. After each upgrade, Braintrust checks the deployment’s health. If an upgrade causes problems, Braintrust supports service recovery and reconciles the deployment through the deployment process.
- Version. To see the data plane version, go to Settings > Data plane, or request the data plane’s
/versionendpoint. - Vulnerabilities. When a scan reports a vulnerability (CVE), Braintrust checks whether it can be exploited in your deployment. Based on that assessment, Braintrust adds safeguards, applies an emergency update, schedules an upgrade, or asks you to take action.
Service health
Braintrust monitors the health of your data plane using telemetry and cloud metrics and alarms, without reading the contents of your data-plane resources. To view throughput, resource usage, and status checks yourself, go to Settings > Infra dashboard. The dashboard is view-only. Service-level commitments, such as uptime targets and service credits, are defined in your agreement.Coordinate changes to your environment
Changes on your side can break deployments and upgrades without an obvious error. Coordinate with Braintrust before you:- Add or change organization policies, SCPs, or IAM Deny policies that apply to the account. An explicit deny overrides the permissions Braintrust’s identities need.
- Change the trust, permissions, or bindings of Braintrust’s identities.
- Change networking, DNS, or firewall rules that the data plane uses.
- Edit resources that Braintrust manages. The next deployment can overwrite your edits.
- Turn off logging.
Incident response
If you suspect that a Braintrust identity is misused or compromised, contain the affected access promptly. In parallel, preserve audit logs and session activity outside the BYOC account. Don’t delay revocation while an export or archive completes. To contain access:- Block new sessions by removing or restricting the trust relationship or bindings for the affected identity.
- End existing access separately. Short-lived credentials that were already issued can remain valid until they expire unless you revoke them. See Revoke access on AWS.
Offboarding
If several data planes share a bootstrap, wait until all of them are removed before revoking shared access or deleting the bootstrap stack.1
Export your data
Export the data you want to keep before Braintrust removes the data plane. See Export data.
2
Deprovision the data plane
Braintrust removes the data-plane infrastructure and keeps the data resources listed in Deprovisioning. Braintrust then confirms removal and gives you a verified list of the resources that remain.
3
Revoke access
Revoke Braintrust’s access, then confirm in your audit logs that no new Braintrust sessions start.
4
Clean up
Decide whether to keep or delete the retained data and the versioned objects in the state and operation log buckets, then delete the bootstrap’s CloudFormation stack. Braintrust doesn’t delete any of these for you.
Deprovisioning
When you leave BYOC or retire a data plane, Braintrust removes its infrastructure usingretain-data mode, the only mode available for customer deployments:
- Removed: The infrastructure that serves the data plane.
- Retained: The Brainstore bucket and its controls, the KMS key and alias that the module created, and an encrypted final RDS snapshot.
- Unchanged: The Terraform state bucket and the operation log bucket.
Next steps
- AWS access and security: Review permissions, audit signals, and access revocation.
- Contact Braintrust: Coordinate changes with your account team.