[prev in list] [next in list] [prev in thread] [next in thread] 

List:       apache-modperl
Subject:    Re: Deadlocking Apache::Session::MySQL
From:       Perrin Harkins <perrin () elem ! com>
Date:       2010-06-15 19:11:50
Message-ID: AANLkTiny174I7mM_356GYwTLbnDLwnjmIEi3t-ctGCzL () mail ! gmail ! com
[Download RAW message or body]

Hi,

> I've come across a scenario where unchanged code works on 5.6/MP1 such that
> through various routes we twice tie a session object within a request's
> life.  Under the hood it seems to invoke get_lock(id) in the DB on when
> doing the tie. I'm not exeactly sure where but I assume the second
> invocation used to force, or follow,  a release_lock(str) (Also invoked in
> Apache::Session's destructor).

I suspect what happened is that you were using the same connection, so
when you asked for a lock you already had, it was given to you.  Maybe
you're now using two separate connections.

> Am on Apache::Session::MySQL (1.01), with differing (old and new)
> Apache::Session versions.

Chances are you can get your code working just by making sure the same
connection is used in both cases, but my suggestions for the future
would be to either:

1) Use an up-to-date version of Apache::Session, and don't use the
canned MySQL one.  Instead, use Flex so you can use
Apache::Session::Store::DBI, with transactions, on InnoDB tables.
Then you can also set the lock module to Null.

or

2) Use CGI::Session, which may not be perfect but is actively
maintained and generally easier to use.

- Perrin

[prev in list] [next in list] [prev in thread] [next in thread]