Re: [Keepalived-devel] Re: Question concerning virtual mac address

[Christian S. <Chr...@gm...>]

Sebastien Renard wrote:
> Quick & quite ugly way to do it : 
> - use separate script to monitor your apache process (from the simple ps | 
> grep to a more complex http GET with parsing http return code or even http 
> body). 
> - one script on each node, each monitoring it own local apache
> - in case of apache failure, the script send a SIGTERM to keepalived or a 
> rc.keepalived stop.
> - try to restart apache
> - in case of apache back to life, start again keepalived.

currently i am doing exactly that with vrrpd. but as keepalived already 
has support for all kinds of "alive" checks it would have been nice to 
use ONE solution instead of a "bricolage".

besides that you then also have to have another script which checks if 
your checker-script (check if apache is alive) is running, ... basically 
a very bad situation. but if it would be integrated into keepalived then 
if keepalived would crash nothing would be lost, because the backup-vrrp 
daemon would just take over the virtual ip.

> The latency of the failover will be quite bad compared to a standard LVS 
> architecture, but you can use service on the same box than the director. 

actually no, why? because if you plug off the cable then anyway the 
original master vrrp daemon is not visible outside anymore and therefore 
the situation is equivalent to not having such a script and if the cable 
is plugged back again then the backup-server will just keep the virtual 
ip a bit longer until the original master vrrp daemon has started up again.
-- 
Christian Schuhegger
http://www.el-chef.de/