Failover strangeness

Greg G ggersh at ctc.net
Wed Oct 11 16:28:30 UTC 2006


David W. Hankins wrote:
> On Tue, Oct 10, 2006 at 04:17:38PM -0400, Greg G wrote:
>   
>> Greg G wrote:
>>     
>>>    The lease didn't transition back, even after a few hours, and the 
>>> peer went down.  Is there a way to make that transition happen?
>>>       
>
> If the server is in partner-down state (not communications-interrupted),
>   
    The server did go to communications-interrupted state.  I never saw 
a partner-down state.
> I think it should happen after STOS+MCLT.
>
>   
   STOS? I'm not familiar with that.  I see a reference to it in the 
dhcpd man page, but not the dhcpd.conf.  I couldn't figure out what the 
key is for the failover-state object to look at it through omshell.  I 
had hoped it was "name", but that doesn't appear to be so.
   Here's what I get in the leases file on the secondary when the 
primary stops and then starts:

failover peer "ontdhcp" state {
  my state communications-interrupted at 3 2006/10/11 16:19:28;
  partner state normal at 1 2006/10/09 18:44:33;
  mclt 30;
}
 
failover peer "ontdhcp" state {
  my state normal at 3 2006/10/11 16:20:48;
  partner state normal at 1 2006/10/09 18:44:33;
  mclt 30;
}

  I can wait hours, but the lease will never go back over to being 
handled by the primary...
> If it doesn't, that's probably a bug.
>   
   I'm inclined to think so, but I wouldn't press the issue. :)

   The really strange things is that if the secondary goes down and the 
primary complains, when the secondary comes back up, the primary takes 
control of the lease again!
> You're running into odd siutations here because failover isn't really
> designed for one-lease pools like this.
>
> Failover survivability all hinges around having a number of leases (free
> and backup states) allocated to each server in the event they should
> lose contact with the other.
>
> When there's only one, that will never be better than a single point of
> failure.
>
>   
    Hmmm.  I'm really looking for redundancy, so that either of the dhcp 
servers can serve up the lease for these clients.   Both servers are 
getting the request (via ip-helper), and I don't want both of them 
offering the lease.  That could get ugly.
> Although you could experiment with 'reserved' dynamic leases, which
> by definition are covered by both the 'free' and 'backup' states
> simultaneously.
>
>   
   No joy, unless I did it wrong.  I put in the "infinite-is-reserved' 
in the shared-network scope on both servers.

   Thanks.

-Greg G




More information about the dhcp-users mailing list