From: lcargill99@comcast.com   
      
   Robert Wessel wrote:   
   > On Thu, 24 Mar 2016 08:56:00 -0500, Les Cargill   
   >  wrote:   
   >   
   >> Robert Wessel wrote:   
   >>> On Wed, 23 Mar 2016 12:32:02 -0500, Les Cargill   
   >>>  wrote:   
   >>>   
   >>>> (directions to a better asking-place always appreciated )   
   >>>>   
   >>>> Suppose I had a pseudo random network of IP nodes. Say,   
   >>>> 50 out of a possible 2000 were on a given network at any given time.   
   >>>>   
   >>>> I want to have a DNS-like solution that uses the static hostname of   
   >>>> each node to populate the DNS tables.   
   >>>>   
   >>>> DHCP may or may not be run on the same server. Dunno yet.   
   >>>>   
   >>>> There may be DHCP for more than one Class C subnet. Dunno that yet,   
   >>>> either. I don't have a clear picture of how coupled DHCP and DNS   
   >>>> really are.   
   >>>>   
   >>>> Human beings would be personally responsible for setting the hostname   
   >>>> per node.   
   >>>>   
   >>>> Sadly, it's better if this runs on Windows, and not necessarily a   
   >>>> Windows server offering. It can simply be something that starts up on a   
   >>>> workstation.   
   >>>>   
   >>>> I would not be adverse to a small-footprint Linux VM on said Windows   
   >>>> machine to do this. Also maybe some small ARM Linux solution perhaps.   
   >>>>   
   >>>> What is this sort of thing called? I don't think the classic DNS BIND   
   >>>> service works for this, but I'd be eternally grateful to be proven   
   >>>> wrong - that's a "one book" solution.   
   >>>   
   >>>   
   >>> I'm not sure I fully understand the problem you're trying to solve.   
   >>   
   >> You are excellent company. I'm not sure I do either.   
   >>   
   >> In one sentence: I observe people using static IP addresses on a   
   >> semi-random network (think an engineering development/test network   
   >> without official IT sanction) and I wonder if there's some way to   
   >> organize this with a combination of DHCP and DNS in a lightweight and   
   >> inexpensive fashion.   
   >>   
   >> I'd like there to be persistent state (hostname) on the nodes   
   >> that is used to influence the choice of DNS name, so that   
   >> pinging "TestTarget42" always gets the same node, but the IP address   
   >> can then be different.   
   >   
   >   
   > How much can the nodes participate in this process? They could, for   
   > example, issue a DDNS update to a small dedicated DNS server (which   
   > could run on a Windows machine).  As I mentioned that's basically what   
   > Windows workstations do in dynamic update mode.   
   >   
   > If that's too heavyweight, could the nodes send some sort of message   
   > to a central machine, which would run a small script in response to   
   > edit a zone file, and then trigger Bind (or whatever) to reload the   
   > zone file?  You need very little if you just want to refer to the   
   > machines by name, just the A records (ignore reverse lookups and all   
   > that).  Just script removing any exist A record with the IP address,   
   > and then add a new one with the IP address and name.  Reserve a whole   
   > subdomain for this (*.testnodes.mycompany.com).  You'd also need to   
   > update the zone serial number (just use the current date/time),   
   > although if you don't have a secondary server you can omit that.  Keep   
   > the TTLs short.   
   >   
   > If your DNS server supports something like nsupdate (Bind does, and   
   > you can run Bind on Windows), just update the DNS server by creating   
   > an appropriate update script executing the it with nsupdate.   
   >   
      
   I rather like that. Will check. Thanks.   
      
   > If you can't do much on the nodes,   
      
   The nodes vary in intelligence.   
      
   > perhaps you could poll the likely   
   > IP addresses, and query the host names at each (which obviously would   
   > require some support at each node).  The do the DNS update as things   
   > change.   
   >   
      
   That's actually one thing I have considered. One node will always have   
   the option of a full ARP table, the rest is moving data.   
      
   I think the "nbstat" command can be used to find Windows hostnames   
   given the IP address.   
      
   > In any event, then all you need is your main name server to point to   
   > the testnodes subdomain.   
   >   
   > If I'm understanding this correctly, it shouldn't be too difficult.   
   >   
      
   I would not think so.   
      
   Thanks for your reply. It's kind of interesting there is no well-trod   
   path here*. You'd think this would be something people had done before.   
      
   *I say that based on other Googling round.   
      
   --   
   Les Cargill   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)