Intermittent "Out of present range" from Classic ASP after migration from SQL Server 2000 to 2008 & IIS6-IIS7

Question

Background: I have just completed a move of approximately 50 classic ASP sites from an IIS6/Sever 2003 and SQL Server 2000 environment to a new virtual environment of 2 machines behind an nginx load balancer. Each MS machine is running IIS7.5 and SQL Server 2008 R2. They current each have 6Gb & 2 VCPUs. The databases are set up in a mirroring configuration (currently without a witness).

During testing all sites appeared to function correctly.

Once live traffic started to hit the sites it became apparent quite quickly that the initial resource allocation (2Gb & 1 VCPU was way too low and was quickly increased). The main problem has come from an intermittent ASP error occuring on approximately 10 (and probably including the busiest) sites on the servers. They will produce a 500 response from an ASP error of

Provider error '8002000a' Out of present range.

All research has pointed to causes such as numbers too large to fit into an integer variable and some people have mentioned some correlation with the newer implementation of RAND and NEWINT() in SQL Server 2008 compared to 2000. The stored procedures that appear to cause the error are relatively simple, with some as simple as accepting a single VARCHAR parameter (well within the limits) and doing a single column select on a table. Most do not even involve INTs at all and if they do, the values are well within range.

The error can appear on one machine for a given amount of time while during this same time the other server will not necessarily have the error, it sometimes will though. After a while the error will stop occurring, this doesn't seem to correlate with excessively overloaded system resources either.

ASP to database is done via a DSN using SQL Server Client 10 drivers. The code is using the ADODB connection and command objects. This code has been working happily for 6+ years on the previous servers. The databases are set to compatibility mode 80 (SQL Server 2000).

Can anyone shed any light on where I should be looking to try and solve this please? If there is any other information I can share, specific code snippets etc please just let me know.

Update: I thought the UPDATEUSAGE answer below had got it but unfortunately it reared up again a little later. After some thinking I've had the following thoughts... There are two instances of IIS, independent of each other, they both talk to a single database whether it be local at the time or not, they both execute identical sync'd code with code that has been working with the same syntax and valid variables for a long time. As the ASP execution through IIS is the only layer in this equation that is not a single point as it were this is where I've headed. When the problem reoccurred, I restarted IIS on the machine at that point that was showing the error (the situation is often that it is only occurring on one of the two servers). The restart of IIS appeared to cure the problem. It then happened on the other server with a different site, again restarting IIS appeared to sort the issue.

Further reading has now lead me to the "Managed pipeline" modes of the app pools. They are currently set to "Integrated". I've done some reading and I'm wondering if they should be set to classic to emulate IIS6. Does anyone have any more thoughts on this?

Many thanks Eric

Answer 1

Did you:

(1) Update usage counters: In earlier versions of SQL Server, the values for the table and index row counts and page counts can become incorrect. To correct any invalid row or page counts, run DBCC UPDATEUSAGE on all databases following the upgrade.

(2) Rebuild all Indexes

• Upgrading from SQL Server 2000 to 2008

Answer 2

I had the same problem and tracked it down to a field definition in my database i had defined as a long integer. the value i had in there was some like 53435534126262 , immediately changed it to a text field and the problem disappeared

try that??

Answer 3

I thought it might be useful to post my findings and solution to this problem as I found no where on the web that mentions the same situation I had.

I went through a number of steps that each seemed to reduce the frequency of the errors but not eliminate them. Firstly I changed the database authentication method to SQL instead of Windows based. At first I changed all the sites to use the same login but later on I changed them to all use a unique login.

I updated the SQL Server with service pack 2 and cumulative update pack 3.

As mentioned, the above steps reduced the frequency of the errors but didn't stop them. I started looking through the class that all the sites use to manage their database connections and use of stored procedures. I came across the line adocommand.parameters.refresh I read up on what this actually does and when called it makes a call to the database to retrieve the parameters of a given stored procedure so that they can be accessed as an object in ASP rather than the parameters having to be given in a particular order and manually assigning the types to them. On the Microsoft page that details this method it has a little footnote that says

Parameters.refresh will fail in some situations or return information that is not entirely correct. Parameters.refresh is particularly vulnerable when used on ASP pages.

This was all it gave and I couldn't find any other details about this. I increased the logging on my sites to, on error, output what parameters.refresh had returned. I caught it in one instance returning the two variables from the stored procedure, with the correct names, but not with the correct variable types. They should have been a VARCHAR and an INT but they came back as both being CURRENCY. Obviously this then errors when you try and assign a string to a CURRENCY. I only managed to catch this one instance of an error before I fixed the problem.

The only way I found that seemed to fix the problem was to change from using an ODBC based driver, both DSN or DSNless, and use the SQL Native Client OLE DB driver with the "PROVIDER" keyword. This had the added benefit of appearing to enable connection pooling when it previously didn't appear to have been working.

One side effect of changing to the driver is that the stored procedures and ASP became susceptible to intermediate results being returned from the stored procedure if there were multiple statements within it and it didn't have SET NOCOUNT ON explicitly set at the top. Rather than trying to update 1000+ stored procedures, I found that the NOCOUNT flag can be set at the database instance level for all databases which solved this problem.

I hope this helps someone, as it was an incredibly frustrating 3 weeks that I spent tracking down this problem. Feel free to ask any further questions and I'll help if I can.

Thanks Eric