Dynamics 365 Finance and Operations: Debugging Batch Processes Gone Mad
One of the more interesting aspects of our job as developers in Dynamics AX/Dynamics 365 for Finance and Operations can be getting enough information to reproduce an issue. This can be especially true if there are multiple processes running via batch, that may affect the data in ‘unexpected’ manners.
Say for instance a user reports an issue with some data. When manually iterating through the process, it works wonderfully, and all results are expected. You go through and attempt to blatantly break the system, and the checks seem to function as designed. You sit and scratch your head, and mumbling under your breath. You go back to the user and ask them how they got the data into this state. They THEN come back to you and state that they didn’t, the system did. You do some examination, and lo and behold, there is a batch process that runs in the background that automates the process you just manually attempted to break.
So, being the good developer you are, you start debugging that process, and it seems to function as expected as well. Back to digging. You then find ANOTHER batch process that runs that may also affect the data. But, how do you tie everything together and prove that this is the cause?
When I see this, one of the things I like to do is create custom logs for the processes. You can write your hooks into the code and provide a ton of useful information. I like to do something similar to the warehouse work creation log history functionality where I provide a form to the end user that has the process information (using descriptive labels of course – this is where the strFmt() function can really be your friend). I like to also put the call stack into the information, and allow the user to add that to the form via a personalization if desired (and possibly add it to a ‘General’ tab on my history form). The call stack can be added into the data via code similar to this:
I also add a time stamp to the data, as we will be using this in the next step.
I also provide methods of disabling the logging, and options for clearing the data, as this can become data intensive rather quickly.
Once this is done for all the suspected processes, the fun begins. Once the issue appears again, you can write X++ jobs or TSQL statements to analyze the data from the different log tables (compare time stamps is usually a good place to start) to see if it can be determined what is going on (specific code running in a certain order, changing data at unexpected times, data rollbacks due to failure, etc.). Dumping the data to Excel can be an option as well if you are an Excel wizard (I am not, so I tend to shy away from that).
Trace parser might be able to help you here as well, but I have actually found that some clients like having the log functionality for other uses as well. It may be enabled periodically and used to see if some processes are taking a while, are being called rather frequently, or possibly being scheduled to run too frequently for the amount of work that is being processed.
So, in summary, I guess I am trying to say that sometimes thinking a bit out of the box may be beneficial in ways we as developers may not initially intend. Use established processes, but if there is a way to provide more information with little or no penalty to performance, go for it.
If you have any questions or if debugging batch processes have you going mad, feel free to contact us here at Stoneridge Software.