|
On Wed, 28 Jun 2000, Julian Anastasov wrote:
>
> Hello,
>
> On Wed, 28 Jun 2000, Dave Tyson wrote:
>
> > Hi,
> > I have just setup a LVS system to try and equalise the load of a group
> > of sun servers running Solaris 2.6. I have got LVS working fine (using
> > the 2.2.15 kernel and appropriate ipvs patch). I am using direct routing
> > with wlc and a weighting value to reflect the performance of each server.
> >
> > At present I am testing and have noticed that telnet sessions established
> > via LVS seem to timeout after about 15 minutes idle time. This does not
> > occur if I connect directly to one of the servers. There seems to be no
> > reference to this in the various FAQ's and notes. We definately would want
> > to alter this timeout value to something > 8 hours to avoid flak from
> > users.
>
> On the director:
>
> ipchains -M -S 36000 0 0
>
> 36000 is 10 hours TCP timeout. man ipchains. You can try
> with different value.
I had a 15min timeout problem like this which took weeks to solve. I found
more timers than I could shake a stick at before I found the right one.
Unfortunately I can't remember the solution right now.
Julian is this same as
/proc/sys/net/ipv4/tcp_keepalive_time
whichdefaults to 7200? If so this is not the cause of the 15 min timeout.
Joe
--
Joseph Mack mack@xxxxxxxxxxx
|
