RE: Feasability?

On Thu, 13 Mar 2003, Posts wrote:

> I guess you're using the DR as a gateway. To do what you want you'll
> actually need 3 network cards in the DR; VIP (eth0), T1 (eth1), Ethernet
> (eth2). Here's a few ideas:
> 
> 1. On DR, "ifdown eth2", cronjob to monitor eth1, if it dies "ifup
> eth2".
> 2. On DR, give RS-eth1 in lvs.cf an extremely high weight and don't use
> scheduler=rr.
> 3. On RS, "ifdown eth1", cronjob to monitor eth0, if it dies "ifup eth1"
> 4. On RS, cronjob to monitor eth0, set database flag or flatfile for
> apache to monitor and respond accordingly. Basically if eth0 is up lvs
> pings get no response unless flag tells it that eth0 is down and go
> ahead and respond.
> 
> I can think of 4 or 5 others. Putting the logic in the director only
> gives you warm standby, the only way to get hot standby is to have both
> paths alive and well and have the RS determine what to do.
> 
> Kawika Ohumukini
> 2-Tier Software, Inc.

Here is the other kicker.. ;) The Real Server is a Windows 2000 box! ;)
 
> > -----Original Message-----
> > From: [email protected] 
> > [mailto:[email protected]] On Behalf Of Greg Boehnlein
> > Sent: Thursday, March 13, 2003 5:34 PM
> > To: [email protected]
> > Subject: RE: Feasability?
> > 
> > 
> > On Thu, 13 Mar 2003, Posts wrote:
> > 
> > > Give the real server two IP addresses on different subnets 
> > & ethernet 
> > > cards and two entries in lvs.cf.
> > 
> > Great.. But I don't want to load balance them at all. I want to do 
> > hot-spare.
> > 
> > So, if IP A is unavailable, then it routes the traffic to IP 
> > B, which is 
> > "local" on our Ethernet.
> >  
> > > Kawika Ohumukini
> > > 2-Tier Software, Inc.
> > > 
> > > 
> > > > -----Original Message-----
> > > > From: [email protected]
> > > > [mailto:[email protected]] On Behalf Of Greg Boehnlein
> > > > Sent: Thursday, March 13, 2003 1:19 PM
> > > > To: [email protected]
> > > > Subject: Feasability?
> > > > 
> > > > 
> > > > Hello,
> > > > 	I have an LVS cluster running in lvs mode with direct
> > > > routing. I 
> > > > have a client that wants a Fail-Over solution for his 
> > > > Webserver so that if 
> > > > his T-1 goes down, we can re-route the traffic over his 
> > > > Ethernet link. 
> > > > Here is a rough diagram:
> > > > 
> > > > 
> > > > VIP -> LVS Router > T-1 > Real Server (x.x.x.x)
> > > > 
> > > > In the event of a failure, I want the same to follow this path:
> > > > 
> > > > VIP -> LVS Router > Ethernet > Real Server (y.y.y.y)
> > > > 
> > > > Essentially what I am attempting to do is see if LVS can
> > > > support failover 
> > > > using a single Real-Server.
> > > > 
> > > > -- 
> > > >     Vice President of N2Net, a New Age Consulting Service,
> > > > Inc. Company
> > > >          http://www.n2net.net Where everything clicks into place!
> > > >                              KP-216-121-ST
> > > > 
> > > > 
> > > > 
> > > > 
> > > > 
> > > > _______________________________________________
> > > > Piranha-list mailing list
> > > > [email protected]
> > > > https://listman.redhat.com/mailman/listinfo/pi> ranha-list
> > > > 
> > > 
> > > 
> > > 
> > > _______________________________________________
> > > Piranha-list mailing list
> > > [email protected] 
> > > https://listman.redhat.com/mailman/listinfo/piranha-list
> > > 
> > 
> > -- 
> >     Vice President of N2Net, a New Age Consulting Service, 
> > Inc. Company
> >          http://www.n2net.net Where everything clicks into place!
> >                              KP-216-121-ST
> > 
> > 
> > 
> > 
> > 
> > _______________________________________________
> > Piranha-list mailing list
> > [email protected] 
> > https://listman.redhat.com/mailman/listinfo/pi> ranha-list
> > 
> 
> 
> 
> _______________________________________________
> Piranha-list mailing list
> [email protected]
> https://listman.redhat.com/mailman/listinfo/piranha-list
> 

-- 
    Vice President of N2Net, a New Age Consulting Service, Inc. Company
         http://www.n2net.net Where everything clicks into place!
                             KP-216-121-ST