Discussion Locked

Caching and DNS Round Robin

0 helpful
5 replies
Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.

The config:

I have 1 web server that contacts 3 COM+ application servers. I am trying to load balance the connections to the COM+ servers using DNS round robin. I have entered the following entries in my DNS:

comserver IN A 10.94.11.201 with 1 sec TTL
comserver IN A 10.94.11.202 with 1 sec TTL
comserver IN A 10.94.11.203 with 1 sec TTL

Round robin is enabled on the Windows 2000 DNS server. The Windows 2000 web server that contacts the WIndows 2000 comservers is on the same logical subnet (10.94.11.0).

I have disabled local cache on the Web server using the following registry key:
LCL_MCHN \ SYSTEM \ CurrentControlSet \ Services \ Dnscache \ Parameters \ MaxCacheEntryTtlLimit = 1

The problem:
DNS round robin works fine if I go around my network and ping 'comserver'. However, from the web server if I ping 'comserver' the round robin DNS sends back the IP address of the next comserver listed for round robin but then the web server is stuck with this address. Even though the web server's cache is set to 1 sec therefore it queries the DNS server for each request AND ipconfig /displaydns shows only localhost, it always gets the same IP address.

Any idea?

Patrick

5 replies

by beermonster

Hi,
Don't know if this is relevant but.... Win2k (and the latest NT)dns uses something called 'LocalNetPriority'. This basically overrides the round robin behaviour to ensure that a client is always directed to it's nearest server. basically, dns looks at the address the request came from, then - if the name the client want's to resolve is round robin'd - checks to see if one of those addresses is on the same subnet as the client. If so, it assumes that you would want the client to use the closest server and returns that address, even if it was not the next in the list. Now, this might not be in action here as all your com servers are on the same subnet. Either way though it's worth looking into. There is a technet article covering this - and telling you how to disable LocalNetPriority. go to

microsoft.com/technet

and search for Q177883

Hope this helps...

by patrick.bisaillon

Hello BeerMonster

I saw this MS Q article but it did not help me because the web server is on the same logical subnet as the comservers.

Thanks for your help.

by ispnetworker

Sounds like the servers are defined as 'sticky' (with persistence) by your load balancer. Making a server sticky will cause all future connections to the VIP from a particular client associate with a specific RIP (real server).

by patrick.bisaillon

Hi ISPNetworker. Sorry for the delay but your suggestion did not help me. I got the fix from Microsoft:

1. DNS Server Properties > Advanced > uncheck Enable Netmask ordering
2. Stop the DNS Client Service on the client workstation

Thanks for your help

This discussion is closed to new replies.