[PVE-User] Migration / Backup Question
david at digitaltransitions.ca
Thu Jan 26 14:51:26 CET 2012
I have a test I am attempting to deploy and am not sure exactly how to do this.
I have the following configuration that I am looking to get working
2 1U Intel Servers with onboard gig ethernet X 2
1 PCI-e Card in each server with 2 Port 10GB Ethernet (RJ-45)
1 NAS device with an added 2 Port 10GB Ethernet Card (RJ-45)
I have the following setup:
Both servers have proxmox 1.9 with latest updates
2 LUNS exported from the NAS to the servers with approx 1.0TB for testing over 10GB Ethernet direct connected to the servers
Both servers have had their /var/lib/vz remapped with the iscsi volumes based on this wiki article here:
By the way that article works great!!
What I would like to accomplish is have my master server (node1) communicate with my child server (node2) over 10GBE as opposed to the 1GB Public ethernet that the rest of the network uses.
This way, if I need to migrate a 40GB container (or any size), instead of it taking minutes / hours over 1GB, it only takes seconds to migrate from one server to the next over 10GB
The problem is this:
1.) I do not have a 10GB Ethernet switch, as I can't find one thats affordable for purchasing yet -- WARNING THIS CONFIG IS GHETTO!!!
2.) 1 Port of each servers 10GB Ethernet card is plugged into 1 port on the 10GB ethernet card on the NAS direct connected
3.) I have the second servers 10GB ethernet card plugged into each other direct connected
So the IP setup looks as follows in my little world here:
Public network (1GB): 172.16.10.0/24 (eth0)
Node 1 10GB Ethernet connection to NAS: 10.0.0.0/24 (eth2)
Node 2 10GB Ethernet connection to NAS: 10.0.1.0/24 (eth2)
Node 1 to Node 2 10GB direct connection to each other: 10.0.2.0/24 (eth3) node1: 10.0.2.10 node2: 10.0.2.11
I can ping all the IP's when I need to, so from node 1, I can ping 10.0.0.1 which is the IP address of the first port on the NAS direct connected to node 1. The same goes for node 2, and obviously neither node can ping the others private IP address range.
I can't figure out how to have the migration / backup communication occur over the 10.0.2.0/24 direct connect network. This would be a full 10GB's of data transfer from node1 to node2 direct connected to each other.
Whenever I add in the child server to the master over the 10.0.2.0/24 range, it still seems to add it over the public 172.16.10.0/24 range, which is not what I am looking to do.
Does this make sense to anyone? Is there a way to do this and I have just missed it?
Thank for any help you can provide me.
More information about the pve-user