[rt-users] RT2 Speed/performance problems.

[Simon Cozens]

Maren S. Leizaola:
> 	Much regretably I've had to pull RT2 out. For us even a machine
> with 2 cpus, raid, 512MB RAMwith 5000 tickets in the database (99% of
> which were closed) the system would take ages to refresh the display. As
> it was costing us more on having to wait for the pages to refresh and do
> maintenance on it than the time required of handlign all e-mail manually.
> I think we are now in a much better position with our e-mail support
> without using RT than with RT.

Did you benchmark it? Did you profile it? Where were the bottlenecks for
you? 

Did you see problems with the command-line interface as well as the web
interface? Was the backend fast enough? How long was it taking for mail
to be processed through the mail gateway?

Were you running Postgres or MySQL? Were your indexes working OK? Was
the database being vacuumed/analysed regularly? Was raw SQL access fast
enough? 

Did you add ?Debug=1 to a URL to determine whether or not it was an RT
problem or a browser rendering problem? (That surprised a few of our
users.) Were the machines you were running the browsers on OK? Were you
using sane browsers?

Was the Mason cache being populated properly? Did you tune your Apache
config? Did you use Apache::Status?

> end is PHP based where deamons do not eat up 62MB of RAM or more...  I

Aha! It's going to be a problem with your mod_perl configuration. We're
seeing the same sort of thing, and after considerable profiling, I'm
convinced it's Apache/mod_perl at fault. 

At a guess, you're running on old version of Perl or mod_perl which
doesn't have the advantage of Doug McEachern's wonderful recent work on
sharable variables or his many memory leak fixes.

Other people have seen RT running incredibly quickly on less powerful
machines, so I'm not going to dismiss it as a problem at RT's end; I
think it's a problem with the way we've set it up.

Simon

-- 
<evilPetey> I often think I'd get better throughput yelling at the modem.