On Tue, 9 Nov 1999, Martijn Pieters wrote:
Although apparently not harmful, I do suspect that eventually the process will run out of threads.
Hmm.. The error occurs within the THUNKed database adapter, it's a problem with the ThreadLock. I think something goes wrong with the _register method of the THUNK class (either register or self._begin fails), causing the freshly acquired lock to be released again.
Then, when the transaction is finished, tpc_finish is called, which releases a unlocked lock, which throws the exception. But because the transaction was already finished, we get the aforementioned log message. Zope just continues to run.
The question is, what, if any, side effects does this cause, and how does this problem occur, so we can put an end to it?
In my case the problem is real. I have a UserDb folder at the top level (/) handling all Zope users. What happens is that a few hours after inactivity (no accesses) Zope is no longer accepting any logins. You can view the public pages, but UserDb seems to deny all login attempts. I have had to restart Zope whenever this occurs to get into the server. My config again: - Red Hat Linux 6.0 - Zope 2.0.1 (binary tarball from zope.org) - Solid RDBMS v2.3 - Jeff Rush's ZODBCDA/Solid (binary tarball) - UserDb 1.3.0 Is there anything I could do to provide more information in order to get help solving the problem? Thank you, -Teemu