|
Malcolm Turnbull wanted us to know:
>>Coz we have 10 nodeserver behind them and near 500 sites thats the
>>reason why :)
>>You can tell me how to make it working?
>That doesn't sound like it would stretch a single LVS active-failover.
>Active/Active is only required if your hardware can't match the load and
>even then you'd be better off having two pairs of active-passive load
>balancers servering two different groups of VIPs.
>It also saves on complexity.
It takes 45 seconds for ldirector/heartbeat to failover from a locked up
master to the waiting slave for a 50 site system with 2 real servers
each. 500 sites at 10 real servers will be a lot longer time than that.
Horms mentioned that another (mon?) might be faster at failing over, but
once this system went live, it became harder to change from one core
program to another. Once we moved all site LB off the router onto this
system, it made it even harder to change things. Not impossible, just
harder :-)
--
Regards... Todd
They that can give up essential liberty to obtain a little temporary
safety deserve neither liberty nor safety. --Benjamin Franklin
Linux kernel 2.6.11-6mdksmp 3 users, load average: 0.68, 1.05, 1.15
|
