Re: [Mldonkey-users] Latest CVS, fasttrack: too many sources, too much memory used.

[b8_bavard]

>  First, sorry about my english.
>  
>  I downloaded and built the latest CVS version of MLnet (26/06/03). (this 
>  thing happens on previous versions too)
>  
>  But when I start to download form fasttrack a file with a high 
>  availability, the list of sources starts to grow over the "max sources 
>  per file".
>  When this thing happens, over 300MB of physical memory was used, and the 
>  downlink bandwidth receives more data than mldonkey write to the file. 
>  Per example: PPPload statistics show 1,8GB received, when the only file 
>  downloaded is 700MB long, and this file was at 50 percent of the total ( 
>  too much junk in the download link?).
>  
>  If I kill the core and manually edit files.ini to cut a large amount of 
>  sources ( thanks a lot for making this file human readable :-)), and 
>  start mlnet again, everything works fine for a couple of hours.

Yes, this is a known issue. The fasttrack plugin does not use any
sources management policy (contrary to the edonkey plugin). The next
step in the development is by the way to make all plugins use the
edonkey sources management, so this issue might be solved
soon. Another problem is that the current swarming mechanism seems to
request the same ranges from several sources, because the ranges are
chosen too early. We are aware of this problem too. 

Thanks for your remarks,

PS: on 2.5-devel-4, non-edonkey downloads don't terminate, because
they cannot be verified... will be fixed too.

- b8_bavard (mldonkey)
-------------------------------------- 
Homepage: http://www.mldonkey.net/
--------------------------------------