DNS Configuration when using Site Recovery Manager

K

KapsZ28

2[H]4U
Joined
May 29, 2009
Messages
2,114
I am setting up a DR site with Site Recovery Manager. All VM's are Windows Server 2008 R2. My main concern is with DNS. The DR site will obviously be using a different subnet. When I setup the IP addresses on the replicated VM's, I don't want them getting replicated back to DNS in the primary site. And of course the current production IP addresses listed in DNS will not be correct in the DR site.

So how do you handle the different IP addresses and DNS when doing DR?
 
When you use the 'customize the IP settings during recovery' it doesn't actually add the IP addresses to DNS at that point. On recovery, DNS information will change on the VM and update the local DNS server with the new info. If you want to test a recovery plan, you're going to need to sever connectivity from the production site, or you're going to have to just do test recoveries using SRM's capabilities using a segregated network.
 
QHalo said:
When you use the 'customize the IP settings during recovery' it doesn't actually add the IP addresses to DNS at that point. On recovery, DNS information will change on the VM and update the local DNS server with the new info. If you want to test a recovery plan, you're going to need to sever connectivity from the production site, or you're going to have to just do test recoveries using SRM's capabilities using a segregated network.
Click to expand...

Cool, that is what I was thinking but wanted to confirm.
 
Also, make sure that your VM's have dynamic entries in DNS, not static.
 
So I do have an issue which I figured would happen. The two sites are connected with a layer 2 connection. My DR site is using 10.100.0.0/24 and production is 192.168.10.0/24. The only way for production to communicate with DR is by adding an additional NIC in production with a 10.100.0.x IP address. No default gateway. This seems to be an issue with the Site Recovery Manager. When I go to configure a connection between the two sites, it fails to connect. Most likely because the two subnets cannot communicate with each other. I am used to configuring two sites with a site to site VPN connection and this wouldn't be an issue. Since this is more like a LAN configuration, how do I get the SRM's to connect to each other?
 
Why can't you just route between the two sites? They need to talk. Should have a management network on each side that is routable to the other side.
 
NetJunkie said:
Why can't you just route between the two sites? They need to talk. Should have a management network on each side that is routable to the other side.
Click to expand...

Not really sure. I am not a network person. I asked if it could be setup that way and the person that does our networking didn't think so. This is a 1 GB connection between CA and VA. Basically like a long ethernet cable connecting LAN to LAN. I don't think there is any routing involved other than a tagged VLAN on each side.

What would be the proper way to setup a site to site connection like this without impacting the performance of the 1 GB link?
 
So if it's just a layer 2 connection why don't you extend the VLANs to both sites and skip the re-IP addressing part?

All you need is something routing traffic between the 192 and the 10 VLANs. It's not magic. Usually a layer 3 switch does that. How are you replicating your array traffic? Are the arrays on the same VLAN stretched across sites?
 
NetJunkie said:
So if it's just a layer 2 connection why don't you extend the VLANs to both sites and skip the re-IP addressing part?

All you need is something routing traffic between the 192 and the 10 VLANs. It's not magic. Usually a layer 3 switch does that. How are you replicating your array traffic? Are the arrays on the same VLAN stretched across sites?
Click to expand...

If we extended the VLANs, what happens to the default gateway? Right now both sides have their own firewall connected to the Internet. If I used 192 on both sides and the network goes does in the primary site, I lose my gateway. I guess I could in the event of a DR reconfigure the firewall on the same subnet as the primary network.

They should be layer 3 switches, but I have to find out.

Arrays won't be replicated. LA uses a NetApp and VA, (DR site,) uses OpenFiler. Also all the VM's resides on local storage on the ESXi servers. This client is cheap, so it is not exactly an ideal setup. I will be replicating the VM's from local storage to local storage. Data that is on the NetApp will be backed up to VA using our backup solution and then restored in a DR.

Right now the replication appliance probably won't even work because of the separated networks. There is just no communication between 192 and 10.
 
You need 192 and 10 to talk to each other. Bottom line. If you explain that to your network guys and they don't understand that, run.
 
KapsZ28 said:
If we extended the VLANs, what happens to the default gateway? Right now both sides have their own firewall connected to the Internet. If I used 192 on both sides and the network goes does in the primary site, I lose my gateway. I guess I could in the event of a DR reconfigure the firewall on the same subnet as the primary network.

They should be layer 3 switches, but I have to find out.

Arrays won't be replicated. LA uses a NetApp and VA, (DR site,) uses OpenFiler. Also all the VM's resides on local storage on the ESXi servers. This client is cheap, so it is not exactly an ideal setup. I will be replicating the VM's from local storage to local storage. Data that is on the NetApp will be backed up to VA using our backup solution and then restored in a DR.

