Integration Solution Overview: Synchronizing Master Data from Kingdee Cloud Cosmic to Jiandao MES Forms
Scenario and Value
In a real engagement, the IT team of a manufacturing company kept hitting the same wall: when master data was changed in Kingdee Cloud Cosmic (materials, suppliers, customers), the Jiandao MES forms were only updated the next morning after manual sync by ops. As a result, shop-floor picking frequently stalled because material codes did not match the work orders. The bottleneck was not ERP data entry speed, but the lack of an automated loop between "master data change → MES base tables."
Sitting down with the business side, we narrowed the scope to three documents: materials (BD_MATERIAL), suppliers (BD_Supplier), and customers (BD_Customer). The goal was simple—after ERP changes master data, MES should see it within ten minutes, without human pushing. Against this goal, we delivered three SYNC strategies, all unidirectional writes from Kingdee Cloud Cosmic to Jiandao, with no reverse flow in this phase.
Integration Architecture and Data Flow
The overall architecture follows a typical two-layer pattern: "ERP master data + low-code business forms," with a data integration platform in the middle handling extraction, transformation, scheduling, and exception handling.
- Source side (Kingdee Cloud Cosmic ERP): three categories of master data—materials, suppliers, and customers—as the authoritative source.
- Integration layer (Qeasy / 轻易云 Data Integration Platform): pulls source data, performs field-level transformation, executes code deduplication, schedules by strategy, and handles retries and dead-lettering.
- Target side (Jiandao low-code platform): hosts the MES material table, MES supplier table, and MES after-sales customer table, consumed by the shop floor and after-sales workflows.
The data flow runs in two stages. Stage one filters incremental records from Kingdee Cloud Cosmic using FModifyDate / FAuditDate / FApproveDate, capturing only audited and non-disabled master data. Stage two maps fields into Jiandao forms through Qeasy's transformation engine and performs upserts using business codes as the deduplication key. The three strategies have no dependencies on each other and can run in parallel.
Interface Inventory
| Strategy ID | Data Object | Sync Direction | Notes |
|---|---|---|---|
| Strategy 1 | MES material master (BD_MATERIAL) | Kingdee Cloud Cosmic → Jiandao | Suggested every 10 min, business key FNumber |
| Strategy 2 | MES supplier master (BD_Supplier) | Kingdee Cloud Cosmic → Jiandao | Suggested every 10 min, business key FNumber |
| Strategy 3 | MES after-sales customer master (BD_Customer) | Kingdee Cloud Cosmic → Jiandao | Suggested every 30 min, business key FNumber |
All three strategies are SYNC type with no execution-order dependency; staggered start is recommended to reduce pressure on the target system.
Implementation Highlights
Staged scheduling. Stage one (master data sync) contains the three strategies above. With no inter-strategy dependency, they can run in parallel within the same scheduling window. Production usage recommends */10 8-22 * * * and */30 8-22 * * * running staggered, with an additional full-volume verification job scheduled during the low-traffic window of 2-4 AM.
Incremental fields and full-volume fallback. Incremental sync uses FModifyDate, FAuditDate, FApproveDate as filter conditions. First-time deployment or data-fix scenarios use full sync; for master data, a weekly full-volume reconciliation is recommended to catch any records missed by incremental logic.
Centralized code mapping management. Material, supplier, and customer codes all use Kingdee FNumber as the business primary key, with idCheck deduplication on the target side. This "centralized code mapping with single-point deduplication on the target" pattern is usually maintained in Qeasy as a unified code-mapping table—when new document types are added, only mapping rows need to be extended, without modifying the strategy bodies.
Exception retry. Network timeouts and target-system 5xx errors use exponential backoff with at most 3 retries (30s/60s/120s). Token expiration triggers a refresh and retry. Missing code mappings or business validation failures (such as duplicate codes) are written directly to the dead-letter queue without blocking the main flow.
Privacy handling. If any synchronized fields contain personal sensitive information (contact phone numbers, ID numbers, etc.), masking should be applied at the mapping layer, or only business-necessary fields should be synced. Qeasy's field-level configuration supports disabling sensitive columns per strategy.
Best Practices and Lessons Learned
-
Do not scatter code mappings across strategies. At one customer site we saw the most common issue: material code mapping lived in strategy 1, supplier in strategy 2, customer in strategy 3. When ERP eventually switched encoding rules, changes had to be chased across three pipelines. The safe practice is to build a centralized code-mapping table in Qeasy, where all strategies reference the same mapping—encoding rule changes then only touch one place.
-
Master data does not need to be "real-time." The first version of the solution tried to push sync frequency down to one minute, which triggered Kingdee Cloud Cosmic API rate limits and caused large-scale strategy failures. Stepping back to 10/30 minute intervals stabilized the system, with no perceptible difference on the business side. This is a classic case of "engineering over-engineering becoming the failure source itself."
-
Separate master data sync from transactional document sync. This phase only delivers master data; actual MES transactional documents (production work orders, reporting records) are handled separately later through header-line phased write strategies. Do not pack master data sync and transactional sync into the same pipeline—when a master data change peak coincides with transactional writes, the target system will stall.
-
Someone must watch the dead-letter queue. Set the alert threshold to 100 messages and attach a daily inspection task to Qeasy's ops dashboard instead of waiting for business feedback. Dirty master data eventually shows up as shop-floor work-order errors, and the troubleshooting path is long.
When to Use Qeasy
When an enterprise uses Kingdee Cloud Cosmic (or similar ERP) alongside low-code platforms such as Jiandao, Feishu Bitable, or Mingdao to host MES, CRM, and after-sales business forms, Qeasy Data Integration Platform (轻易云) can handle master data distribution, cross-system code mapping, scheduling orchestration, and exception fallback—eliminating the script maintenance cost of building ETL in-house. Solutions at the three-strategy scale are typically live within one to two days.