In the Wiki FAQ about performance, I found the following run on
sentence. I’ve reread it a dozen times and I’m still not sure that I
completely understand what the author meant. If someone would be
willing to parse this out for me, I’d be happy to rewrite it for the
wiki.
Memory. RT uses a lot of memory in total. You need adopt amount of
process(Apache, FastCGI, DB) to fit into 70% your memory under
average workload. For example, if you have RT on Apache/mp/mysql
with ~20-30 end-users, then it’s enough to have 4-6 Apache processes
this also control DB processes/threads because eache Apache instance
with Apache::DBI has own DB connect each DB connect require one or
more DB process/thread to support. Also people report that one
FastCGI process can serve 2-3 Apache processes. You also have to
leave room for memory usage growing, example: Apache/mp process has
10-25MB size under average workload, but on big incoming message
memory footprint could grow up to 200-300MB. Out of memory is most
perfomance problem it’s blow up HDD activity for swap out/in memory.
In particular, “adopt amount of process to fit into 70% your
memory”. Does this mean calculate the memory usage of Mod_Perl or
FastCGI processes and come up with a number of processes which
wouldn’t exceed 70% of your total memory? Ie,
60mb per process
1024mb total ram
Set Apache MaxServers = 11 ? (60mb * 11 = 660mb of ram)
Jo Rhett
Net Consonance : consonant endings by net philanthropy, open source
and other randomness