Ryan Bloom wrote:
>
> I have a problem with tagging the way the tree is right now. The threaded
> MPM is completely bogus,
Bull. After fixing the breakage you introduced late in June and early
July, it's working pretty well actually.
> and I don't believe that the group should be
> encouraging it's use in production servers. As the code stands right now,
> the threaded MPM can not be gracefully restarted or shutdown on a
> lightly-hit server.
You obviously haven't tried it in a while.
> I see two ways to fix this. Hold the tag/roll until the worker MPM is finished,
> hopefully sometime this week, or switch the default MPM to prefork.
>
> I want to stress that I am perfectly fine with just switching the default MPM
> to the prefork MPM. Leaving the default at threaded, makes people think that
> we expect this MPM to work properly, which we really don't.
I expect threaded to work properly, except for two issues:
* not getting core dumps on Linux, and
* processes staying in mutex land with a "G" in the scoreboard on a
server with no incoming requests (yawn).
If anybody has problems with the threaded MPM on servers that actually
have incoming http requests, please let us know.
Greg
