Active lease that does not move to expired

[John Wobus]

On Jul 21, 2006, at 1:12 PM, David W. Hankins wrote:

> On Wed, Jul 19, 2006 at 12:57:52PM -0400, John Wobus wrote:
>> On the primary:
>> lease 128.84.152.133 {
>>    starts 3 2006/07/19 09:49:44;
>>    ends 3 2006/07/19 10:48:44;
>>    tstp 3 2006/07/19 10:48:44;
>>    tsfp 3 2006/07/19 10:48:44;
>>    atsfp 3 2006/07/19 10:48:44;
>>    cltt 3 2006/07/19 09:49:44;
>>    binding state free;
>>    hardware ethernet 00:02:2d:53:32:1e;
>>    uid "\001\000\002-S2\036";
>> }
>>
>> On the secondary:
>> lease 128.84.152.133 {
>>    starts 3 2006/07/19 09:49:44;
>>    ends 3 2006/07/19 10:48:44;
>>    tstp 3 2006/07/19 10:48:44;
>>    tsfp 3 2006/07/19 10:48:44;
>>    atsfp 3 2006/07/19 10:48:44;
>>    cltt 3 2006/07/19 09:49:44;
>>    binding state active;
>>    next binding state expired;
>>    hardware ethernet 00:02:2d:53:32:1e;
>>    uid "\001\000\002-S2\036";
>> }
>
> That's a neat trick.
>
> The primary's entry indicates clearly that the secondary
> acknowledged the lease as being in that state.  There's
> no reason for the primary to retransmit that lease's
> state.
>
> The secondary's entry looks as though it never got the
> expired binding update from the primary (whose ACK it
> would send to tell the primary to move from expired to
> free).
>
>
> Did you roll back to an older leases database?
>
> Has the secondary been having a bout of restarts?
>

We never rolled back either leases file.

We restart the secondary (actually, both) servers
frequently for configuration updates.

Over the course of a couple of days, I've seen the following:
we had six or seven IP addrs in this situation, and the
count dwindled down to just one, as of now.  Over that
time, I specifically looked for any additional IP addrs going
into this situation and I watched at least one do so.

I'm monitoring the leases files at this point, to look for these.

My most obvious concern is whether the secondary
would ACK a renew request.  I already assume
that the primary could allocate the IP at any instant.

I suspect that in practical terms, in the vast majority of
instances, the first time either server does anything with
the addr, intercommunication takes place and the
situation apparently is straightened out.

John