[mnet-devel] Re: [mnet-cvs]this 1-line patch should fix the "we dont hello quickly en...

[Zooko O'Whielacronx]

 Arno wrote:
>
> this won't do it. It still seems to set that timestamp without knowing 
> that the MT was reachable. The failure is usually a timeout, not one of 
> the cases checked in that function above the newly inserted line.

I thought that the bad case we were seeing was that you were behind firewall
and didn't decide on your relay server soon enough, so it didn't even try to
send hello.  You seem to be saying that the bad case is that you tried to
hello, but your hello failed.  Is this common?  Why does it fail?  Because you
are sending to MTs that are currently down ?

In any case, the 1-line patch I posted should be an improvement -- you should
definitely not touch the "lasthellosent" timestamp when you in fact didn't even
try to send a hello!


> I am afraid one has to put that into hello response handling....

That would be good.  Shouldn't be too tricky to implement something that says
"whenever a hello fails, initiate another hello with a randomly chosen MT" or
something like that.

> I am testing other patches in the moment, which produce nice results, 
> although maybe more traffic than necessary.

Please tell us about these!

--Z



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
mnet-devel mailing list
mnet-devel at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mnet-devel