Hello, I have several Windows Server 2008 SP2 servers, some of them are running in the NLB cluster. I enabled IPv6 on all servers using DisabledComponents=1 registry settings (which makes only native IPv6 protocol enabled, without Teredo or 6to4). Everything works fine on servers without NLB. Machines with NLB, however, have problems talking to other IPv6 hosts on the same /64 subnet. Communication from a non-NLB host originates in IPv6, traverses across the network to the IPv6 default gateway and then returns back to the destination NLB host in IPv4... Also communication between two NLB hosts goes through the IPv6 default gateway and eventually returns in IPv4. This behavior is the same when using DNS names and IPv6 addresses of destination servers. All servers are in a single /64 subnet. Router advertisement on the gateway is not enabled, all servers are statically configured and there is no DHCP on the network. Tracerts using IPv6 are 1 hop only each time, but somehow all other communication like file copy via SMB goes through the default gateway :( I'd like to stress that only communication involving NLB hosts is affected. In testing I found out that clearing the IPv6 neighbor cache solves the problem for a limited time. When the cache repopulates in a few minutes, exactly the same problem returns. I'd appreciate any help on this issue as this prevents me from continuing my IPv6 deployment on a larger scale. Thanks, Docentus

Hi Docentus, Thanks for posting here. >traverses across the network to the IPv6 default gateway and then returns back to the destination NLB host in IPv4... Also communication between two NLB hosts goes through the IPv6 default gateway and eventually returns in IPv4. This behavior is the same when using DNS names and IPv6 addresses of destination servers. Could you first show us the ipconfig /all results and route table form both nodes here when this issue occurred ? meanwhile, how did you verify that it was returned in IPv4? If you have sniffed any data could you also share with us ? Was there also any problem when configured this IPv6 NLB ? Network Load Balancing (NLB) can not connect with IPv6 addresss on Windows Server 2008 http://support.microsoft.com/kb/2000919 Regards, Tiger Li TechNet Subscriber Support in forum If you have any feedback on our support, please contact tnmff@microsoft.com. Please remember to click Mark as Answer on the post that helps you, and to click Unmark as Answer if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread.

Hi, Thank you for your answer. I uploaded my "ipconfig /all" and "route print" outputs in one config.txt file on my SkyDrive: https://skydrive.live.com/redir.aspx?cid=8bda2ed637ed7df5&resid=8BDA2ED637ED7DF5!140 I also included "netsh interface ipv6 show neighbors" output for both NLB hosts as I thought it may be useful. This data is gathered directly after copying a big file and noticing the traffic going through the router. A little bit of explanation to this config.txt file: I changed my external IP addresses. I changed first two IPv4 octets to 123.123 and first three groups of IPv6 to 2001:aaa:bbbb. My default gateway is 2001:aaa:bbbb::a. Address from range 2001:aaa:bbbb::c:0/112 are dedicated, address 2001:aaa:bbbb::cc:1 is shared NLB address. NLB is configured using IGMP multicast. Answering your questions: 1) I have MRTG on my default gateway installed. It monitors IPv4 and IPv6 traffic separately and I can see that on the IPv6 graph there is only incoming traffic on Internal interface and the same amount of traffic egress through the same Internal interface but on the IPv4 graph. The Internal interface on my default gateway is dual-stacked. 2) I've already seen this KB but I think it's not applicable here - it's about connecting to the NLB Manager using IPv6 addresses. I have no problems connecting to the NLB Manager using DNS names of NLB hosts so I assume NLB Manager chooses IPv4 addresses automatically. I'm looking forward to your further questions and suggestions. Regards, Docent

Do the AAAA records exist in DNS for the NLB hosts, NLB VIP and what is the addressing on the IPv4 on the network, public or private. Are the hosts added to the NLB manager with both the IPv4 and IPv6 addresses? Sumesh P - Microsoft Online Community Support

AAAA records are automatically registered in DNS by all NLB hosts and the records are present in the zone. VIP does not have AAAA record but I don't use VIP in e.g. copy process when the problem occurs. As I said, the problem occurs also when using pure IPv6 address in the UNC path (and no DNS names). I have public IPv4 addressing on the network. All NLB hosts are configured with dedicated IPv4 and IPv6 addresses using NLB Manager. Cluster is configured with a few IPv4 addresses and one IPv6 address. These are all public (global) addresses. I would like to emphasize that: - the problem occurs with or without IPv6 default gateway configured on the NLB hosts - clearing the neighbor cache solves the issue for some time (until neighbor cache repopulates) If the gateway address is irrelevant here, I think the problem is wrong interpretation of neighbor advertisements that are exchanged between NLB servers and/or router. Also I'm wondering why you cannot specify subnet length when adding dedicated IPv6 address in NLB Manager. You can specify subnet in dedicated IPv4 addresses but this is not the case in IPv6. Somehow, however, the subnet length is guessed correctly by NLB Manager and configured properly when I open the TCP/IPv6 Properties window. Regards, Docentus

Can you try reducing the basereachable time per KB 949589. Let me know if that helps. Sumesh P - Microsoft Online Community Support

I reduced it to 10000 on both NLB hosts but still file copy between them sometimes goes through the default gateway (and sometimes not) in a manner that I described earlier... Regards, Docent

Can you try the command to set the neighborcachelimit as well. It has a default of 256 and when it is purged, the machine works. Change it for IPv4 and also run it for IPv6. See if that helps. Sumesh P - Microsoft Online Community Support

Hi, Thank you for another hints. I set the global neighborcachelimit both for IPv4 and IPv6 to 4096 as in the KB949589 and then purged the cache. Unfortunately it didn't help :( In a few hours again the traffic is coming through the gateway :( Regards, Docent

No, 4096 is too high, I would cut the 256to a very small TTL if not zero then start upping the value, 64, 128 possibly. Try that please. Sumesh P - Microsoft Online Community Support

Thanks for clarification. I set the neighborcachelimit to 0 but it didn't help either. Regards, Docent

At this point I think that this issue requires a more in-depth level of support to get to the root cause. Please visit the below link to see the various paid support options that are available to better meet your needs. http://support.microsoft.com/default.aspx?id=fh;en-us;offerprophone Sumesh P - Microsoft Online Community Support