Fwd: inconsistent behaviour with failover on 2 combo router-server nodes?

Hello All,

I'm running keepalived on two nodes that work both as routers and servers,
using lo:0 interface for virtual ip and iptables to mark packets for real
servers. Server 1 has priority over server 2 and thus control the virtual
ip. The workload consists of clients establishing a long-running tcp
connections to the servers.

I've noticed an interesting behavour during failover: when server 1 fails
over, all existing tcp connections to server 1 are gone and clients
reconnect to server 2, which is 100% correct and expected behaviour.
However, when server 1 comes back and takes over virtual ip, some clients
on server 2 get disconnected and then have to dial back to server 1. This
seems to happen only for clients that connected to server 2 when server 1
was offline. The clients that were connected to server 2 before server 1
went down remain connected to server 2.

Is this due to the fact that before server 2 takes over virtual ip on eth0,
clients bind to lo:0 and after virtual ip takeover they bind to virtual ip
on eth0? If so, is there any way to teach them to bind to lo:0 instead so
they do not disconnect the moment another server takes over the virtual ip?

Many thanks,
Alexander

-- 
Alexander Ufimtsev
School of Computer Science and Informatics
University College Dublin, Belfield, D4, Ireland
Phone: +353 1 716 2926  Fax: +353 1 269 7262

------------------------------------------------------------------------------
CenturyLink Cloud: The Leader in Enterprise Cloud Services.
Learn Why More Businesses Are Choosing CenturyLink Cloud For
Critical Workloads, Development Environments & Everything In Between.
Get a Quote or Start a Free Trial Today. 
http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk

_______________________________________________
Keepalived-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/keepalived-devel