Right now the replication appliance probably won't even work because of the separated networks. There is just no communication between 192 and 10.
Click to expand...

You use something like HSRP so each side has a local gateway then do filtering to make sure you don't send traffic back and forth across that doesn't need it. This is where DR planning gets more involved. It's easy to replicate the VM data and fail over. Getting the network right is tougher.

I assume you're using Host Based Replication, then?
 
I am not replication the entire host. Only replication certain VMs through Site Recovery Manager. The datastores will be mapped, and VMs replicated. That is, once the networking is corrected.

I can probably force everything to use the 10.100.0.0 network in both sites, but I would rather they use their own subnets.
 
KapsZ28 said:
I am not replication the entire host. Only replication certain VMs through Site Recovery Manager. The datastores will be mapped, and VMs replicated. That is, once the networking is corrected.

I can probably force everything to use the 10.100.0.0 network in both sites, but I would rather they use their own subnets.
Click to expand...

SRM doesn't replicate data. It automates failover and testing. Either your arrays replicate the data or you used Host-Based Replication in vSphere.
 
NetJunkie said:
SRM doesn't replicate data. It automates failover and testing. Either your arrays replicate the data or you used Host-Based Replication in vSphere.
Click to expand...

Correct that SRM doesn't actually replicate. I installed the vSphere Replication Appliance which integrates with SRM. It is still not host-based. It is per VM. Replication is not an issue and I have done this before.

What I haven't done is set this up with two sites connected via a layer 2 MPLS. I am trying to understand what is the correct setup.

Like QHalo said, the two networks should be able to talk to one another. Unfortunately my network guy seems to disagree and I am stuck adding a second NIC to my production vCenter and forcing replication, vCenter, and SRM to communicate over the 10.100.0.0/24 network.
 
vSphere Replication is called Host-Based Replication because..well...it's the hosts doing the replication. Doesn't mean you are replicating the entire host.

Your network guy is confused and doesn't understand the question. What you're asking is very simple. Just tell him you need routing between the two networks.
 
NetJunkie said:
vSphere Replication is called Host-Based Replication because..well...it's the hosts doing the replication. Doesn't mean you are replicating the entire host.

Your network guy is confused and doesn't understand the question. What you're asking is very simple. Just tell him you need routing between the two networks.
Click to expand...

Understood. Host-Based Replication.

He understands, just doesn't agree or can't make it happen.

Here is the other problem I have with it. Obviously I need a domain controller setup in our DR location. The only way I could do it with the current setup was to add an additional NICs to the domain controllers in the primary site and add an IP from the 10.100.0.0/24 subnet. Setup Site and Services, and install the new domain controller on the same subnet. So now all three domain controllers are communicating over 10.100.0.0/24. It works, but I don't feel like this is the correct configuration.

I just wish I had more information to give him in order to correctly setup routing over the MPLS. He has never worked in a large enterprise where this is normally setup. I come from large enterprises, but have never been involved in networking.
 
It's time to bring in a partner/consultant. Someone needs to bridge this gap. Sounds like there wasn't any real talk of design before this started.
 
NetJunkie said:
It's time to bring in a partner/consultant. Someone needs to bridge this gap. Sounds like there wasn't any real talk of design before this started.
Click to expand...


Haha. That is an understatement. They started planning this before I started here and everything planned so far was wrong. The plan now is good, I just don't have the network setup the way I want it.
 
Yeah, I talked to another network engineer that said all his MPLS connections are routed. Unfortunately my guy just doesn't know how to do it I guess.
 
The MPLS connection is going directly into a switch on each end, and not through a router. I believe these switches are capable of layer 3, but not currently setup that way. If they were configured as layer 3, would we be able to static route the subnets?
 
Somewhere...somewhere in your network is a router. It doesn't have to be the switch that MPLS plugs in to. Just somewhere you need a router that can route between those subnets. You really want one on each side.

Or just trunk the damn server VLAN across the MPLS and have the same IPs on both sides.
 
Time to talk to the boss. Project stalls that no one can resolve, time for a 'come to Jesus'.
 
QHalo said:
Time to talk to the boss. Project stalls that no one can resolve, time for a 'come to Jesus'.
Click to expand...

Ha, he is the boss. It is a small business. I think the issue is this is a dedicated connection between two datacenters and not specific for the client I am setting up DR for. I am giving up trying to tell him how the network should be. So much stuff here is done incorrectly.
 
Finally got the CTO to agree to setting up static routes through the firewall. Now I have to make the changes myself....
 
Took almost an entire week, but at least the sites are now setup correctly and I can finish my SRM configuration.
 
You must log in or register to reply here.
Back
Top