Hi,

it depends on the worker model you use with Apache.

When you use the forking model than every child of the Apache master process should create a database handle as every request gets handled by one child. The request->do-something on the database->return-response-cycle should be safe. In this scenario you only have to be sure that each cycle is terminated by a correct commit or rollback. It would be bad if you open a transaction with one request and close it with another.

When you use the threading model in Apache it gets IMHO more problematic. You have to synchronize the access to a global DBI handle and you have to be sure (as above) that ONE transaction is not used by several requests.

I just looked at the docs of Apache::DBI explaining the same in more detail.

Best regards
McA