|
> Andrew -- the reason I use DR for my apache cluster: since realservers
> have
> the VIP (hidden on the "lo" interface), all the realservers can use
> exactly
> the same config file (in my case, off an NFS volume, along with the
> content). The only trick is to have the log and pid files stored on a
> local
> filesystem (/usr/local/apache/logs symlinked to /var/run/httpd is
> where I
> stash it all) -- NAT would require that each realserver have a
> specialized
> config for its RIP address.
>
> Or did I miss something in the docs on nat vs tun vs dr ?? (horms ??)
I use DR for exactly the same reason (with tomcat servers, Apache servers, etc.
Now my cluster begin to grow fast), and you missed something, I have to say we
are both, who want to join the club ?
François.
P.S.: I was also aware about performance. I need to reduce all the latencies.
So, no big rewriting at the arrival, no rewriting at all at the departure,
should reduce the latency of some milliseconds...
|
