|
Graeme Fowler wrote:
> On Thu, 2008-03-06 at 12:51 -0500, LDB wrote:
>> What HOWTO????
>
> *The* HOWTO. The one documenting just about all there is to LVS:
>
> http://www.austintek.com/LVS/LVS-HOWTO/
>
> It's been compiled over the years from many mailing list posts, pieces
> of documentation and other sundry sources. It's *very* detailed. I
> recommend you read at least some of it.
>
> However, I think you have fundamentally misunderstood one of the
> principal problems of LVS-DR. Back to first principles:
>
> For LVS-DR to work properly, the VIP must be present on both the
> director(s) and the realserver(s). In systems where the director's IP
> (the DIP) and the realservers' IPs (the RIPs) are on the same network -
> like yours - then the realservers can respond to ARP requests for the
> VIP, which results in the issue you see.
>
> Essentially, all the traffic ends up being handled directly by a (or
> more than one) realserver and the director doesn't see it.
>
> This means you have to "hide" the loopback interfaces on the
> realservers. See:
>
> http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.arp_problem.html
>
> for a much more detailed explanation of the problem, and the myriad ways
> to fix it. For a 2.6 kernel, use the arp_ignore/arp_announce sysctls.
>
> I really, really recommend you read the HOWTO. Alternatively, read
> Horms' excellent synopses and examples for Ultramonkey, the most
> relevant of which is probably:
>
> http://www.ultramonkey.org/3/topologies/hc-ha-lb-eg.html
>
> Graeme
>
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
>
>
Graeme .. Thank you much!!! :)
LDB
|
