[lvs-users] Question about UDP timeouts

[Malcolm Turnbull]

If you are load-balancing a DNS server with LVS via UDP ...
and you delete the real server from the LVS table because of a health check
failure
i.e.
ipvsadm -d

Then the current users in the connection table keep going to that server
until the 300 second UDP timeout has expired.

Now I can do a :

ipvsadm --set 0 0 30

to reduce the UDP timeout to 30 seconds, but i'd rather not set that globaly

I thought the following options may help....
echo 1 > /proc/sys/net/ipv4/vs/expire_quiescent_template
echo 1 > /proc/sys/net/ipv4/vs/expire_nodest_conn
But I'm not using persistence or quiescence....

Am I missing something?

Thanks in advance.


-- 
Regards,

Malcolm Turnbull.

Loadbalancer.org Ltd.
Phone: +44 (0)870 443 8779
http://www.loadbalancer.org/