|
The issue:
Periodically (can't find a pattern) lvs01 and lvs02 (my identically
configured, software and hardware) load balancers bounce back and
forth. lvs01 will be active, then for some reason lvs02 detects it as
dead and takes over. then it stays on lvs02 until lvs01 detects it as
dead. I'm not sure if I just need to set my ping times higher or what
the issue is..
Background:
Recently I changed my LVS servers over from being CentOS w/ heartbeat
v1 + ldirectord I believe to Ubuntu w/ heartbeat v2 + ldirectord.
They seem to run great, very low resources of course. The machines are
overkill too - Dual-core Opteron 1.8ghz w/ 2 gigs RAM, 250 GIG SATA,
dual gigE onboard NICs.
They only run 3 services:
- postfix
- djbdns/tinydns
- LVS
Soon I will be removing postfix and moving it to it's own server. So
it will be PLENTY of resources.
With this new setup, I may have misunderstood a new configuration
parameter or something. I want lvs01 to be primary and lvs02 be
secondary. Right now it does not "prefer" lvs01 like I thought it
should according to the docs.
I have posted all the files (with my public IP addresses changed to
"my.main.ip" and "my.gateway.ip" and "LVSOTHERIP" and "LVS01IP" etc)
here:
http://mikehost.com/~mike/tmp/lvs/
Any help is appreciated. If you need any more information or for me to
try anything just let me know. I don't know if changing to the new CRM
style management would be a good thing but at the time it seemed
confusing and something I could not do on a production environment. If
there's a conversion tool or someone wants to do it I'd be willing to
pay for your time. I want this to be setup as optimal as possible!
ldirectord version 1.143
heartbeat 2.0.7
kernel 2.6.19.2
ubuntu 6.06 LTS (i think)
Thanks!
|
