I want to know which tables are being read by a query.
for each Customer where CustomerID = 12345.
Eventually this customer will be found in the following example, but progress must 'read' many tables before getting to customer 12345.
How do I know exactly which tables are read (By CustomerID), prior to getting to customer 12345?
*NOTE: I do not have access to modify the code being run for this selection. Ideally I would run a separate set of code that is executed at the same time as the customer query above to track the reads.
EDIT: More clearly - Can you track reads from a given program (.p) OR ProcessID and output either a RECID or the PrimaryKey to a file?
I understand the information is being read off the Disk and probably stored in a database buffer. So how would I get at the information in the database buffer?
You seem to be mixing up a few different things.
In a situation like your example where you FIND a specific record in one, and only one table then there is just a single record read. Progress will find that record by first scanning a relevant index. That might be 2 or 3 "logical reads" of the b-tree to get to the proper node. The record block and index blocks may, or may not be read from disk - that depends on what has happened previously.
There are "Virtual System Tables" available that can tell you how many READ operations take place against a particular table or index. But they do not trace the specific ROWID or other identifying data. _TableStat and _IndexStat are aggregates for all users on the system, _UserTableStat and _UserIndexStat are specific to a particular user's activity. You do need to set the -tablerangesize and -indexrangesize parameters adequately to take advantage of these.
If you have enabled the table and index statistics then you can use a tool like ProTop - http://protop.wss.com to get insight into this activity. Or you can write your own code.
OpenEdge Auditing does not track reads. That would be prohibitively expensive.
It's probably not really a good idea but, in theory, you could write FIND triggers for the tables you are interested in. That doesn't require access to the application source but you would need a development license. It will probably kill performance to do this though - so unless this is a non-production test environment that you just want to fiddle with I wouldn't really do that.
You mention wanting to know how you got to that point. That sounds more like you might need to have a "4gl trace". One easy way to get the stack trace of a running process is to execute:
$DLC/bin/proGetStack PID (UNIX)
or
%DLC%\bin\proGetStack PID (Windows)
This command will generate a "protrace.pid" file containing a 4gl stack trace and other interesting information.
There are also more complicated ways to get that info like using PROMON and the "client statement cache" or setting various log entry types at session startup. But proGetStack is pretty convenient and requires no code or scripting changes.