Ai! That was exactly it. I increased the VM mem, and all is running
fine (34K queries and rising!)!
Thanks a lot,
- Bram
On Mon, Jun 2, 2008 at 3:32 PM, Chris <[EMAIL PROTECTED]> wrote:
> Maybe you jetty need to turning....
> how many memory in your system ?
> Can you show the processes information with the java processes ?
>
> above
> Chris
>
> 2008/6/2 Bram de Jong <[EMAIL PROTECTED]>:
>
>> Hello all,
>>
>>
>> Still running tests on solr using the example jetty container. I've
>> been getting nice performance. However, suddenly between 15400 and
>> 15600 queries, I get a very serious drop in performance, and this
>> every time I run my test, independent of what I'm searching for. The
>> performance STAYS low and doesn't come up again until I restart
>> Jerry/Solr.
>>
>> This what I'm getting. The number between parentheses is the nr of
>> queries done until "now".
>>
>> <snip>
>> average query time this batch ( 2798 ) : 21.7171502113
>> average query time this batch ( 2998 ) : 21.556429863
>> average query time this batch ( 3197 ) : 20.7244367456
>> average query time this batch ( 3397 ) : 20.9529149532
>> average query time this batch ( 3597 ) : 21.7199647427
>> <snip>
>>
>> Then sudenly around 14K my average time goes up 3 fold:
>>
>> <snip>
>> average query time this batch ( 15183 ) : 22.5312757732
>> average query time this batch ( 15383 ) : 27.6089298725 <-
>> average query time this batch ( 15583 ) : 66.8137800694 <-
>> average query time this batch ( 15783 ) : 67.5224089622 <-
>> average query time this batch ( 15983 ) : 68.210555315 <-
>> </snip>
>>
>> I tried taking another set of searches (I'm replaying searches done on
>> our website), but exactly the same pattern occurs. The cumulative
>> evictions for all caches is 0 before and after the slowdown, so my
>> initial thought (i.e. full cache) was not it. I did some further
>> investigating, and it looks like only once every few searches becomes
>> slow. This is the batch of searches for block nr 15583. Every search
>> string is mentioned and then the query_time as reported by Solr:
>>
>> ('electricity', 19), ('killed', 16), ('radio static', 179), ('monster
>> killed', 15), ('heavy machinery', 16), ('killed', 179), ('chimes',
>> 17), ('video games', 17), ('sword', 16), ('construction machine', 17),
>> ('graveyard', 15), ('people', 16), ('yard', 179), ('horn', 15),
>> ('bugle', 14), ('trumpet', 17), ('grass walking', 177), ('walking',
>> 17), ('horn', 15), ('clunk', 14), ('hydraulic', 178), ('jet landing',
>> 16), ('o fortuna', 14), ('large crowd', 180), ('dj', 15), ('hallway',
>> 15), ('scrach', 13), ('jet tires screeching', 177), ('tires
>> screeching', 14), ('ambient', 16), ('electricity', 180), ('tires',
>> 15), ('hospital', 15), ('chimes', 17), ('win chimes', 178), ('wind
>> chimes', 15), ('sex', 15), ('SPACE', 17), ('river', 180), ('thunder
>> storms', 16), ('crash', 15), ('boss', 16), ('thunder', 16), ('car
>> braking', 16), ('vocal', 17), ('vocal dance', 182), ('vocal', 16),
>> ('stream', 16), ('whale', 14), ('space ambient', 183), ('animal', 16),
>> ('pad', 15), ('body', 16), ('crickets', 180), ('fall', 16), ('camera
>> flash bulb', 15), ('arctic', 178), ('flash bulb', 13), ('camera', 15),
>> ('drawn', 16), ('next level', 180), ('timbale', 14), ('navigation',
>> 14), ('bass', 14), ('blips', 179), <snip>
>>
>>
>> Any hints of things I can try would be superb. Having to wait 15000 *
>> 25ms every time I want to try something else to fix this is becoming a
>> bit annoying :)
>>
>>
>> - Bram
>>
>> PS: if you are curious: the things people search for are sounds :)
>> hence the very varied set of search strings.
>>
>> --
>> http://freesound.iua.upf.edu
>> http://www.smartelectronix.com
>> http://www.musicdsp.org
>>
>
>
>
> --
> Chris Lin
> [EMAIL PROTECTED]
> Taipei , Taiwan.
> -----------------------------------------------------------
>
--
http://freesound.iua.upf.edu
http://www.smartelectronix.com
http://www.musicdsp.org
