[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: Pathological TCP behavior running ls(1) over SSH
2008/1/28, Andreas Gustafsson <gson%gson.org@localhost>:
> Darren Reed wrote:
> > > This looks similar to what Tomas Svensson reported in 2002 [...]
> > This is a different problem/bug to what you are seeing.
> I think the cause of all those tinygrams being sent in the first place
> is probably the same, but the lack of fast retransmit when one of them
> is lost is a new and separate issue.
> > I'm going to bet that this is somehow tied up with SACK.
> > All of the retransmitted (duplicate) ACKs include 1 or more
> > ranges of bytes from SACK.
> > So it seems worthwhile testing to see if that is a cause here:
> > sysctl -w net.inet.tcp.sack.enable=0
> I think you're right. If I disable SACK, the pauses disappear, and
> tcpdump shows that fast retransmit is now happening:
Nice to know. Since I turned off sack, my idle ssh connection doesn't
drop after 5 minutes anymore. I always thought it would be my
Main Index |
Thread Index |