Long-running BizTalk migration requires a plan for work already in progress, not only a replacement workflow for new messages. Identify the active business transactions, what each instance is waiting for, how later messages are correlated, and which platform will own completion. Do not assume that deploying a new workflow transfers existing runtime state.

An order, claim, or partner exchange may span hours or days and involve several responses. The migration plan must account for those unfinished obligations. Otherwise, a technically successful deployment can leave the organization uncertain about which system should receive the next message or finish the business process.

Inventory business work as well as applications

For each long-running flow, describe the business transaction it represents and the conditions required for completion. Identify waiting points, expected responses, timeouts, exceptions, and any manual intervention.

Use supported administrative and diagnostic approaches appropriate to the environment. Do not edit BizTalk databases to manufacture a migration state or treat this article as an operational runbook.

Connect technical instance identifiers to the business identifiers needed by operations, within the organization's data-protection rules. The team needs to know not only that instances exist, but what obligations they represent and who can confirm the business outcome.

Understand persisted state and correlation

Microsoft describes orchestration dehydration as persisting an instance's state to a SQL Server database, with rehydration restoring it when processing resumes. That is runtime behavior inside BizTalk, not a portable promise about another platform.

Microsoft's correlation documentation explains how promoted-property values can associate later messages with a particular orchestration instance. Those associations need attention when traffic is split between old and new processing paths.

Document the actual keys and initialization behavior used by your application. A new workflow that understands the message format may still lack the state and routing context needed to continue an existing transaction correctly.

Decide which system owns new and existing work

Establish a clear boundary. One possible design is to send new transactions to the target while allowing existing BizTalk instances to finish. Another situation may require a different, explicitly engineered transition approach.

These are architecture options, not universal product features or instructions. Evaluate the dependencies, duration, routing, and business requirements before selecting a pattern. Some flows may be practical to drain; others may include long waits or unresolved cases that need separate treatment.

Write down how ownership is determined for every later message. A date-based cutover rule alone may be insufficient when a response arriving today belongs to a transaction started earlier.

Examine the response path

List the systems and partners that can send follow-up messages. Determine whether they use stable business identifiers, technical correlation values, endpoint-specific assumptions, or a combination.

Ask how the receiving path will identify the correct owner during coexistence. Avoid a design in which both platforms can act on the same response without a defined control. Also avoid assuming that a partner can change its behavior immediately to accommodate your migration.

A hypothetical purchase process might begin before the new platform is introduced and receive a later invoice afterward. The transition design must explain how that invoice reaches the process that owns the original transaction, not simply whether the new platform can parse it.

Define treatment of exceptional instances

Separate normal waiting work from suspended, expired, abandoned, or otherwise exceptional cases. Each category may require different business and technical decisions.

Do not terminate instances merely to reduce a dashboard count. Before any operational action, establish the business status, potential side effects, recovery options, authorization, and recordkeeping required by the environment.

Some cases may need a controlled business reconciliation rather than a technical continuation. Others may require further investigation. The important point is that the decision is owned and documented, not hidden inside a migration script or an assumption that old work no longer matters.

Plan the coexistence responsibilities

If BizTalk and the target operate together for a period, assign monitoring, incident ownership, release coordination, and support escalation for both. Coexistence is an operating arrangement, not simply an architecture diagram.

Review how shared dependencies will be maintained. A certificate, lookup service, schema change, or downstream update may affect both platforms. The team should know who assesses the impact before a change is made.

Avoid using “temporary” as a reason to omit documentation. Temporary arrangements often carry the most uncertainty because they combine old and new behavior. Give them an owner, review point, and explicit conditions for ending them.

Define completion evidence before retirement

Decide how the organization will know that the old flow no longer owns unfinished business. A low message count is not enough if exceptional cases or downstream obligations remain unresolved.

Combine supported platform evidence with business reconciliation. Confirm which transactions completed, which were intentionally handled through another approved process, and which still require attention.

Keep an evidence record that operations and business owners can understand. The retirement decision should not depend solely on the person who designed the migration remembering why a few remaining instances were considered harmless.

Keep rollback expectations realistic

Discuss what can and cannot be reversed after the target begins processing new work. Business side effects may persist even if a workflow deployment is rolled back.

Separate configuration rollback from transaction recovery. The first changes software or routing; the second addresses what happened to actual business work. Both need a plan appropriate to the selected design.

Do not promise a universal zero-downtime or one-click rollback path. Evaluate the specific flow, ownership boundary, data state, and downstream behavior. The review should make uncertainty visible before the transition is approved.

Common questions

Can a new Logic Apps workflow automatically continue a dehydrated BizTalk instance? Do not assume that. Runtime state and continuation require a specific supported design and validation.

Should we wait for every instance to finish naturally? That may be one option, but long waits and exceptional cases can require another reviewed approach.

Who should approve the plan? Include technical owners, operations, and the business owners responsible for the transactions, with security and other specialists as needed.

Review the unfinished work before choosing the sequence

BeyondBizTalk supports discovery, modernization planning, and operational continuity for BizTalk environments. A focused orchestration review can connect runtime behavior with the business work that must continue. Bring the instance patterns, correlation dependencies, and current support concerns to the conversation so the migration sequence reflects the real obligations of the system.

Microsoft documentation

← Back to All Insights