horms at verge.net.au
Wed Apr 21 02:20:38 BST 2004
On Tue, Apr 20, 2004 at 09:52:29AM -0700, Sean Roe wrote:
> Hi All,
> I am having an issue with my LVS setup. I have 12 web servers behind 2
> LVSes (setup for redundancy and fall over) they are set up for LV-NAT
> and they work fine. My issue is when the primary fails over to the
> secondary. It seems to take several minutes (~5) for the fallover to
> complete. Any ideas why?
> The director interface (Real) on lvs1 (eth1): 10.4.2.43 The director
> interface (Real) on lvs2 (eth1): 10.4.2.44
> The virtual interface shared between lvses (eth1:0): 10.4.2.46
> Here is the real weirdness: The http servers are set to route to
> 10.4.2.46 but when I do a traceroute it hits 10.4.2.43 as its first hop.
> I am thinking that two real interfaces need to be moved to a separate
> network (10.9.xx.xx) or something like that. I thought I read something
> about this a couple of nights ago but I cant find it now. BTW the
> directors are RH 2.1AS Advanced servers.
That is quite strange. First how are you doing your failover.
Are you using heartbeat or keepalived or something like that.
If so it should work.
Also, can you check to see if lvs1 (10.4.2.43) is sending
out an ICMP redirect. That might explain why your real servers
are sending packets via 10.4.2.43 instead of 10.4.2.46.
More information about the lvs-users