|
Sébastien Arnaud
arnaudsj at emedialibrary.org
Thu Dec 15 23:30:12 EST 2005
Thank you Graham and Jim for the pointers! I guess I was directly affected by the bug in question in 3.1.4 ! Since I was running 2 apps as virtual hosts, 10 PythonOption instructions per app, I also have 1 Load Balancer in front of those app servers making 1 request each minute to poll each app, it was making me loose 2x10x25 = 500 bytes / min ! As soon as 3.2 is golden, I will grab it and run it on our test farm before I push it out to our prod environment, sounds like 3.2 is full of sweet bug fix and enhancements! Any idea when it will be officially released? Also, I noticed that Apache had released 2.2 officially, how does mod_python 3.2 handles this, will it work on apache 2.2? I am asking this, because I just did a port sync on my Mac and I just saw that apache 2.2 was already ready to replace my 2.0.55 install! Thanks again for the help! Sébastien Arnaud eMedia Library, inc sebastien at emedialibrary.org On Dec 15, 2005, at 5:07 PM, Graham Dumpleton wrote: > FWIW, you might consider upgrading to mod_python 3.2.5b if you can. I > believe that that version eliminates a few memory leaks in certain > circumstances in mod_python itself, although I don't remember the > details. This may or may not be contributing. My understanding was > that > the mod_python leaks weren't that great, but if you are handling a lot > of requests it may be noticeable. You also may want to configure > Apache > to recycle child processes after a certain number of requests to avoid > overblown processes. > > Sorry, I know this doesn't really help with tracking down the > source of > any problem if in your code. :-) > > Graham > > =?ISO-8859-1?Q?S=E9bastien_Arnaud?= wrote .. >> Hello! >> >> I am running mod_python on 2 identical production server for 2 small >> apps I wrote which basically render over 100 of different SVG charts. >> I am running Gentoo on those servers, mod_python 3.1.4, apache 2.0.55 >> in prefork mode, each server has 1 Go of mem. >> >> Both applications have been in beta testing for like 1 month or so, >> and I just noticed today that the memory usage was really high for >> each apache process running (max of 10 set in the conf file). When I >> started those apache process 30 days ago they stabilize at 2.0 % of >> memory used, but somehow after 30 days of users testing the app, it >> went to 10% per apache process, which made the swap usage go to 1Go >> (over 2Go avail). I am doing quite a lot of caching of XSL files in >> the app, but this is suppose to happen at launch time, meaning I >> instantiate all the objects I need at the first call of the >> mod_python handler and they persist through the life of those apache >> processes. Does somebody know anyway to inspect each apache process >> to see where the memory is being used? Basically I am trying here to >> figure out if I have some kind of memory leak going on, or if there >> is something I miscalculated when caching the large amount of XSL >> files in persistent Python objects. >> >> Apart from that, those 2 little applications fly in terms of speed! A >> lot of people (including me) got really blown away to witness the >> speed of those 2 apps on those "old" servers (which are quad PIII >> 550Mhz btw). >> >> Thank you in advance for any insights so that I can get to the bottom >> of this! >> >> Cheers! >> >> Sébastien Arnaud >> eMedia Library, inc >> sebastien at emedialibrary.org >> >> >> >> >> _______________________________________________ >> Mod_python mailing list >> Mod_python at modpython.org >> http://mailman.modpython.org/mailman/listinfo/mod_python > _______________________________________________ > Mod_python mailing list > Mod_python at modpython.org > http://mailman.modpython.org/mailman/listinfo/mod_python
|