Marc Schmidt wrote: > > hihi :-), > > sorry, it's probably my fault. > > again: > all i need are some good - already tested - values for a bunch of > configuration parameters of the > squid.conf, so that squid will not stop working when i run 20 > simultanous clients against it. > this can't be too difficult??? Depends, system related issues could be the cause of this too. Check process mem. size and swap usage , etc. Since you say there's nothing in cache.log, also check 'dmesg' output on the box and /var/log/messages , check for suspicous errors.
M. > > thanks marc > > Marc Elsen wrote: > > >Marc Schmidt wrote: > > > > > >>thanx for the fast reply! .-) > >> > >>the version is 2.4 stable 7 > >> > >>the access.log reports round about 500 up to 1000 requests than it > >>hangs. > >> > >>there is nothing reported in the cache.log > >> > >>actually i did set up a few things in the squid.conf. http_port and a > >>few other > >>things are among these changes as well, but i have not changed > >>anything in terms > >>of object sizes. squid is configured to not cache anything > >>( acl all src 0/0 > >> no_cache deny all > >>) > >>in the no_cache section. that is the only important thing to mention. > >> > >>one thing i have encountered after inspecting the log files from > >>apache. (the requested url) > >>long time after squid stopped working there are a couple of requests > >>appearing > >>in the access.log from apache: > >> > >>"IP-blablabla" - - [12/Feb/2003:13:37:01 +0100] "-" 408 - > >>"IP-blablabla" - - [12/Feb/2003:13:37:01 +0100] "-" 408 - > >>"IP-blablabla" - - [12/Feb/2003:13:37:01 +0100] "-" 408 - > >>"IP-blablabla" - - [12/Feb/2003:13:37:01 +0100] "-" 408 - > >>"IP-blablabla" - - [12/Feb/2003:13:37:01 +0100] "-" 408 - > >> > >>this means squid did open a socket but didn't send a request, right? > >>that's why there is > >>a 408 (timeout). > >> > >> > > > > All of this I don't understand completely (my fault , probably :-). > > Are you testing an accelerator setup ? > > > > M. > > > > > > > >>ok, but this doesn't help me. i need a hook from somebody where to > >>start reconfiguring > >>the proxy > >> > >>cheers marc > >> > >>Marc Elsen wrote: > >> > >> > >> > >>>Marc Schmidt wrote: > >>> > >>> > >>> > >>> > >>>>hi all, > >>>> > >>>>after writing and starting a little performance test client and > >>>>running > >>>>it against squid, > >>>>the poor little fish stops doing what he is supposed to do: > >>>>serving the > >>>>requests. > >>>> > >>>>the setting is something like this: > >>>>the test client is written in java (using jdk1.4.0) > >>>>there are 20 threads (each simulating a web client) > >>>>each thread requests 50 times the same url > >>>>the os is linux suse 7.3 > >>>>the squid configuration is the one that gets shipped (standard > >>>>squid.conf) > >>>> > >>>>when using 20 threads with 10 iterations per thread everything is > >>>>fine. > >>>> > >>>>so, for my five pens this is more or less a configuration issue. > >>>>isn't it? > >>>> > >>>>anybody out there with a proper squid.conf file that is prepared > >>>>to > >>>>startup squid in > >>>>a high performance mode? or does anybody know what conf parameters > >>>>to screw? > >>>> > >>>>help's appreciated > >>>> > >>>>cheers marc > >>>> > >>>> > >>>> > >>>> > >>> Which version of squid are you testing ? > >>> > >>> What's in access.log during the problem test window ? > >>> > >>> More important : anything in cache.log ,during the problem phase ? > >>> > >>> There is no high performance mode squid.conf so to speak, because > >>> squid is always high performant... > >>> > >>> A standard squid.conf as shipped can not work I think, > >>> at least a listening port for requests must be specified. > >>> > >>> What about cache sizes used etc ? > >>> > >>> > >>> > >>> M. > >>> > >>> > >>> > >>> > >>> > > > > > > -- 'Time is a consequence of Matter thus General Relativity is a direct consequence of QM (M.E. Mar 2002)
