Hi Robert,
In the Jobs list, I don't see a way to define a job to reference an individual test item either, or to pass a parameter to one.
(where 'test item' is defined as an those in the Test Items tab of the Project).
In the Test lookup, the only thing I can find are procedures in the project.
To make that work, I would have to define a different procedure for each group, that internally calls the list of tests (csv files) in that group. That's certainly possible if it's the only way to do it, but would involve some structural rework, especially since I have some custom code in GeneralEvents_OnStopTest that relies on Project.TestItems references.
If there was a way to launch Test Items directly from scripts I could define a procedure for each group, but that doesn't seem to be supported.
Still thinking...
Allen
How about a single slave project that includes a routine or something that first of all queries the machine name and then, based upon the machine name, executes a particular set of tests? Yes, this would take some restructuring, but I think it would, over all, be less complicated. What you would do, then, is have the main routine of the project use a DDT object that would execute an SQL query against your CSV file, that SQL query altered based upon the machine name. So, Machine1 would get all tests where GroupName = "Machine1", Machine2 would get all tests where GroupName = "Machine2" and so on.
Execute that slave project on your slave machines.. they are now context-aware..., they know what machine they are running on and will then execute any tests you need to. What this would mean is that the slave project, instead of having a whole list of test items, would ahve a single test item that would call your test, parameterized based upon machine name.
Given time, I could write up the code for you but, perhaps, this will trigger some thoughts so you can fiddle with it yourself?
Hi Robert,
I understand your proposed solution, and it is something that I could implement myself.
Thinking through some possible cons:
Test groupings would be somewhat harder to visualize and maintain, most likely maintained in Excel instead of natively.
Built in timeout capability for individual tests would be lost. I would need to implement my own if needed (rare, but it does happen).
So far nothing insurmountable, just a little work.
Error logging: That's a problem. How could test failures be broken out and reported for individual tests out of a group?
Another possibility would be to implement keyword tests to call my CSV launcher script or a group driver script directly, just like a test item. I haven't used them before, but they are selectable individually for a distributed project. Seems a bit of a kludge, but might maintain some of the capability above.
FYI: I did query SmartBear support directly and got the official answer to my initial query:
"You cannot select test items or an item group to run on a slave machine. As a test in task properties, you can choose a keyword test or a function in a script unit:
https://support.smartbear.com/viewarticle/68430/"
Anyway, my original question is answered (even if I don't like the answer) and I have some options now, either pure code + CSV, keyword tests, or some hybrid approach. I'll play around for a while and see which one works out.
Thanks!
Hello all,
I have my distributed test working now, and wanted to follow up by presenting my solution. I didn't follow Robert's advice exactly, but it's something like that.
I have a single slave project, my original project where each test item calls my ScriptLauncher procedure and passes the CSV file representing the test as a parameter.
What I have done is add a second parameter to pass a machine name. The ScriptLauncher procedure checks the machine name, and only loads the CSV file and runs the test if the parameter matches the current hostname.
Then from the master NetworkSuite, I just call each host with the same project.
This allows me to distribute my test item groups across multiple host machines just by adjusting the test item parameters.
I still get full logs of each individual test item recorded in the network suite.
The main drawback so far is that even the skipped test items get logged as a Pass, but it's a lot better than the alternatives I tried, and I really only care about the failures anyway.
Thanks.
Allen
