|
On Wed, 2009-07-22 at 10:29 +0200, Martijn Marsman wrote:
> We have noticed that node 1 gets all the connections, after half an hour,
> node 1 goes idle and node 2 gets all the connections. User complain about
> beeing thrown out of their session on the site and go back to the frontpage.
This is a classic example of "the ARP problem".
This is an LVS-DR setup, isn't it? The realservers (nodes 1 and 2) are
responding to ARP queries for the VIP, and the traffic is bypassing the
director.
There are several ways to solve this but the best way is to use the
appropriate sysctls as follows:
# LVS switches
net.ipv4.conf.lo.arp_ignore = 1
net.ipv4.conf.lo.arp_announce = 2
net.ipv4.conf.all.arp_ignore = 1
net.ipv4.conf.all.arp_announce = 2
You may need to amend that depending on the interface you put the dummy
VIP onto. That stops the realserver kernel replying to "ARP who-has"
requests for the VIP.
Graeme
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users
|
