|
Hi Cyrille,
> Do you think the ARP problem can be a consequence of all my
> problem ???
I don't think so. If it is the ARP problem what you will see is all servers
being directed to one real server.
> I didn't applied the hidden patch on my real servers... ?
What does redhat do with Pirahna setups? With LVS-DR you definitely need to
handle it somehow. I personally prefer the --REDIRECT (Horms) approach.
> I'm using direct routing, do you think NAT could be work better ?
I wouldn't switch things around again. DR is a good method. Some people
prefer NAT for security & simplicity issues. I don't think it will solve
your issue.
Did you see Malcolm's post about the http check timeout/interval? Turn
debugging on and check the logfiles to see if it is marking all of your real
servers out of service. Maybe this is occurring during heavy load? I'm no
pirahna expert, but pb is and wrote a HOWTO:
http://peterbaitz.com/pbs_howto.txt
Peter
|
