[Dnsmasq-discuss] Lease comes back after removing

[A C]

On 2019-06-26 13:32, Geert Stappers wrote:
> On 24-06-2019 02:38, A C wrote:
> 
>> On 2019-06-23 11:08, Geert Stappers wrote:
>>> On Sun, Jun 23, 2019 at 09:33:21AM -0700, A C wrote:
>>>> On 2019-06-23 03:32, Geert Stappers wrote:
>>>>> On 23-06-2019 08:06, A C wrote:
>>>>>
>>>>>> I'm having trouble removing leases from my system.
>> I have:
>>
>> /etc/dnsmasq.conf:
>> read-ethers ###(there is a dash)
>> dhcp-range=10.0.200.100,10.0.200.105,255.255.0.0,12h
>> ### My netmask is actually 255.255.0.0 on the network, all dhcp leases
>> are in 10.0.200.x, static assignments are in other subnets
>>
>> /etc/ethers:
>> 00:25:9C:1C:0A:C3 ap.lan
>> ### and others
>>
>> /etc/hosts:
>> 10.0.0.252  ap.lan
>> ### and others
>>
>>
>>>>>> How can I fully purge the lease and get it to start over?  The client is
>>>>>> an embedded device and has no memory of the lease once I power cycle it
>>>>>> so somehow dnsmasq is remembering even though the lease file was purged.
>>>>>>
>>>>> Recently I had simular sympthoms, infact still have them.
>>>>>
>>>>>
>>>>> My only issue is when I forget to add a client to the list
>>>>> prior to first putting it online and it acquires a lease from the DHCP
>>>>> lease pool instead of the IP I want to give it via the combination of
>>>>> ethers/hosts.  After that I have extreme difficulty purging the lease to
>>>>> force a new address.
>>> My /etc/ethers has a line like
>>>
>>> 00:02:b0:ef:f0:ef  kornuit
>>>
>>> Doing `host kornuit` returns correctly the IPv4 address that is in DNS.
>>> (other DNS as dnsmasq  ( no kornuit entry in /etc/hosts ))
>>>
>>>
>>> I also stopped dnsmasq, removed the lease file and restarted dnsmasq.
>>>
>>> Did another PXE netboot  ( consider it an embedded device, it has no
>>> memory of previous lease )
>>>
>>> Server kornuit got the IPv4 address as before, one from the Dnsmasq
>>> DHCP range. Where it looks like a persistent DHCP lease, do I think
>>> that it is algoritme that hands out each time the same IPv4 address.
>>>
>>> Thing I need to verify is if dnsmasq does do a DNS-lookup on 'kornuit'
>>> after reading /etc/ethers.
> 
> Now using  `dhcp-host=MA:CA:DD:RE:SS,192.0.2.42
> 
>> host ap.lan localhost (run on the same server as dnsmasq) returns
>> 10.0.0.252 as expected.
>>
>> Turned off the device, stopped dnsmasq, removed the lease from the lease
>> file, restarted dnsmasq and then restarted the device.  I get the old
>> DHCP range lease instead of the one defined in hosts.
>>
>> If I add a host entry (with an IP outside of the DHCP range) and an
>> ethers entry for a device bore I ever plug it into the network I will
>> correctly get a static lease of the IP I define in hosts.  If I forget
>> to do that, the device picks up a lease from the DHCP range and then
>> it's a massive fight to get rid of it.  Sometimes I have to leave the
>> device disconnected for a few days so the lease expires on its own
>> (regardless of deleting the lease) and then I can switch the IP.
>>
>> There seems to be some kind of persistence that exists outside of just
>> the lease file but I don't know where that is.  This isn't the first
>> time this has happened and it's happened with different embedded devices
>> and non-embedded devices (cameras, AP's, IoTs, phones, etc.) there just
>> seems to be sticky lease data somewhere.
>>
> You are probadly right about the  "sticky lease data somewhere"
> 
> (I'm not comfortable to say "you are right about sticky lease data" )
> 
> 
> The lease file I removed had size of zero bytes.
> 
> 
> My (educated??) guess is that lease data is stored in Dbus.
> 
> My only point to back that up, is `dnsmasq -v` showing  "DBus"

Yes, my copy does have DBus support.  I've turned on dbus-monitor to see
if anything shows up at the next lease update.  There's enough devices I
shouldn't have to wait long.