|
Hello,
On Wed, 10 Oct 2001, Roberto Nibali wrote:
> Hi Julian,
>
> > No, I don't want it, it is a crap that is already present,
> > I never use it. Of course, I didn't used CONFIG_IP_MASQ_DEBUG may be
> > from 2 years ago :) It is not compiled by default
>
> So bloody take over the code and drop this crap. Think about people
> trying to understand the TCP/IP stack in linux only having a limited
> time of one lifespan.
:) What to say, you see how many changes adds LVS to the masq
code in 2.2 and always by trying to keep the masq code when the LVS
is not compiled. May be this is the only change that touches the
2.2 masquerade without preserving it for !LVS. For good.
> > Nobody from the maintainers wants/have the time to touch the
> > masquerade
>
> Ok, I mean it's working, but some cleanup work would be a piece of
> cake for you.
:)
> > > Thanks for the clarifications, I will certainly urge the people
> > > attending
> > > the course to patch in your fresh patch :)
> >
> > Very good :)
>
> Well I set up 10 Sunblade 100 today (sponsored by the swiss federal
> institute of technology) and did some preliminary testing for the
> workshop tomorrow and: It didn't crash. I made 2.5 Million requests
> over the LVS_NAT and downloaded stuff over another interface, also
> NAT, the same machine and kernel. No crash. But I couldn't verify if
> it is faster then the old one. From to code, yes, but not from any
The same is here. I didn't found today a way to flood my cluster
properly. I feel a little difference but for now this is only a feeling :)
I can't fully load my director. Any difference will be visible with
long packets.
> tests. I think this is a thing Joe can do a lot better since he's
> already done extensive tests.
Yes, may be results from netpipe will be useful.
> Good work, Julian. Cheers,
> Roberto Nibali, ratz
Regards
--
Julian Anastasov <ja@xxxxxx>
|
