On 18 Aug 2005, at 18:18, Jonathan Vanasco wrote:
Instead of connecting to the DB and having DBI cache the handle
after-the-fork, would it make sense to subclass Apache::DBI, have
the parent process create a defined number of connections, and then
use some sort of roundrobin/checkout to sign out a handle during a
request?
You're asking "why?" and i'm responding that i dont really know
i just remembered that i'll likely have to up my connection limit
to my db based on the number of apache children
No, don't do that. Front-end your fat apache with squid or a thin
apache and
MaxClients-limit the backend. Keeps concurrency issues on the backend
server
to a minimum too. It's all in the guide.
