Hi, > > > > At the moment I only turn DBI tracing on in the connect_cached() call and > > turn it off again before the call returns. I did that because I was worried > > about the amount of output that would be produced if I left tracing on. > > But perhaps that is what I will have to do. > > Perhaps you don't need to output it. Call $stacktrace = Carp::longmess > and only output it if the stacktrace is different to the last one, or > some similar logic.
I decided I had enough disk space to just leave debugging on for a bit and a quick look at the debugging I'm getting now shows something that seems odd and potentially may be interesting. I'm now getting debugging for more than just calls to connect_cached(). It seems that as well as calling DBI->connect_cached() for it's connections to one database, our application is also calling DBI->connect() for connections to a different database (I think this may be buried in the CGI::Application package that the application uses, which is why I hadn't noticed it before). Both databases are on the same mysql server. I haven't looked at it in detail yet (and as it's nearly home time on Friday evening here won't have time to look at it properly until later!). But the quick look I have had seems to reveal that the driver handle changes after each call (or maybe only some calls?) to DBI->connect(). Perhaps a more in-depth examination will show I've misinterpreted the debugging and this isn't what is happening. If so, I'll post a followup message later. But I thought it worth posting this message now to ask if this sounds like something that could happen. And if it could, is it a feature or a bug? Thanks for all the help/pointers so far... Duncan
