On Jul 3, 2009, at 1:25 AM, Bob Tanner wrote:

> $ dig metaserver2.netrek.org
> metaserver.psychosis.net.
> 18.181.1.140
> 199.233.98.170

Pointed at two machines under my control as you noted.

> $ tools/metaget XVM-ONE-140.MIT.EDU
> r,1
> pickled.netrek.org,2592,2,41,6,0,B
>
> $ tools/metaget neural.psychosis.net.
> r,2
> continuum.us.netrek.org,2592,3,660,0,0,B
> pickled.netrek.org,2592,2,65,6,0,B

Not sure why that is at the moment. They're running identical configs  
with an identical meta codebase.

I SUSPECT that it's because continuum, in this case, has decided to  
solicit one of the IPs and not the other due to not applying the DNS  
RR properly. (When I tested this when I initially set it up, it seemed  
to work correctly.)

I'll try removing the entry for neural.psychosis.net from DNS in the  
morning and see if that fixes things with soliciting onto the MIT  
machine. If so, then I'll just list one of the two until we figure out  
what's seemingly caching a solitary entry... (We could just have both  
machines in the rotation, but I actually have some custom scripts set  
up with multiple replicated DNS masters that let me do RRs and have  
"dead" IPs pulled out of the RR as needed; kinda a hack, but it works  
well for connectionless UDP or short-lived TCP sessions. I'm using  
that for multiple DNS entries, including some IPv6 testing I'm doing,  
so I'd like to see it working properly rather than just doing a  
workaround.)

So, I could just hardcode both servers in as non-solicit permanent  
entries, but I'd rather try to figure out what exactly is going on...