Method Node

[jiangjunjie01]

If the method node is ns=1;i=1001, which is located in a custom node manager MyNodeManager, and its parent node is the Objects node.

When executing the method call, the MasterNodeManager iterates through all registered NodeManager instances and invokes each one's CallAsync method.
However, since CoreNodeManager is called first, its Call method will first look up the parent node via GetLocalNode(methodToCall.ObjectId). If it finds the node, it sets methodToCall.Processed = true, causing all subsequent node managers to skip the call. But the method node ns=1;i=1001 is not in CoreNodeManager; it is in the custom node manager MyNodeManager.
Since CoreNodeManager cannot find the method node during its Call execution, it will directly throw an exception.

Is my statement correct? Is this a bug?

[cookesan]

Yes, that code path can happen if the ObjectId in the Call request is the standard Objects node.

MasterNodeManager.CallAsync sends the same CallMethodRequest list through each node manager. AsyncCustomNodeManager.CallAsync then checks methodToCall.ObjectId, not methodToCall.MethodId, and marks the request Processed = true as soon as that object belongs to the manager. Since Objects is namespace 0 and belongs to the core node manager, the core manager can consume the call before your custom node manager sees it. If it cannot resolve the custom method under that source object, the result becomes BadMethodInvalid/similar rather than falling through.

The safer model is to put the callable method on an object owned by your custom node manager and have clients call with:

ObjectId = <your ns=1 object>
MethodId = ns=1;i=1001

I would avoid hanging the method directly under Objects for this case. Browsing can show an external reference there, but method-call dispatch is still driven by the object node that is passed in the Call request.