Issues while pinging IPv6 address

[Chauhan Amit]

Hi All,

I am facing a very wierd issue working with IPv6 address on HP-UX system.
Scenario is like , I want to enable communication between two systems via IPv6. I have Configured my lan1 and lan2 on both the systems (/etc/rc.config.d/netconf-ipv6) for IPv6 address(Link-Local).Lan1 is connected to lan1 and lan2 is connected to lan2 of the other systesm and that has been verified via link-loop command.
But the strange thing is I am able to ping only 1 IPv6 address(lan1) not the other one(lan2) from other system. Trend is like whatever IPv6 interface comes UP later only that is pingable. Tried configuring IPv4 address , no such problem was seen and even I have tried the same on different set of machines as well.
Let me know if the problem is not clear. Has anyone anytime faced susch issue ? Do I need to do any extra route settings for IPv6.

Need Experts help/advice on the same.

Thanks,
Amit

[Robert-Jan Goossens]

Hi Amit,

Could you post the /etc/nsswitch.conf?

http://www12.itrc.hp.com/service/cki/docDisplay.do?docLocale=en&docId=emr_na-c00926204-1

HP-UX: resolver ignores lookup policy in /etc/nsswitch.conf

Regards,
Robert-Jan

[Chauhan Amit]

Here you go:

#
# /etc/nsswitch.files:
#
# @(#)B11.23_LR
#
# An example file that could be copied over to /etc/nsswitch.conf; it
# does not use any name services.
#
passwd: files
group: files
hosts: files dns
ipnodes: files
services: files
networks: files
protocols: files
rpc: files
publickey: files
netgroup: files
automount: files
aliases: files

Thanks,
Amit

[rick jones]

In very broad handwaving terms two interfaces with IPv6 link-local addresses is like having two IPv4's in the same subnet. The routing code will have "problems" deciding which interface to use, and as you can see, which one it uses will determine which remote IPv6 addr you can ping. I'm guessing that when you used IPv4 addresses you happened to put them in separate subnets?

In linux-land, one passes a "scope id" with the addresses to help it pick the interface when playing with v6 addresses.

I don't know if UX uses something similar. You could try some explicit host routes, or you could try picking IPv6 addresses yourself rather than relying on the autodiscovery stuff. I'm assuming there are RFC1918-like IPv6 address ranges from which to choose, but I don't know what they are at present.

[Chauhan Amit]

Hi Rick,
Replies are INline:

a)I'm guessing that when you used IPv4 addresses you happened to put them in separate subnets?

We used IPv4 address also of same sunet.

Althought we can try using some other address then link-local address. I will try that out.

But I still feel that I am missing something on the IPv6 routing part.

Regards,
Amit

[rick jones]

If the v4's were same subnet, and the connections were pt-pt, I suppose it could be the case that the HP-UX system was willing to receive the v4 on the "wrong" interface, or perhaps in all the shuffling some of the correct information got into ARP. My recollection is that UX does use a weak end system model and will accept datagrams for any of its IP's on any interface, but I didn't think that ARP was willing to respond that way. (Linux ARP is by default, which is a royal PITA when trying to do benchmarking and "know" on which interface traffic is flowing, even when they are in different subnets...)

Might look at the ARP tables (arp command) and then lanadmin stats when doing the pings to see on which interface(s) things are transmitted and received. That, or use tcpdump to do the same.