Graham and Jim, thanks for the advice. As to prefork: as long as APR has threads and mutex and all these fine things, I am pretty certain the mod_h2 itself does not have a problem with prefork.
The concern I have is: if people use prefork to have code running which is not thread-safe, that will not make it safe in mod_h2. Or if people use prefork as an indication to avoid mutex and other sync helpers, they will be surprised. I am just adding a WARNING log on startup (once) that says that mod_h2 in prefork might be a source of unpleasant surprises. But I leave mod_h2 running. //Stefan > Am 09.06.2015 um 17:04 schrieb Graham Leggett <[email protected]>: > > On 09 Jun 2015, at 3:37 PM, Stefan Eissing <[email protected]> > wrote: > >> My understanding is that one uses mpm_prefork when threads are to be >> avoided. This is not what mod_h2 is about and so I consider disabling the >> module in a prefork configuration. My thinking: if threads are not a >> problem, why not run worker/event? > > Getting prefork to work would be a good thing. > > The prefork MPM is the “tank” MPM - regardless of what the client does, > regardless of what code is being spawned by httpd, any leak/crash affects > that connection only. > > mod_h2 might not be fully featured or fully performant under prefork, but > that’s fine in theory. > > Regards, > Graham > — > <green/>bytes GmbH Hafenweg 16, 48155 Münster, Germany Phone: +49 251 2807760. Amtsgericht Münster: HRB5782
