Ok, so here's my theory. Not positive but it is the only thing that makes sense.
On the evening of August 30th, I changed the DHCP server for the staff subnet from my ASA to my 3750 core switch, in an effort to centralize all DHCP servers. On the ASA, the lease time was configured to 24 hours. I then disabled DHCP on the ASA and moved it to the 3750. However, all of those leases the ASA had already given out were still active, so clients had no reason to ask for a new address. This is why there was no impact when I initially moved it over to the 3750.
I told the 3750 the pool range for distribution was .75 - .233, same as when it was on the ASA. When the 3750 started giving out it's own leases, it would do it's ping checks before making the offer to make sure that the IP it wants to give out is not in use. However, when it received a reply to the address it wanted to give out, since it wasn't the DHCP server that originally gave it out, it just views it as a conflict, and moves it to the blacklist table.
Rinse and repeat this over and over and eventually we have an exhausted DHCP pool with a combination of the conflict table and the active leases, which explains why users were able to get addresses at first, but then were no longer able to do so.
Sounds like what I need to do is either A) Find a way to make all the client as for new leases once the DHCP server is moved, or B) Clear the conflict table, and then disable conflict logging. I think option B is more feasible.
Your thoughts are appreciated.
