[lvs-users] Failover and resets
ABhat at trustwave.com
Fri Oct 22 00:32:22 BST 2010
Many thanks Simon.
I'll look into the synchronization daemon if that's needed.
The sending of resets however appears to be the correct and desired
behaviour in this case so that's great.
Do you know of some documentation that may help answer any further
questions about how failovers work and the finer details about it?
Once I get into the office, I'll send in our config for ssh load balancing
in case someone sees any obvious mistakes.
On 10/21/10 5:53 PM, "Simon Horman" <horms at verge.net.au> wrote:
>On Thu, Oct 21, 2010 at 03:50:17PM -0500, Anoop Bhat wrote:
>> Just a couple questions if someone has experience. We have two real
>> servers in an LVS-DR setup with arptables_jf load balancing ssh
>> connections. These connections are sometimes kept open for very long
>> periods of time and we're wondering how failover works with a setup like
>> 1. Does the secondary lvs server which just started ipvs and sent
>> gratuitous arp to take over the VIP send resets to those connections
>> which it doesn't know about or does it simply not respond because it
>> doesn't know the state of the connection?
>They will be reset when their next packet arrives
>unless you use the synchronisation daemon to mitigate this effect.
>> 2. Anyone have any experience with ssh load balancing and learned any
>> good lessons that we can take advantage of?
>My advice is to start simple and work from there.
More information about the lvs-users