Multi-warehouse inventory: upgrades for operations handover

Review upgrades for operations handover in Multi-warehouse inventory. Check document scope, failure paths and source evidence through controlled tests.

What the documents look like

Multi-warehouse allocation happens at the order routing stage, where the system decides which warehouse ships an order. The working data comes from WMS, OMS, ERP inventory and platform inventory interfaces, and the people involved are supply chain planners, WMS administrators and e-commerce operations staff. Daily documents include allocation rules, inventory synchronization tables, transfer orders and stockout alerts. Compared with courier handover, the permission boundary in allocation is easier to overlook: being able to view an inventory synchronization table does not mean being able to change allocation rules, and submitting a transfer order does not mean the transfer has been approved. During deployment and upgrade, old allocation rule names, inventory synchronization table fields and transfer order statuses need to be checked item by item, and current behavior cannot be inferred from an old configuration name.

Constraints at the upgrades step

Before upgrading, check the current version, the target version and the corresponding upgrade notes. The V4.17.2 upgrade guide requires updating the fastgpt-app and fastgpt-pro image tags, and it includes two manually triggered upgrade scripts: historical chunk training task migration and knowledge base rebuild status migration. The second depends on completion of the first. Neither script blocks service startup, and both must be clicked only after confirming that all old-version App and Pro instances in the cluster have fully stopped and old task leases have expired. During operations handover, the authorization state of materials, operations and requests must be confirmed item by item: a transfer order pending review does not constitute approval, a submitted stockout alert does not constitute a replenishment instruction, and a value collected by retrieval or a form does not constitute authorized execution.

Hypothetical drill sample: on one day, supply chain planner A submits an allocation rule change request reading "propose changing the East China warehouse priority from 2 to 1, applicable to new order routing from 1 October 2026, requiring approval by operations supervisor B." On the same day A submits a handover request: "the inventory synchronization table has been updated daily, please confirm and then execute transfer order D-20261001." B has not replied. Draft sentence to be assessed: "A has been authorized to adjust East China warehouse priority, and transfer order D-20261001 can be executed." In this sample, A's submission of the request is a factual operation; the available material contains no approval or confirmation record from B, so the related conclusion remains pending. The draft sentence treats submission as authorization and a request as execution, and neither is supported.

Configuration and implementation checks

Configuration itemAssumed starting point for this drillVerification method
Image tagTarget v4.17.2Compare against the fastgpt-app and fastgpt-pro tags listed in the upgrade guide
AIPROXYAPIENDPOINTAssumed filled as http://aiproxy:3000Confirm no /v1 is appended and that it matches the AI Proxy service root address
AIPROXYAPITOKENAssumed filled with this deployment's administrator tokenConfirm it is non-empty and matches the server configuration, and is not used as an application invocation credential
Manual upgrade scriptsRun historical chunk training task migration firstConfirm old instances have stopped and old task leases have expired before execution
Post-upgrade verification sampleReuse the same allocation rule and synchronization table as before the upgradeRecheck import, retrieval and application invocation with the same sample

These values are assumed starting points for this drill, used to illustrate the comparison object; they do not represent official default values.

Three common errors

  1. Treating a submitted request as granted authorization. The input is the allocation rule change request submitted by A; the comparison object is B's approval status; the available material contains only the request and a note that no reply has been given, with no approval record, so the authorization claim is unsupported.
  2. Treating a handover request as executed. The input is the transfer order execution request submitted by A; the comparison object is B's confirmation status; the request does not show execution, and neither confirmation nor an actual execution record has been provided, so both remain pending.
  3. Inferring current behavior from an old configuration name. The input is the pre-upgrade allocation rule field name; the comparison object is the target version upgrade notes; the observable result is that fields and script requirements need to be checked item by item, so the old name cannot be carried over directly.

How to verify the setup

  1. Check the image tags. The input is the currently deployed fastgpt-app and fastgpt-pro image tags; the comparison object is v4.17.2 as listed in the upgrade guide; the condition is met only when the two match.
  2. Check the preconditions for the manual scripts. The input is the state of old-version App and Pro instances in the cluster and the state of old task leases; the comparison object is "fully stopped and leases expired" as stated in the upgrade guide; the condition for clicking execute is met only when both are satisfied.
  3. Check the post-upgrade verification sample. The input is the same allocation rule and inventory synchronization table as before the upgrade; the comparison object is the import, retrieval and application invocation results after the upgrade; the condition is met only when the sample can be reviewed and differences can be recorded. When a retrieval result is empty, check the downstream design against an explicit paper input, and do not claim that a fictional rule must return empty.