bob, regarding the memory leak question. if you are seeing 1.6gb of before and after..then that may not be the problem. if it's a memory leak the IIS process should start eating into that 1.6gb available. My question is how do you know that is what is availble before the crash?
something just hit me that may indicate you are on the right track with the recordset...i had a very similar issue with an application. it would crash once a week, but then started happening more frequently sometimes twice a way as the number of users increased. It had to do with ODBC calls. the applications was using a class that wrapped the ODBC API. the only way i got to the bottom of it was writing a utility that monitored the application's memory at certain interval..and it showed that with more activity (increase in volume of processing) the memory was being drained, but not released.
In an event...the issue was this
1) when using parameters, the class was not calling SQLFreeStmt with SQL_RESET_PARAMS
2) when excution of queries were done, i was calling SQLFreeStmt with SQL_CLOSE but not a SQL_DROP.
what does this all means?
1) the easy one is..it means parameters are resources and need to be freed. so if you are using the Parameters collection, be SURE TO iterate thru the collection and set the objects = to nothing and then the collection itself. Infact, you should do this with all collection. simply setting a collection = nothing, does not free the resources of the objects in that collection.
2) the second i am not exactly sure how to translate to ADO..but this is what i think. using the ODBC API, doing a SQL_CLOSE close the statement handle (analygous to a recordset), BUT it does not release the resources of that handle. so you could reuse that handle if you wanted. to free the resources, you have to do a SQL_DROP. I think with ADO, this may translate to...make sure the recordsets are closed AND make sure you set them = nothing.
it still bugs me that all this used to work fine on the older machine...it's not like they day ou switched to the new server, that was when the volume on the site increased, was it?
