Re: Director Failover

> No, lvs doesn't work. I can't telnet to the VIP from a
> client pc. It could be the director cannot detect the
> realserver as seen in the log, so when telnet request
> come in, it can't redirect to the realserver. That's
> why I get "connection refused" when I telnet the VIP
> from the client pc. Is my assumption correct ?

if ipvsadm -ln doesn't show the RS, you're right. If it shows the RS,
then you haven't configured the VIP on the RS !

> I think it has to do with my PC setup or kernel setup
> as I don't encounter the same problem in another PC as
> director (even with the expect/send=""). I'm sure
> because the same problem occur when I fall back to the
> old Piranha without patch in this PC. But I think I
> did the same for the ipvs patch in both PCs.

Check (with md5sum) you have the same binaries. If not, rebuilt piranha
on both machines !

> The funny thing is I can telnet/ping the realserver
> from this PC. So whst does nanny actually do to poll
> the realserver.

AFAIR, if you don't specify send nor expect, it will just try to connect
to the RS.

I've just tried with the following setup and it works fine.

virtual test-telnet {
        active = 1
        address = 172.22.48.212 eth0:2
        vip_nmask = 255.255.240.0
        port = 23
        send = ""
        expect = ""
        load_monitor = "/etc/sysconfig/ha/ruptime-piranha %h"
        scheduler = wrr
        protocol = tcp
        timeout = 5
        reentry = 15
        quiesce_server = 0
        server cluster1 {
                address = 172.22.48.208
                active = 1
                weight = 500
        }
        server cluster2 {
                address = 172.22.48.209
                active = 1
                weight = 500
        }
}

> Thanks
> Chee Chye
> 
> Because of this
> 
>  --- Sébastien Bonnet <[email protected]>
> wrote: > > I have another problem here. I got the
> following
> > message in the log:
> > > nanny[1228]: READ to 192.168.123.198:23 timed out
> >
> > I guess that because you don't have a send string,
> > nanny just connects
> > to telnetd, which doesn't reply anything. If you
> > really want to setup a
> > telnet cluster, consider using approriate
> > send/expect string or an
> > external test program.
> >
> > > I can telnet/ping 192.168.123.198 (the realserver)
> > from the
> > > director. What could be the problem ?
> >
> > Exposed earlier. Just to make sure I've understood :
> > it does work, but
> > you get these nasty lines in the logs, right ?
> >
> > > rsh_command = rsh
> > >          load_monitor = uptime
> >
> > Too bad ! This is the least efficient way to collect
> > remote loads.
> > Consider using (a modified) ruptime : start rwhod on
> > all realservers,
> > name them by their real, not FQDN, hostname in
> > lvs.cf, and use the
> > following script as load_monitor.
> >
> > [sbt@ssca4010 sbt]$ cat
> > /etc/sysconfig/ha/ruptime-piranha
> > #! /bin/sh
> >
> > ruptime | grep -E "^$1[[:space:]]+" | sed -e 's/
> > load / load average:
> > /g'
> >
> > --
> > Sébastien Bonnet
> >   Centre de contacts - Experian France
> >
> >
> >
> > _______________________________________________
> > Piranha-list mailing list
> > [email protected]
> >
> https://listman.redhat.com/mailman/listinfo/piranha-list
> 
> __________________________________________________
> Do You Yahoo!?
> Promote your business from just $5 a month!
> http://sg.biztools.yahoo.com
> 
> _______________________________________________
> Piranha-list mailing list
> [email protected]
> https://listman.redhat.com/mailman/listinfo/piranha-list

-- 
Sébastien Bonnet
  Centre de contacts - Experian France