On Tue, Jun 21, 2011 at 5:03 PM, Rolf Schaufelberger <r...@plusw.de> wrote: >> My approach to this problem would be to run separate httpd servers for >> each version of the code you need to support and use a proxy in front >> to do the dispatching. > > yes, I've tried this too, but this leads to a big memory footprint too. If > you have 10 sites and make separate (preforking) instance for each vhost, > with StartServers = 2 and set the other parameters (Min/MaxSpareServers etc) > to low values, you quickly get 20 oder 30 fat apache processes. Ok, you can > get this running with enough memory, but the idea of instance pools seems > to be a better and more elegant solution for this.
I don't understand what would be different. My suggestion was to run one httpd server for each version of the code that you need, possibly with virtual hosts on each if there are multiple hosts that use the same version of the code. If you had an "instance pool", wouldn't it also be multiple processes (or threads if you prefer) per version of code? I don't see how you could avoid that. Were you thinking you'd reset the perl interpreters completely between requests so that one could be shared between multiple versions of code? > I've discussed this once with Torsten Förtsch who is a mod_perl expert > nearby, his opinion was, that it wouldn't be very much code that has to be > added for this. Torsten is a major contributor and I'm sure that whatever he had in mind would be cool. - Perrin