moving to a new thread on dev@, which is probably the better place to do this :)
Stas Bekman wrote:
Geoffrey Young wrote:
[...] in the meanwhile I'll see if I can't prototype a perl interface for killing both processes and interpreters.
I'd suggest discussing this before spending any time prototyping things. I can't quite see how are you going to kill interpreters and how are you going to make the decision which one to kill. You don't want to kill an interpreter you are in, because you have no idea whether it's one that is of the biggest size. The idea was to have an garbage collection thread which will kick in when the size of the process goes beyond the limit and measure the idle interpreters in the pool and kill the big ones. Or do you have something else on your mind?
yeah, I don't know how to handle it. as I mentioned back in vegas, I don't see the reason for killing off interpreters - the process is really the only thing you can measure, so once it gets too large probably the best way is to kill it gracefully and let another one spawn in its place. there were some other things we discussed (such as a two-tiered limit, where the first tier knocks off an interpreter, while the next tier knocks off the process) but I'm not sure if that is realistic or not.
Remember that on winFU there is only one process, meaning that you are going to do a gracefull restart every time the limit is passed. And it's going to be very bad, because no new requests will be handled untill all threads will quit, and considering how slow perl_clone is, it's not so practical.
certainly, just limiting the process as a whole is easier to implement for the reasons you just listed, and I'm content to leave it at that. everyone else seemed to want the thread-killing api, though...
I say, just implement the preforking for now (should be a trivial porting once child_terminate is added), and leave the threaded mpm alone untill we will be forced into providing the sizelimit for it or will have the tuits to work on it. This is really not that important for now, as hardly anybody uses threaded mpms yet. Havind a solid prefork support with all the tools as in mp1 is what we should strive for 2.0 release. Next we can work on getting the threaded things better.
__________________________________________________________________ Stas Bekman JAm_pH ------> Just Another mod_perl Hacker http://stason.org/ mod_perl Guide ---> http://perl.apache.org mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com http://modperlbook.org http://apache.org http://ticketmaster.com
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
