On Jan 23, 2008, at 8:36 PM, Peter Jakobsson wrote:
On the load test, the stress testing package reports about 84%
'Socket Errors'. Can anyone throw any more light on this with
respect to how 4D behaves, namely:
- what exactly is a 'socket error', is it an OS level problem,
nothing to do with 4D? or a 4D overload ?
- are there any settings in 4D Prefs (or OS prefs) which can
influence this aspect of performance ? (e.g. Max web processes,
Number of requests per connection etc)
b.t.w, when I tried this last using my Mac Powerbook as a web
server, I didn't experience anything like the dropped connections,
just slow page loading (and that's what I'm looking for).
Environment:
4D 2004.5 Windows (Standalone & Client Server [Client as web server])
Active 4D 4.2
I had a lot of problems with 2003(client as web server), but have not
seen any in 2004. My main problem was with with file uploads using
Safari. Research on the socket error pointed to tcp_keepAlive, which
was not configurable in 2003. I switched to using Apache for file
uploads, so I can't verify that 2004 fixed that problem.
2004 has a setting for tcp_keepAlive, but not sure how it works - you
might try changing the setting. I also put Apache in front as proxy,
but did not move static pages to Apache - just don't have that many.
Did move them to "web decoy" to take load off of Active4D. Moving to
Apache would shortcut the process one step better.
Steve Alex
_______________________________________________
Active4D-dev mailing list
[email protected]
http://mailman.aparajitaworld.com/mailman/listinfo/active4d-dev
Archives: http://mailman.aparajitaworld.com/archive/active4d-dev/