spawning class on ethernet

[Eugene M. Zheganin]

Hi.

On 11.02.2015 12:40, Eugene M. Zheganin wrote:
> Hi.
>
> On 09.02.2015 19:47, Frank Sweetser wrote:
>> I would guess it's because the hardware address isn't visible in
>> DHCPv6, unless you're using a relay agent with RFC6939 support.  You
>> may have better luck using the DUID instead.
>>
> Thanks, it helped.
>
Nah.... it does something other.
I created it as this:

class "default" {
    spawn with duid;
    lease limit 1;
}

subnet6 fd00::0300/120 {
    pool6 {
      range6 fd00::332 fd00::3fe;
      allow members of "default";
      option dhcp6.name-servers fd00::301;
    }
}

And the pool degraded to "Unable to pick client address" state very quickly.

In order to investigate this I did the experiment - I deleted the
dhcp6.leases on the server in a test network, and the server was unable
to give any address right from the start. All the clients received the
"status code for NA-0: no addresses" message (this one is from dhcp6c,
for example), zero leases were written to the dhcp6.leases.

So I'm still fighting the degrading pools - I have like 20-30 clients in
each, each has at least 200 leses available - and the pool degrades to a
statet when dhcpd think that there's no free leases available each 2-3
days. For now I restart the dhcpd each night - it helps.

I have even more questions.
1) Is there any way to determine the pool state - number of free and
used leases ? For example dhcpv4 says this at the start.
2) Why there isn't any indication of where the request for a new lease
was received ? dhcpv4 indicates the interface. I took a brief look at
the dhcpv6 sources, seems like there is no trace of interface in data
structures - why ?

Thanks.