Skip to main content

Trading Systems

Designing a Safe Rollout for a New Broker Connector

Stage a new broker connector through capability checks, account-specific verification, explicit enablement, monitoring, rollback, and reconciliation.

By Sunbot Labs

Updated

5 min read

Illustration of a broker connector capability matrix with permissions and verification checks

Adding a connector to an existing trading system creates several distinct milestones: code deployed, account authorized, capabilities verified, operating checks accepted, and submission enabled. Making those states explicit prevents a planned or partially tested connection from being mistaken for one that is ready to use.

Separate deployment from operational enablement

A connector can be present in a deployment without being ready for an account to use. Track whether its code is deployed, credentials are authorized, permissions are confirmed, expected capabilities are verified, operating checks pass, and a named owner has enabled it.

Keeping those states separate prevents a successful health check or visible menu item from being mistaken for permission to submit requests. The operating interface should show the effective state for the current environment and account.

Build a capability matrix for the actual account

Broker support is rarely one yes-or-no flag. Record the capabilities verified for each environment and account, including data access, order types, time-in-force values, cancellation, status updates, and any market or permission restrictions.

  • Market, account, order, and position data that can be read
  • Order types and modifiers accepted in the intended market
  • Status delivery through streams, polling, or both
  • Account-specific restrictions and permission evidence
Illustration of a broker connector capability matrix with permissions and verification checks

Stage enablement by environment and account

Move from contract fixtures to a broker-provided test environment where available, then to client-controlled verification with a restricted account or operating mode. Each stage should have acceptance evidence and a reversible enablement decision.

Gate submission separately from read-only connectivity. This lets monitoring and reconciliation be verified before the system is permitted to create a new broker request.

Illustration of staged broker connector enablement with monitoring, rollback, and reconciliation

Make availability visible in the product

A product can show an upcoming connector without presenting it as selectable. Use explicit availability states and explain whether the limitation applies to every account, one environment, or a specific capability.

  • Planned: design and access work remain
  • Testing: verification is limited to named environments or accounts
  • Available: approved capabilities can be enabled deliberately
  • Degraded or disabled: submissions are blocked while visibility remains

Plan rollback and reconciliation before release

A rollback must define what happens to new submissions, open orders, pending cancellations, stream processing, scheduled account checks, and operator access. Disabling the connector should block unsafe actions without removing the evidence needed to understand venue state.

Reconcile local records with broker-reported orders and positions before resuming. Record the reason, decision owner, affected account, configuration version, and acceptance evidence for each enable or disable event.

Use explicit rollout states and acceptance evidence

Store rollout state per connector, environment, and account instead of relying on a global enabled flag. A state change should reference the capability matrix, configuration version, verifier, time, known limitations, and recovery owner. Credentials remain in the approved secret store and do not belong in the rollout record itself.

Treat a failed acceptance scenario as evidence that one capability remains unavailable, not a reason to edit the expected result or enable the whole connector. A restricted rollout can allow verified read operations while order submission remains blocked.

Rollout states keep product availability aligned with verified behaviour.
Rollout statePermitted behaviourEvidence to advance
PlannedNo connection or submissionApproved design and access path
ConnectedHealth and authorized reads onlyIdentity and permissions verified
VerifyingNamed test scenariosCapability-specific acceptance results
AvailableExplicitly enabled capabilitiesOperating owner and recovery checks
DisabledBlock new submissions; retain evidenceReconciliation and reviewed recovery

Practical questions

Questions that often come up

Can a deployed connector remain unavailable to users?

Yes. Deployment only places code in an environment. Account authorization, capability verification, operating checks, and explicit enablement should remain separate gates.

How should a planned connector appear in the interface?

It can be visible as planned, but it should not look selectable or available. The same area should explain the current state and any environment or account boundary.

What should happen when a connector is disabled?

Block new submissions, preserve monitoring and audit evidence where safe, identify uncertain orders, reconcile broker state, and require a reviewed recovery decision before re-enablement.

Work with Sun Cluster

Planning a similar system for your organization?

Sun Cluster develops broker connectors with secure access, order-state handling, logging, and reconciliation.