Solved

How to set up name resolution on redundant networks

Posted on 2009-05-13
17
729 Views
Last Modified: 2013-11-25
System configuration:

Some PC's (Servers/Clients), running Win2003 Server, two of them set up as WINS Servers. Applications not of interest for problem. Every station is equipped with two network interfaces, configured to separate IP network segments, thus resulting in two physically and logically independent, fully redundant networks.

Problem: The WINS servers seem to return only one of the possible IP addresses if asked for the address of a given system. Which address is the WINS server returns is IMHO dependent on random issues, possibly which address the server has seen last from that station. If one card fails in a system (or a cable/router/switch/NIC gets defective etc.) a random number of stations is not reachable by name (but surely by the remaning of it's IP's directly), because the station is provided with the addresses in the defective network.

I need a way to teach Windows to try to resolve a name with any IPs known ?
0
Comment
Question by:frankhelk
[X]
Welcome to Experts Exchange

Add your voice to the tech community where 5M+ people just like you are talking about what matters.

  • Help others & share knowledge
  • Earn cash & points
  • Learn & ask questions
  • 5
  • 5
  • 4
  • +1
17 Comments
 
LVL 10

Expert Comment

by:lanboyo
ID: 24374679
Configure all server IP addresses as replication partners with other WINS servers.
Configure Both addresses in the interface wins server configurations and no other servers.
0
 
LVL 14

Author Comment

by:frankhelk
ID: 24375388
@lanboyo:

I'm not sure how you meant that - so I'll try to beef up my explanation of the problem a little bit ....

The current WINS server configuration is that the two WINS servers are replicating push/pull to each other. On every network interface on every computer I have configured to ask both WINS servers with their respective IP addresses.

Example:

WINS server 1:
    IP address 1 = 192.168.110.5/24
    IP address 2 = 192.168.120.5/24

WINS server 1:
    IP address 1 = 192.168.110.6/24
    IP address 2 = 192.168.120.6/24

Sample other PC:
   IP address 1 = 192.168.110.10/24, here WINS = 192.168.110.5, 192.168.110.6
   IP address 2 = 192.168.120.10/24, here WINS = 192.168.120.5, 192.168.120.6

If I pull the plug from NIC 1 on the sample PC, and try to ping it from another Station (even from the WINS server) with

PING SAMPLE

the ping might succeed - or not, depending on which address the currently asked WINS server gives back. Trying

PING 192.168.120.10

will work in any case.
0
 
LVL 51

Expert Comment

by:Keith Alabaster
ID: 24465710
Firstly, are all systems using the same default gateways? You cannot have a default gateway on both nics so requests should always take the default path. As you have all machines having a second nic that is on the same subnet as the secondary wins servers then they should be directly contactable outside of the default route without any static routes - as they are directly attached.

Have you cionfirmed that all devices have the nic with the default gateway in the same bind order?
0
Get MySQL database support online, now!

At Percona’s web store you can order your MySQL database support needs in minutes. No hassles, no fuss, just pick and click. Pay online with a credit card.

 
LVL 14

Author Comment

by:frankhelk
ID: 24466226
Keith,

since the system is designed as standalone network, no standard gateway is needed and none is defined at all.

All needed systems could be reached by either direct connection (only one or two switches in the middle) or via routing servers defined by static routes on any node (in fact there are 3 separate networks, one is "central" and the remaining two are attached to servers with nic's in both nets. But that's secondary, because the problem exists in the "central" network, too.)

I'm not sure what's meant with "bind order", but I suppose that is obsolete in absence of a standard gateway.

Just to clarify the problem again - both WINS servers are connected to both networks, but the system asking seems to get only one IP as answer from the WINS server. That works proper if all connctions are up. If the connection related to that IP is broken, no attempt seems to be taken on the other one (which is still alive), thus making the redundancy concept senseless.


0
 
LVL 57

Expert Comment

by:giltjr
ID: 24466309
Unless MS has changed something WINS does not have the same capability as DNS when you have multiple computer names with the same IP address.

Bind order has nothing to do with standard gateways, it deal with the order of the NIC's.  Windows has the NIC in a "list" it the registry.  For certain function, DNS and WINS being two of them, it will send out the name resolution request over the 1st NIC in the list.  If it gets NO response, it then goes to the second NIC in the list.  By NO response, I mean NO response, not a "I don't know" response.

Why are you not using DNS?  I would suggest that you use DNS, the only thing I am not sure of, is if dynamic DNS will work with the same host registering its name and IP address via two NIC's.    You might have to have static DNS entries in order for the one name, multiple IP address function to work.
0
 
LVL 14

Author Comment

by:frankhelk
ID: 24467067
giltjr,

the problem seems to be that, when Windows asks on the first NIC, it might get an answer (not the connection to the WINS server is broken), but the destination may be unreachable by that IP due to network malfunction. But instead of trying the next NIC, it seems to simply abort the request.
0
 
LVL 51

Expert Comment

by:Keith Alabaster
ID: 24467120
Bind order does not affect gateways but does affect the first route looked at - it will be on the first nic listed.
Thanks for the additional info that these are non-DG networks. You say though that you are using static routes? Are you giving different prioirties to these are all they all the same?



0
 
LVL 57

Expert Comment

by:giltjr
ID: 24467134
Sorry, but it is working as designed.  It did a name look-up, it got an answer, it is going to try and talk to the address that was returned.  If that address does not respond, there is no "try again" or "look for another host by that name" function/option.

