Alternate IP Address causing Domain connections to drop

zzz2496 said:

RedBirdDad said:

When you added the alternate addresses, did you put in a default gateway address as well? I've used the alternate address stuff as well. Going back to W2K like you have.

While I haven't tried it on W7 yet, one thing that used to get me into trouble is having more than one default gateway. Adding more than one will get windows all confused. As long as you have *one* default gateway (in this case your domain) it should know which is which.

Just my 2 cents.

From what Jeffs said, I think he doesn't use 2 gateways, he just added a new alternate IP address. Having 2 gateways does indeed make Windows go weird at times, but it can be cured by putting a number in the "Interface metric" text box in the advanced properties. It simply saying that this gateway is farther than that gateway, so Windows will automatically use the shorter distance gateway. By the way, I'm currently using 3 different IP addresses for my current computer, one for IBM IMM interface, one for my VMM connecting to my IBM server, and the last one is for connecting to the virtual servers inside the IBM, nothing wrong so far (I only access FQDNs/IP addresses here).

zzz2496

To build on what zzz2496 said, I'm pretty sure this is the correct assessment. It just looks like the Browser service is getting confused.

An alternate solution is to physically segment your 192.168.0.x network from the 192.168.10.x. This would require that you get a separate Wired Ethernet card.

Nice, Jeffs :)

Anyway, if you still want to use the hostname, you just add:

Code:

192.168.10.200      Serv01

in your HOSTS file, it should then resolve the "Serv01" by HOSTS file instead of broadcasting over the network.

zzz2496

I mentioned this to the IT guy this morning, before you did. Neither one of us was sure if that would work. I'll take your word for it that it does.

For now I'm going to leave the hard coded IP addresses in. Any additional network load from what I'm doing won't matter anyway.

I'm going to add in the two additional class C subnets now at 25.x and 26.x

Well, the HOSTS will work (my Microsoft training curriculum said so, and it does work like that). The order of which Windows will use to resolve an IP address are: at the very first is HOSTS/LMHOSTS file, then when it failed, it will query the Master Browser in the network, that's where the "Computer Browser" service comes in. If the "Computer Browser" service failed, it will then query the first prefered DNS server it has, and when failed, it'll query the next prefered DNS server, and so on...

Jeffs, I just remembered something... as I understand it, you need more IP address, so you need to expand your address space. Now, if you read my guide about internetworking, I didn't have the time to talk about CIDR, Classless Inter Domain Routing.

If you've read my guide, the next topic after this won't be that confusing, if you haven't read the guide, I suggest you read it first, then continue read this post after you read the guide.

As I talked about in my guide, IP address is defined by 2 elements, the IP address number and the subnet mask. These 2 numbers do have a regulation body regulating it, BUT for LAN that sits behind a NAT/PAT router, you can comply to that regulation loosely. What I mean is, Class C private address is defined by (I just pick a common one from the Class C range) 192.x.y.z with subnet mask of 24bits (255.255.255.0), we all know that, BUT this is a closed network, you are using Class C private addressing because it's, well... private, Local Area Network... So, assuming this... you can freely define your subnet mask to accomodate your network size growth needs. Here's an example: for now you're using a 24bit subnet mask, that will result your network to have maximum of 254 hosts, the .0 is for network address, the .255 is the broadcast address. Here's where CIDR comes in. Classless means, it ignores the "regulation" of Class C private address, you can have, say... 20bit subnet mask, of even extreme (to simplify calculation) use a 16bit subnet mask (which according to regulation, it's a Class B network), but hey... it's private, you don't route your host directly to the internet, so it's OK. By reducing your subnet mask, you are extending the number of hosts you can have in your network.

Here's a test case:
Your current subnet mask is 24bit, that will give you 254 hosts available for your network. Say you need around 300 hosts in the net and you don't have the network infrastructure or human resources to maintain a routed network, you can just simply remove 1 bit from your network subnet mask. That means changing EVERY COMPUTER'S SUBNET MASK on the network. Say you're reducing it by one, so it's now 23bit (255.255.254.0), you are now able to stuff 512 hosts in your network, nothing is changed drastically, just the subnet mask. Say in the future you need to expand on the network some more, and you still don't have the infrastructure nor the resources, reduce another bit from your subnet mask, you'll double the number (at 22bit [255.255.252.0], you can have 1024 hosts in the network).

By using CIDR type address, you don't need to "define" another Class C network to use, and all will be simpler. You can test these type of addressing in a closed lab if you want, I have a network like this running in a production system, and it's routed through several network backbones. It works for those who are economically limited :)

Anyway, that was just an idea, feel free to test it :)

zzz2496

zzz2496 said:

Nice, Jeffs :)

Anyway, if you still want to use the hostname, you just add:

Code:

192.168.10.200      Serv01

in your HOSTS file, it should then resolve the "Serv01" by HOSTS file instead of broadcasting over the network.

zzz2496

This would work also.

@zzz

Thanks for the tutorial, but I'm up there with you already. Both the guy in the IT dept, and myself realize the subnet mask needs to be opened up to allow more hosts on the network. However, he doesn't want to do it out of fear of breaking something. Also, while we're not a huge operation, there are several hundred devices in two buildings that would need to be modified, all at once, for this to work.

At some point in the future he may be (actually will be) forced to do this when he runs out of addresses on the single 253 address segment, but that will be when machines that are part of the "business" can't be used on the network and there is no other choice.

Right now, I am using "non business related" Ethernet devices on the network that only I need to access, so the logicial workaround is for me to assign alternate IP addresses to my machine (and a few other machines) to access these devices rather than modify the entire network addressing scheme.

Ahh, I see :)

zzz2496