What front-end and storage mode do you use?
Roy Nasser wrote:
>
> We have about 60,000 URLs in total... and only about 2,300 are indexed... I
> think that the problem is not really in the DB, i think that it is more of
> an indexer problem, stalling at URLs that to not answer, and maybe stopping
> in the middle of indexing...
>
> I have let 3 indexers run for about 1 hour and it has indexed only about 200
> sites, on a dedicated line (i.e. it is not a bandwidth limitation, i am
> sure...
>
> Oh well... hehehe... I am a bit frustrated, but i will try what you
> suggested...
>
> On another note, I am still unable to make any Frontend search my index...
> No matter what backend, no matter what keyword, i always get 0 found... Can
> someone help me? (I dont get any errors, just get 0 found)...
>
> Please help me, aterall, there is no use indexing if i cant read the
> index...
>
> -----Original Message-----
> From: Shane Wegner
> To: Roy Nasser
> Cc: '[EMAIL PROTECTED] '
> Sent: 08/10/2000 22:15
> Subject: Re: UdmSearch: Question: way to compile for more performance
>
> On Sun, Oct 08, 2000 at 06:45:07PM -0300, Roy Nasser wrote:
> > Hi,
> >
> > I can see that it indexes the documents, but it sometimes stalls for
> over 2
> > minutes on a single URL... We have about 58,000 pending URLs, and it
> really
> > takes a long time to index...
>
> How many URLs do you have in the database? 2 minutes is a
> bit long but try adding an index on url.next_index_time.
>
> Indexer if I recall correctly does select * from url where
> next_index_time<=unix_timestamp() limit 32;
>
> So every 32 URLs you will get a pause if you have many URLs
> in the database. I don't have it indexed but only have
> 74,000 URLs being indexed. The pause is about 10 seconds
> or so every 32 entries.
>
> Anyways, you can give that a try and see if it helps.
>
> > Is there a special compilation option, or should i pass a special
> parameter
> > to the indexer for it to fetch like, 5 or 10 URLs simultaneously?
>
> Well there may be gcc options which could improve your
> performance. The various -m options come to mind but from
> my work with UDMSearch, it's pretty fast and is generally
> the database which is the bottleneck.
>
> Cheers,
> Shane
--
Alexander Barkov
IZHCOM, Izhevsk
email: [EMAIL PROTECTED] | http://www.izhcom.ru
Phone: +7 (3412) 51-32-11 | Fax: +7 (3412) 51-20-80
ICQ: 7748759
______________
If you want to unsubscribe send "unsubscribe udmsearch"
to [EMAIL PROTECTED]
