• Bugs and Features of MXCS (SQL/MX Connectivity Services).

1. Turning trace or statistics on and off takes effect only when the datasource process first connects. It does not apply to any connected server datasource process and there is no script like the ./trace script for httpd in web server. If the client is using pooled connections the data source processes will always be connected. If the client executes a statement that was prepared prior to your oldest ems log and you need to see what that query is, you must recycle the connected server datasource processes to engage the trace and see the prepared statement. You must do a stop immediate in NSM/web but that will not stop the connected processes. See the next problem.

2. A stop immediate in NSM/web will appear to work but none of the connected server datasource processes will be stopped. They will continue to service requests but that will not be visible in NSM/web. They must be stopped manually.  Because these processes will have various owner-ids, the stop will usually have to be done as super.super. Using super.super is often subjected to a great deal of approval.

3. The database access protocol (the vender protocol) is not available to anyone (even GCSC), so a tcp/ip trace is no help in solving problems.  The MXCS trace events, once you get them going, are not much better. See next problem.

4. While the format of the trace events is documented, their contents and meaning are not. Often that content has content and contains codes that cannot be interpreted by anyone, including GCSC. Given a great deal of experimentation some of the content can be translated and put into your personal book of spells.

5. The support from development for this product seems to be non-existent. No one seems to own the product. There is no data on how many GSCS cases have been closed or cancelled because it was clear that no answer was forthcoming or the customer simply gave up. There have been many.

6. The server data source processes have been known to put one (and sometimes two) cpus at 100% busy with high queues. Given problems 1 thru 5 above, it is difficult to determin how it happens. Some users who have banned the product which means no JDBC type-4 allowed.