[Spread-users] Invalid message in high load group

[Marcin Kuthan]

Hi

My application send between 1000 and 10000 messages per second). On 150
per second all works correctly. Yuval, can you test your pure C client
with 10000 messages per second? My testing topology is very simple ->
one spread daemon (on eth0 interface, not loopback) in group and client
receiving self messages.

Yair, I think that is not in Java interface because invalid message has
always start with valid group name. This is not accidentally situation
with accidentally data. This is not corrupted message, this is other
messages that spread client expected.

Scott, If the problem is caused with SO_SNDBUF and SO_RCVBUF settings
(http://commedia.cnds.jhu.edu/pipermail/spread-users/2005-April/002496.html) or other Java IO error, invalid messages should be randomized (as I wrote - invalid message always has identical structure and content).

Best regards,
Marcin

> Hi,
> 
> This would be my guess too - probably something in the Java
> interface...
> 
> Cheers,
> 
> 	:) Yair.
> 
> Yuval Lifshitz wrote:
> > I'm not sure the problem is in the daemon.
> > I use spread to send 100 bytes messages at 150 per sec
> > with the C interface (which is much thinner), and does
> > not encounter malformed messages. The problem might be
> > in the Java interface or maybe topology depended.
> > 
> > Yuval.
> > 
> > --- Marcin Kuthan <mkuthan at pit.edu.pl> wrote:
> > 
> > 
> >>Hi Pradyumna
> >>
> >>Thank our for your help. I study attached patch, it
> >>seems that temporary
> >>resolve my problem. My application trace messages
> >>order on high
> >>abstraction level, and losing messages on transport
> >>layer is safe for
> >>one.
> >>
> >>This problem look like a bug inside spread deamon.
> >>Unfortunately I'm not
> >>an advanced C developer, and I can't fix the spread
> >>deamon. Is the any
> >>person on groups who could help (maybe spread
> >>developers)?
> >>
> >>
> >>
> >>>Hi Marcin, 
> >>>
> >>>We had a similar problem with spread when we
> >>
> >>loaded it
> >>
> >>>quite a bit. First of all I suggest that you use
> >>>message type "UNRELIABLE_MESS".
> >>>
> >>>When we did this the frequency of the problem did
> >>>decrease but continued to happen.
> >>>
> >>>In this case there a workaround has been
> >>
> >>implemented
> >>
> >>>in the file SpreadConnection.java which I have
> >>>attached with this file.
> >>>
> >>>Note: There exists a high risk of losing messages
> >>
> >>and
> >>
> >>>is *NOT* a very safe workaround.
> >>>
> >>>Hope this helps,
> >>>Regards
> >>>Prady
> >>>
> >>
> >>-- 
> >>Telecommunication Research Institute
> >>Poligonowa 30, 04-051 Warszawa, Poland
> >>
> >>Marcin Kuthan
> >>mail:     mkuthan at pit.edu.pl
> >>phone:    +48 (22) 486-53-46
> >>
> >>
> >>_______________________________________________
> >>Spread-users mailing list
> >>Spread-users at lists.spread.org
> >>
> > 
> > http://lists.spread.org/mailman/listinfo/spread-users
> > 
> > 
> > 
> > 		
> > ____________________________________________________
> > Start your day with Yahoo! - make it your home page 
> > http://www.yahoo.com/r/hs 
> >  
> > 
> > _______________________________________________
> > Spread-users mailing list
> > Spread-users at lists.spread.org
> > http://lists.spread.org/mailman/listinfo/spread-users
> > 
> > 
> 
> 
> _______________________________________________
> Spread-users mailing list
> Spread-users at lists.spread.org
> http://lists.spread.org/mailman/listinfo/spread-users
>