Alright, I got into a pissing contest with the Network Engineer here about why a device wouldn't complete an IP lease, here is my win:
I've placed X's in IP addresses for security.
1. DHCP Block in
/etc/dhcp.master (includes failover config for our HA dhcp cluster, dont get lost/overwhelmed):
Quote:
subnet 19x.60.23x.120 netmask 255.255.255.248 {
option subnet-mask 255.255.255.248;
option broadcast-address 19x.60.23x.127;
option routers 19x.60.23x.121;
pool {
failover peer "dhcp-failover";
deny dynamic bootp clients;
range 19x.60.23x.122 19x.60.23x.126;
}
}
|
2. An example lease in
/var/lib/dhcpd/dhcpd.leases(the file that shows all leases for your defined DHCP ranges):
Quote:
lease 19x.60.23x.122 {
starts 4 2011/06/16 14:18:36;
ends 4 2011/06/16 20:18:36;
tstp 4 2011/06/16 11:18:36;
tsfp 4 2011/06/16 23:18:36;
atsfp 4 2011/06/16 23:18:36;
cltt 4 2011/06/16 14:18:36;
binding state active;
next binding state expired;
hardware ethernet 00:24:8d:82:4e:04;
uid "\001\000$\215\202N\004";
}
|
Now, since my DHCP network is 19x.60.23x.120 255.255.255.248, as defined by Step 1., .120 is the network, .121 is the Gateway. .122-126 are usable addresses, and .127 would be the broadcast address.
Imagine there being a lease for .122-126 in the dhcp.leases file as defined in step 2., so I dont inflate this reply with superfluous confusion.
when a 6th device tries to lease an address on this network via DHCP, you'll see this in
/var/log/dhcpd:
Quote:
Jun 15 15:48:10 sprodhcp01 dhcpd: DHCPDISCOVER from 64:31:50:38:f0:bc via 19x.60.23x.121: network 19x.60.23x.120/29: no free leases
|
going back to Step 2., notice that there is a stard/end (end being expiration) to the lease:
Quote:
starts 4 2011/06/16 14:18:36;
ends 4 2011/06/16 20:18:36;
|
Plain and simple, a new device isn't giong to get an IP address until leases have expired. Adding a "new" device to the network and you're not able to get an IP via DHCP (even if there are under the limited amount of devices on the DHCP network)? Clear the leases out of
/var/lib/dhcpd/dhcpd.leases, save, and try again.
No, arp caches are not the issue here. clearing the arp cache on your routers/switches doesn't free up DHCP space, and it doesn't give a single switch port a free slot for your new device to ARP up with.
This entire battle came from the following arp entry in a cisco router:
Protocol Address Age (min) Hardware Addr Type Interface
Internet 20x.8x.9x.2x4 0 Incomplete ARPA
I hope this is helpful.