|
I just wanted to give a bit of performance feedback:
Our two node load balance cluster (two PIII 700 machines) automatically
fails over the following resources in an LVS-DR configuration:
55 VIPs (IPAddr)
93 virtual services, mostly http and https (ldirectord)
A failover (a node becoming active) lasted from 22:03:08 to 22:03:50, so
it took a total of 42 seconds to bring up the VIPs and virtual services.
>From 22:03:08 to 22:03:38 (30 seconds) was heartbeat bringing up VIPs
and from 22:03:38 to 22:03:50 (12 seconds) was ldirectord bringing up
the virtual services.
A graceful shutdown of an active node lasted from 21:46:30 to 21:46:44, so
it took a total of 14 seconds.
Starting a node when the other node was active (auto_failback off) took
only 1 second.
Traffic seems to be peaking somewhere around 300 KBytes (but we're not
at the heaviest traffic point of the day, so we'll see if it averages
higher). The traffic across two of the main machines being load
balanced averages around 10-15 Mbits each (according to iptraf).
Load average on the active node is 0.00 except when I login :-)
I just wanted to give a big virtual beer to all LVS and Linux-HA
developers and HOWTO makers and in general question answerers. We took
95% of the services that were being load balanced on a Cisco with SLB
running at 80-90%+ CPU, and it is now running at 10% CPU. It is much
happier, and we are much happier. :-)
--
Regards... Todd
OS X: We've been fighting the "It's a mac" syndrome with upper management
for years now. Lately we've taken to just referring to new mac
installations as "Unix" installations when presenting proposals and
updates. For some reason, they have no problem with that. -- /.
Linux kernel 2.6.8.1-12mdkenterprise 1 user, load average: 0.19, 0.10, 0.02
|
