LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: [Linux-HA] local node

To: linux-ha@xxxxxxxxxxxxxxxxxx
Subject: Re: [Linux-HA] local node
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: Jeremy Hansen <jeremy@xxxxxxxxxxxxxx>
Date: Sun, 26 Oct 2003 21:59:58 -0800 (PST)
On Mon, 27 Oct 2003, Horms wrote:

> On Sat, Oct 25, 2003 at 10:19:24AM -0700, Jeremy Hansen wrote:
> > 
> > Let me ask my question and try and put some more details into it.
> > 
> > 
> > Three boxes:
> > 
> >  First LVS machine: eth0 - 208.x.x.x
> >                     eth1 - 192.168.1.40 <- used for backnet traffic
> >                        (like NFS and LDAP)
> >                     eth2 - 10.1.1.1 <- used for heartbeat
> > 
> >  Failover LVS machine: eth0 - 208.x.x.x
> >                        eth1 - 192.168.1.50 <- used for backnet traffic      
> >                            (like NFS and LDAP)
> >                        eth2 - 10.1.1.2 <- used for heartbeat
> > 
> >  Client machine: eth0 - 208.x.x.x
> >                  eth1 - 192.168.1.20 <- used for backnet traffic, etc
> >              no eth2 needed here.
> >   
> > The First LVS machine and the Failover LVS machine are also primary and 
> > replicated LDAP servers.  What I would like to do is use LVS in local node 
> > mode to load balance LDAP traffic on the backnet traffic interface, eth1 
> > from the client machine.  My goal is more for fail over rather then 
> > actually load balancing.

Thanks for the reply Horms.

> In that case perhaps you don't need LVS at all, perhaps you just need
> heartbeat to control heartbeat.
> 
> I would suggest looking at these and working out which 
> best suits your needs.
> 
> http://www.ultramonkey.org/2.0.1/topologies/sl-ha-lb-overview.html

I'm checking this out and I believe I have everything setup in this way

TCP  192.168.1.42:636 rr
  -> 192.168.1.40:636             Local   1      0          0         
  -> 192.168.1.50:636             Route   1      0          0         

on 192.168.1.40 I have:

lo:0      Link encap:Local Loopback  
          inet addr:192.168.1.42  Mask:255.255.255.255
          UP LOOPBACK RUNNING  MTU:16436  Metric:1

on 192.168.1.50 I have:

lo:1      Link encap:Local Loopback  
          inet addr:192.168.1.42  Mask:255.255.255.255
          UP LOOPBACK RUNNING  MTU:16436  Metric:1

cause lo:0 is being taken up by something...

when I telnet from 192.168.1.30 on the same network to port 636, just 
connection refused.  Ldap is listening on both real server IP's:

tcp        0      0 192.168.1.40:636        0.0.0.0:*               LISTEN

tcp        0      0 192.168.1.50:636        0.0.0.0:*               LISTEN

I'm doing local node on these same machines using the outside routable IP 
on eth0 to balance pop and it works just fine:

TCP  208.37.31.234:110 wlc
  -> 208.37.31.237:110            Route   1      0          0         
  -> 208.37.31.236:110            Local   1      0          0         

so ...

I don't get it.  My head hurts.  Thanks for the help!

> http://www.ultramonkey.org/2.0.1/topologies/ha-overview.html
> 
> > I've been trying to use the local node feature without much luck.  I've 
> > done the following from what I've read:
> > 
> >     VIP on eth1 192.168.1.41
> >     Forced ldap to listen on 127.0.0.1
> >     Configured ipvs to look like this:
> >     
> >     TCP  192.168.1.41:636 wlc
> >       -> 127.0.0.1:636                Local   1      0          0         
> > 
> > and then I've tried telneting to port 636 from the client machine and the 
> > connection is refused.  Now this sounds like an arp issue because the 
> > client machines are on the same network as the VIP.  This is my first 
> > guess but this is where I need help.  It would be great to get this going 
> > so I don't have to worry about what happens when LDAP dies.
> 
> That sounds suspiciously like LDAP isn't listening on 127.0.0.1:636.
> Or perhaps you don't have the machine set up to recieve packets for
> the VIP.

when you say machine setup to receive packages on the VIP, you mean have 
the IP alias?  I do:

eth1:1    Link encap:Ethernet  HWaddr 00:07:E9:1B:54:07  
          inet addr:192.168.1.42  Bcast:192.168.1.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          Interrupt:23 Base address:0x10c0 Memory:f4120000-f4140000 


> It almost certainly isn't an ARP issue unless you have another
> machine configured with the VIP, which obviously isn't going to work.
> 
> > Also, I'm confused about using the noarp loopback device in this scenerios 
> > on the Failover LVS server.  I can add this to the ipvs config:
> > 
> > TCP  192.168.1.41:636 wlc
> >   -> 192.168.1.50:636             Route   1      0          0         
> >   -> 127.0.0.1:636                Local   1      0          0         
> > 
> > 192.168.1.50 is the IP of the replicate ldap server and also the IP of 
> > eth1 on that server.  Ldap is listening.  Do I need to add a noarp device 
> > to 192.168.1.50 for the VIP on the Primary LVS server?  I mean I've tried 
> > and this just locked me out of the box because I lost access to ldap.
> 
> Yes.

Ok, did this and no go...

Thanks again.  Any other info I can give you, please ask.  I've been 
working on this for two days straight now.

-jeremy


<Prev in Thread] Current Thread [Next in Thread>
  • Re: [Linux-HA] local node, Jeremy Hansen <=