SG300-10P DHCP problems

[becherovka]

Hi,

I am using a SG300-10P in L3 mode to power 3 Cisco WAP321s and provide connectivity through a trunk to an ISP-provided broadband router. The switch has been working well except for the DHCP server, which has not been very reliable so far. The DHCP server periodically stops handing out DHCP leases, despite the availability of expired or non-allocated addresses in the network pool. To get the DHCP server "unstuck", I disable and then enable the DHCP server. After doing this, I immediately see a burst of traffic at each of the APs and a bunch of new leases are issued. 

The APs are usually quite busy, with up to 30 clients at each AP during brief periods. Most users stay connected for an hour or less, so there's a lot of turnover. Leases are configured to last for 1 hour. 

Any suggestions? I can post my configuration and a couple of odd screen shots I took when the DHCP was hung. I can also allow access to someone from Cisco, if they want to look around.

I really need to solve this soon or replace the SG300 with something else, because the business where it is installed is getting tired of their customers complaining about the WiFi.

Thanks,

Bob

[becherovka]

Just to follow-up, I ended up disabling the DHCP server in my SG300 and configuring DHCP relay to a Linux server running dnsmasq. This arrangement has been working without incident for over a month. dnsmasq has a lot of useful features, including the ability to configure it as an "authoritative" DHCP server. 

[Brandon Svec]

Thanks for coming back to update this.  I was following this post with some interest.

[viningele]

I've been experiencing similar issues on one installation with an active guest vlan.  It seem the table binding table fills to the pools capacity and even though most leases have expired the server just stops serving.  It I clear the vlan's binding entries it seems to work but I haven't been able to test onsite. It seem to me that if a client is already in the binding table as expired it's re-issued its ip but if the table is full and a new client makes a dhcp request even if 90% of the binding table entries are expired it won't clear the oldest expired client and issue to the new requesting client.  Still a buggy switch IMHO.