Can processes share a DBI connection?

[jecouto]

Is it safe to open a connection to a database and then use that same connection in all the children your process forks? Or is it necessary to open a connection for each child?

Originally posted as a Categorized Question.

[Zaxo]

Each process should open its own DBI connection. Processes will readily share one if it's opened before fork, but that has bad side effects, and no good ones that I know of. Two processes talking over one connection may get a db server confused, and the first process to end will take the serverside connection down, with no notification to the surviving process. Further confusion may result when the survivor tries to close its copy.

Moral - Don't call &DBI::connect before a fork, even if one process never uses the connection.

[grantm]

I agree with Zaxo - don't do it. If you have a need to multiplex multiple processes into one DBI connection, you might want to investigate DBI::ProxyServer

[pajout]

I agree with previous answers. If problem is many opened connections, think about Apache::DBI (if forked processes are httpd's :) ) or about SQL Relay, which is daemon realizing connection pooling against various DB engines.