This is a known issue related to the winnt MPM. You're likely seeing your worker segfault. On linux, there are usually extra processes so if one segfaults there's another process that can step in fast enough. But on the winnt MPM, there's one master and exactly one multi-threader worker. So when that worker blows up, the master needs to completely tear it down, and bring it back up, which is what you're seeing.
Look for a bug in user-code, not in the mod_perl binary. Issac On 29/04/2013 13:49, Michiel Beijen wrote: > Hi Steve, > > On Sat, Apr 27, 2013 at 6:27 PM, Steve Hay <steve.m....@googlemail.com> wrote: >> For those who are interested, I've now uploaded binary builds of >> mod_perl-2.0.8 and libapreq2-2.13 components compatible with Apache >> Lounge 2.2.24 Build 25 February 2013 (win32) and each of: > > Thanks for providing these binaries. > > Using my mod_perl based application I get apache restarts from time > to time. This is not related to your binary per se (I also have it > with kmx'es mod_perl binaries for perl 5.12) but it is related to > mod_perl on Win32; Linux does not have this. > The bad thing is that I don't have any specific information in the log > file; after some requests, the process just restarts. If I use the > application as CGI, this does not happen; but of course the whole > thing is very slow. > > Can you maybe also provide a debug version of mod_perl by rebuilding > mod_perl with MP_DEBUG=1 ? > > -- > Mike >
