"Healthy" servers issuing leases for length of MCLT

[Matt Causey]

On Thu, Sep 9, 2010 at 2:39 PM, Matt Causey <matt.causey at gmail.com> wrote:

> On Sat, Jul 10, 2010 at 1:07 AM, Glenn Satchell <
> glenn.satchell at uniq.com.au> wrote:
>
>>
>> The initial response to a DHCP DISCOVER will be to issue a lease of MCLT,
>> then the next renew will be your specified lease time. It works this way
>> because the DHCPOFFER is sent before communicating to the partner server.
>>
>
> Thanks for the info.  We have a very short MCLT (300 seconds) and 12 hour
> lease time.  I can confirm that our 'normal' clients behave in the way you
> describe:
>
> tc at box:~$ sudo udhcpc -i wlan0 --foreground
> udhcpc (v1.14.3) started
> Sending discover...
> Sending discover...
> Sending select for 172.24.120.141...
> Lease of 172.24.120.141 obtained, lease time 300
> deleting routers
> route: SIOCDELRT: No such process
> adding dns 10.4.4.10
> adding dns 172.24.98.22
> adding dns 172.24.98.21
> adding dns 10.5.40.1
> Sending renew...
> Sending renew...
>
> <wait...>
>
> Sending renew...
> Lease of 172.24.120.141 obtained, lease time 43200
> deleting routers
> route: SIOCDELRT: No such process
> adding dns 10.4.4.10
> adding dns 172.24.98.22
> adding dns 172.24.98.21
> adding dns 10.5.40.1
>
> So, that part looks good.
>
> Now our Cisco access points seem to be actually sending the 300 second
> lease time in their DHCPREQUEST:
>
> 19:20:06.439558 IP (tos 0x0, ttl 254, id 18499, offset 0, flags [none],
> proto: UDP (17), length: 336) CLIENT.bootpc > SERVER.bootps: [udp sum ok]
> BOOTP/DHCP, Request from 00:26:0b:c9:eb:e4 (oui Unknown), length: 308,
> xid:0x97f, flags:
>  [none] (0x0000)
>           Client IP: phx3-wap-97-189.phx3.amazon.com
>           Client Ethernet Address: 00:26:0b:c9:eb:e4 (oui Unknown)
>           Vendor-rfc1048:
>             DHCP:REQUEST            MSZ:1152
>             CID:[ether]00:26:0b:c9:eb:e4
>             LT:300
>             HN:"ACCESS-POINT"
>             PR:SM+NS+DN+WNS+DG+LOG+SR+T150+VO
>             VC:"Cisco AP c1250"
>
> which seems to be followed by the reply containing the 300 second lease
> time:
>
> 19:20:06.445402 IP (tos 0x0, ttl  64, id 0, offset 0, flags [DF], proto:
> UDP (17), length: 331) SERVER.bootps > CLIENT.bootpc: [bad udp cksum aa72!]
> BOOTP/DHCP, Reply, length: 303, xid:0x97f, flags: [none] (0x0000)
>           Client IP: BLAH
>           Your IP: BLAH
>           Server IP: BLAH
>           Client Ethernet Address: 00:26:0b:c9:eb:e4 (oui Unknown)
>           file "filename"
>           Vendor-rfc1048:
>             DHCP:ACK
>             SID:HOST
>             LT:300
>             SM:255.255.255.0
>
> The DHCPREQUEST from other clients does not ask for a particular lease
> time, so they get the 12 hour one from the server.
>
> Soooo, I intend to bug Cisco about their dhcp client implimentation.  But
> is there anything I can do to fix this from my end?  I see the
> min-lease-time option.  Will that help?
>

Hum.  Well adding the min-lease-time parameter causes the desired behavior.
My access points are getting 12-hour leases now - even though they are
asking for 5 minute ones, and this does not appear to interfere with the
MCLT-assigned 'initial lease' that is assigned to a device.

Are there any potential negative implications to enforcing a min-lease-time
that's the same as the default lease time?

--
Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/dhcp-users/attachments/20100910/82bef688/attachment.html>