In fact that is they way that DNS works also.

If I understand what you are attempting to accomplish you need to take another approach.  The approach you need to take is to use NIC teaming.  One host name, one IP address, two NICs, two paths into the "same logical" network.  The network must be on the same VLAN and IP subnet and there must be a common connection point some where in the network (generally the "center").  Something like:

                  /-> SWA1 <--> SWA2 <--> SWA3 <-\
                 /                           /\                          \
Desktop <-                             |                            -> Server
                \                            \/                         /  
                 \-> SWB1 <--> SWB2 <--> SWB3 <-/


This way there is no single point of failure.  With all switches in the same VLAN and IP subnet, if any single switch fails, there is a backup path.  If fact, if the "A" and "B" switches are close enough you could interconnect all "A" and "B" switches.
0
 
LVL 51

Expert Comment

by:Keith Alabaster
ID: 24467175
But if the server that should be providing the answer is down ie the wins server - then the alternative wins server should be checked by the requestor and the alternative ip given.
0
 
LVL 57

Accepted Solution

by:
giltjr earned 250 total points
ID: 24468279
Yes, but say comp102  has IP address 1.102 (NIC1) and 2.102 (NIC2), WINS1 knows about 1.102 and WINS2 knows about 2.102.

I'm on comp101 and I have addresses 1.101 and 2.101, with NIC1 having 1.101 and it is in the bind list first.  So I will send out WINS lookup for comp102 to WINS1 first and if it is up it will return 1.102.  Now what happens if comp102's NIC1 is down?  I can't talk to comp102.

With NIC teaming there is only a single IP address with two paths.  As for my first network diagram, in order to have full redundency you really need to do either:

                  /-> SWA1 <--> SWA2 <--> SWA3 <-\
                 /         /\               /\               /\         \
Desktop <-            |                |                |          -> Server
                \          \/               \/               \/        /  
                 \-> SWB1 <--> SWB2 <--> SWB3 <-/

or:

Have each of the "1" switches connected to both of the "2" switches and have each of the "3" switches connected to both of the "2" switches.  Basically the "2" switches will be your "core" switches and the "1" and "3" will be edge switches.

This does depended on the size of your network.  The more computers you have, the more edge switches you need.
0
 
LVL 51

Assisted Solution

by:Keith Alabaster
Keith Alabaster earned 250 total points
ID: 24470391
I think we are saying the same thing - as even with that redundancy - wins will only return one entry as a response to the request.

0
 
LVL 14

Author Closing Comment

by:frankhelk
ID: 31585193
Thanks ... even for letting me know that the current system layout won't let it work as initially expected. Unfortunately some application software relies on the "tow separate networks" stuff and is fully capable of using it, and the Windows part of the thing (file handling, remote console, etc.) is secondary.

Thanks anyhow.
0
 
LVL 57

Expert Comment

by:giltjr
ID: 24472292
Thanks for the points, but could you answer a few questions so that I could learn.

What software do you have that relies on two separate networks?

How does it handle name resolution?  Even with IP DNS you would have the same basic issue that you have with WINS.  Although IP DNS can return multiple IP addresses for a single host name, every client software package I have seen only tries the 1st IP address in the list, I have not run into any that will actually try more than one.

In fact the only name resolution process that I can think of that would work is a broadcast type name resolution, like NBNS broadcasts.
0
 
LVL 14

Author Comment

by:frankhelk
ID: 24474496
giltjr,

I've got answers, so you've earned your points. Thanks.

The software is a process monitoring system. It is configured fixed to the IP addresses of the partner systems, and where needed it deals with names in its own protocols. Most communication is done by multicast, and the software is capable of monitoring both NICs for the packets. It won't be interrupted by if one connection is lost. It doesn't need name resolution, and bypasses it entirely.

The name resolution issue arised from the needs of the imperfect human beings, who prefer to use machine names instead of IPs ...
0
 
LVL 57

Expert Comment

by:giltjr
ID: 24474572
multicast is different than unicast.  In fact multicast is a modified broadcast, which is why it works, it is not bound to a machine specific IP address or to a specific NIC.  It will listen to a multicast IP address on all NICs.

Switching to a single IP address in a dual NIC enviroment will not break this setup.  In fact it may make it easier, as you will now only need to configure the monitoring side with one IP address instead of two.

Glad to have been some help.
0

Featured Post

Simplifying Server Workload Migrations

This use case outlines the migration challenges that organizations face and how the Acronis AnyData Engine supports physical-to-physical (P2P), physical-to-virtual (P2V), virtual to physical (V2P), and cross-virtual (V2V) migration scenarios to address these challenges.

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Configuring Remote Assistance for use with SCCM
Sometimes clients can lose connectivity with the Lotus Notes Domino Server, but there's not always an obvious answer as to why it happens.   Read this article to follow one of the first experiences I had with Lotus Notes on a client's machine, my…
Finding and deleting duplicate (picture) files can be a time consuming task. My wife and I, our three kids and their families all share one dilemma: Managing our pictures. Between desktops, laptops, phones, tablets, and cameras; over the last decade…
In this video we outline the Physical Segments view of NetCrunch network monitor. By following this brief how-to video, you will be able to learn how NetCrunch visualizes your network, how granular is the information collected, as well as where to f…
Suggested Courses

615 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question