Hey. One time I tried recording video while the job was running but that didn’t work since the robot session is blacked out when it connects and runs.
Firstly, you want to make sure you have fully tested your process in the robot environment while you are watching the job run. If you have not done this and are confident that your coding is good to go with no glitches as an attended process, then you can move to testing unattended.
If your attended tests were successful and coding is robust, you technically won’t need to troubleshoot in unattended. From my experience, after I got over these learning curves and understood how an unattended logon session works on the environment, unattended tests went very smoothly.
But, here are some tips for unattended troubleshooting.
-first need to identify where the issue is happening. If you have good logging and exception information being outputted, this will be pretty easy.
-then, you need to place a TakeScreenshot step before the error is happening… you can also publish a TakeScreenshot workflow into Orchestrator to run when jobs fail (however, if you use a good framework, failures will end up closing all applications, in which you would need to custom place a TakeScreenshot before the error)
-The screenshot will show you things like what resolution the logon session has, what size is the window you are interacting with and what elements can be seen.
-it is sometimes beneficial to start your job with a Launch Workflow Interactive to set resolution and color depth, or the newer Orchestrator can set those settings for each Robot.
-Typically, issues that you will find will be related to the following:
—Internet settings on the robot allowing for popup blocker and notifications that interrupt a step
—Browser crashes showing notification to restore which should be disabled
—Popups loading in weird sizes due to resolution and website functionality
—Windows Error reporting crashes… sometimes a challenge to get around unless you disable the service
—Elements not visible where a scroll or maximize is necessary… ideally you don’t want to code your job in a way where you need to maximize the window, but depends. Also, avoid image and text clicks as element clicks are more reliable.
Those are just a few things I can think of. The key is taking a screenshot right before the problem area.