RWT Stress Tests with JMeter

[Sebastian-Habenicht]

I am working on stress testing an RWT application with JMeter. The setup is done, tests can be recorded and replayed in the JMeter GUI. This works reliably for all initially created widgets. However, when 'transient' widgets are created later on (e.g. Menus, Dialogs, GridItems, ...), the widget id of these widgets (resp. their child widgets) is not always the same as the one recorded for these widgets, which causes the test to fail (with an UnsupportedOperationException in WidgetOperationHandler, probably because the actual widget with the passed widget id is of a different type than the one during the recording).

I do not know how the widget id is generated by the framework - I assumed that the widget ids would be the same for each session, as long as the order in which the widgets are created remains the same, which is the case when replaying the recordings, but that is obviously not correct for these 'transient' widgets.

Correlation (like what needs to be done for the cid in the first request) will not help here, as there is no identifier for these widgets (like $.head.cid for the cid) and creating correlation rules based on text or other properties is not really an option.

So basically, the question is: are stress tests limited to the initially created widgets or am I missing something?

[ifurnadjiev]

Hi,

if the UI session is replicated 1:1 the widget ids should stay the same. For "transient" widgets like Menus, it depends on the time when widget is created. The ids are created sequentially by incrementing a number and append it to "w" during widget creation.

The connection id (UI session identifier) is created on the server and sent to the the client as a response to the first UI request. In JMeter you have to extract the "cid" with every new session and provide it as a URL parameter.

Hope that helps.