Yes, there is a selection process for differentiating between vA and vB. I have a MySQL table that lists the hostnames that the virtual machines run on, the ip addresses for the virtual machines, and the "cluster id" for the virtual machines. So, if I specify cluster-id '0', I want to be able to make HTTP requests to vA. And, if I specify cluster-id '1', I want to be able to make HTTP requests to vB. I have a perl script that takes one parameter: a URL to GET from vA or vB. The URL looks like this: http://<vA or Vb's ipv4 address>/<cluster_id of vA or Vb>/some_webpage.html So, given that URL the perl script knows the ip-address of vA, it knows that vA is on sA, and it knows the ipv4 address of sA (it figures this out by using the cluster_id to do a lookup in MySQL). Now, I feel that if I have vA and vB on different VLAN's, the perl script should be able to use the information it has to route the HTTP request correctly. Once I figure that out, I can figure out how to do likewise for rsync and scp to vA and vB. My target application is load-balancing. I'm modifying APSIS Pound. I'm using virtual machines because I want to be able to transfer virtual machines from one physical server to another (using rsync) based on the load of a server. After I rsync a virtual machine to a different server, I update the virtual_machine_ip_address/clusterid--->hostname mapping in MySQL so that the perl script still works. Only problem is, I can't figure out how to use VLANS to deal with the ipv4 address overlap. Thanks again! On 10/8/06, Peter Stuge <stuge-vlan@xxxxxxx> wrote: > On Sat, Oct 07, 2006 at 07:45:53PM -0700, EKC wrote: > > I have two servers with different ipv4 addresses on an > > 802.1Q-capable ethernet switch. > > > > Each server is running a VMware virtual machine running an apache > > web server. However, both of the virtual machines have the same ip > > address. Changing the ip addresses is not an option. > > > > I want server A to be able to choose whether to make HTTP requests to > > the virtual machine on server B or the virtual machine on server A. > > Likewise, I want server B to be able to make HTTP requests to > > whichever virtual machine it chooses. And, I want both server A and > > server B to still be able to communicate with each other (ssh traffic, > > rsync trafffic, etc...) > > I'm calling your physical machines sA and sB (server A/B) and your > virtual machines vA and vB. > > sA and sB have different IP addresses so they can communicate with > eachother without any problems. > > You want vA and vB to both use the same IPv4 address but you still > want to be able to differentiate between them using only IPv4 > addressing, is that correct? > > Unfortunately it can not succeed, since there is no way of knowing > which of vA and vB is really refered to when the shared IP address is > used. > > This is not a straightforward setup. I assume you're doing some kind > of HA system. Perhaps you should have a look at ha-forum.org? > > > If my description above is incorrect and you in fact have some way > besides IPv4 to differentiate between vA and vB then there's > certainly a solution to be found, and VLANs are likely part of it. > > > > However, how do I get the physical servers to selectively route > > traffic to two identical ip addresses on different VLAN's? > > What is the selection process? > > > //Peter > _______________________________________________ > Vlan mailing list > Vlan@xxxxxxxxxxxxxxx > http://www.candelatech.com/mailman/listinfo/vlan > >
