The mid-tier's not my strongest area, so thought I'd throw a design issue out for feedback:
Mid-tier configuration assumption (please confirm or verify): a mid-tier SAS web app (e.g, SASServern_1) uses a shared/fixed JDBC connection to access RDBMS data, therefore all users running the SAS web app will share the common JDBC connection ID to the data.
Issue: Internal regulatory powers-that-be of the data insist that different business groups cannot share the one common JDBC ID.
Solution/Question #1: Can the settings for the JVM be configured to conditionally set the JDBC ID to be group-dependant? (Not sure how..)
Solution/Question #2: The web app could be vertically clustered, in which case, could each instance be configured to use a different JDBC ID, and also configured to ensure each group only uses their specific instance?
Hi Andrew, Nice question 🙂 I am not an expert either, but I will try my best with my current knowledge.
Assumption/logic 1: The JDBC is a setting on a web application level, not on a web application server level.
Assumption/logic 3: And as far as I know, only 1 web app of each type be registered on a single metadata server (horizontal clusters counts as a single web app).
With those 2 assumptions, I would say that both of the possible solutions are not an option.
In other hand, besides the expensive option to have a solution or deployment level per group, I can think on an alternative, but it would require an extra development on the current architecture:
The current configuration is set as a global variable/setting in an application level. But I think that variable can be overwritten on a user session level or authentication domain, such as the web authentication.
This option is used on web development several times (I am aware if your requirement). But: 1- I think that flavor of change would not be supported by SAS Technical Support. Most likely. You can check with them and they might have even an additional option.
2- you can ask a web app developing team at your company, just for fun and giggles, what they would do in your scenario.
Thanks for the clarification re web application level, not web application server level - quite correct.
I'd still like to quiz the admins about vertical clustering (which is possible - horizontal and vertical clustering is one of the topics I cover when teaching SAS Administration), but although I've set up and used horizontal clustering in the past, I haven't set up vertical clustering - plus this requirement would mean they're not identically configured (different JDBC IDs).
I'm actually working with an installation team from SAS here - this was one solutions we considered, so we thought we'd throw it out here for commentary.