[Xapian-discuss] Remote database search issues

[Ron Kass]

Olly Betts wrote:

> On Sun, Oct 28, 2007 at 02:23:19AM +0200, Ron Kass wrote:
>   
>> CAN you run searches with tcpsrv in parallel?
>> or can you NOT run searches in parallel with progsrv?
>>     
>
> Both can run searches in parallel.  In the case of xapian-tcpsrv, it
> forks for each new connection (on Windows a new thread is started),
>   
It would be interesting then to find out why tcpsrv is so much slower 
then. But lets deal with the first slowness first then..
>>> Until we track down your uninitialised weights issue, I don't think
>>> performance testing is going to give reliable results.
>>>       
>> What can we do to help track this? If it requires debugging part of the 
>> code, I would be happy to. If you can let us know what to do or at least 
>> point us in the right direction, that would be great.
>>     
>
> Could you rebuild xapian-core without optimisation and with assertions
> enabled, and then rerun under valgrind?  Details in this message and the
> followup:
>
> http://thread.gmane.org/gmane.comp.search.xapian.general/5295/focus=5322
>
> Cheers,
>     Olly
>   
We use gnu c++ compiler and valgrind 3.2.1

We configured and recopmpiled with...
configure/configure CXXFLAGS='-O0 -g' --enable-assertions ; make; make 
install

Then ran valgrind again and then with --leak-check=full just in case

Here are both outputs:
http://www.pidgintech.com/other/fts/test/valgrind_2-4-6-8-10_nooptimization.txt
http://www.pidgintech.com/other/fts/test/valgrind-full_2-4-6-8-10_nooptimization.txt


Note:
* Used the Dell server (the non-xen server, with the latest svn head 
version, which had 0.2 seconds for this searches).
* Although not showing specifically for these 6 runs (in the 2 tests), 
estimates still vary. (We used a custom set weighting scheme for the tests)


Cheers,
Ron