DHCP Issues with RV345P

[tech4shcs]

My organization is a small private school that uses the Cisco RV345P as Router and DHCP server.  We currently use a /23 mask.  We have about 40 devices that have a static DHCP setup, otherwise everything else is assigned as needed.  The network runs great when school is not in session, but when we have students trying to utilize their chromebooks we seem to have an issue with the DHCP server not providing addresses quickly, or at all in some occasions.  A little about our setup, we have wired connections to all rooms that go through 1 of several switches.  Each switch connects back to the RV345P.  We also have Ubiquiti access points connected via Ethernet to the switches as well.  Due to recent increases in the amount of technology in the building, we are looking to reconfigure our setup to utilize a /22 mask.  This reconfiguration won't take place until we have an institutional break that allows time for the network to be down.  Anyhow, we currently keep running into the issue of devices not getting ip addresses when we have a large volume.  The current configuration was in place when I arrived, so I'm looking for advice on troubleshooting this situation, as well as tips and suggestions for reconfiguration.

[Enes Simnica]

@tech4shcs man, I've run into this exact problem at a school I help with. That morning Chromebook stampede is brutal on a router's DHCP server. Honestly, the /23 mask size isn't your issue u've got plenty of addresses. The RV345P is just getting overwhelmed when 100+ devices all ask for an IP at the same time.

So before your big reconfig, try this one thing: lengthen your DHCP lease time. If it's set to a few hours, bump it up to 8 or even 24 hours. This drastically cuts down on the renewal traffic and should help it survive the morning rush.  Also, do a quick check to make sure no one has plugged a cheap home router into ur network (it happened to me..) somewhere, those can become rogue DHCP servers and cause this exact problem.

The move to a /22 is smart for growth, but it won't fix this. If the longer lease time helps, you'll know the router itself was the bottleneck. Good luck

hope it helps and PEACE!

-Enes
CCNP x2 Enterprise
Your Friendly Networking Ninja

[tech4shcs]

First, thank you for your reply!  We have a lease setting of 24 hours currently.  It has been set for longer in the past, as much as 30 days lease...  

Would it help to do some static dhcp for the chromebooks, or possibly go static ips for the chromebooks?

Currently all switches are connected directly to the RV345P.  Should I have a distribution switch between my access switches and the RV345P?  

Will the /22 move still be manageable by the RV345P?  Should I consider moving to a windows based dhcp server?  Is there a better cisco router that would handle this network more efficiently?

Any other thoughts or suggestions on my current setup or changes that you'd suggest?

Thanks again!

[Jens Albrecht]

---

@tech4shcs wrote:

We have a lease setting of 24 hours currently.  It has been set for longer in the past, as much as 30 days lease...  

The lease duration of 24 hours is pretty short because the first renewal request is due after 12 hours already (50% lease time). This means that all DHCP clients need to renew their IP address every day.

Increasing the lease duration e.g. to 7 days will reduce the load on your router because the first renewal is due only after 3.5 days. Depending on the total number of clients you might have to move to a /22 network as your router will block the IP addresses for these 7 days.

---

Would it help to do some static dhcp for the chromebooks, or possibly go static ips for the chromebooks?

Sure. Every client that has a static IP address reduces the load on your DHCP server/router.

---

Currently all switches are connected directly to the RV345P.  Should I have a distribution switch between my access switches and the RV345P?  

This does not affect the load on your DHCP server/router. If you have large amounts of traffic between your clients and the router has problems to handle the throughput, then it may help. Regarding the DHCP problem there will be no effect.

---

Will the /22 move still be manageable by the RV345P?  Should I consider moving to a windows based dhcp server?  Is there a better cisco router that would handle this network more efficiently?

Yes, your router can handle a /22 network. The problem is the number of concurrent DHCP requests and this is independent of the network size. If you reduce the bursts e.g. by increasing the lease duration, then you can move to a /22 network as mentioned above.

A Linux or Windows DHCP server would definitely be able to handle the DHCP requests in your network without problems. The same applies to the enterprise-class Cisco routers and switches but these are much more expensive than a linux server.

As a result, an action plan might look like this:

• Configure critical clients with static IP addresses
• Increase the lease time to 7 days to reduce the load
  If the total number of DHCP clients is larger than about 400 you probably have to move to a /22 network
• Prepare a linux or windows DHCP server and use the RV345P as a backup DHCP server only.

HTH!

[tech4shcs]

Thank you for this information!  Another question I’ve been thinking about is moving my Guest WiFi access to a different IP address range.  Is that something that I could do by setting up a VLAN for the guest network, and then configuring my guest network within the Ubiquiti controller (self hosted) to operate my main WiFi on the primary VLAN and the guest network to operate on a separate VLAN?  If this would even work, would I need to change anything in my switch configurations or just on the dhcp server/router and the Ubiquiti controller?

thanks again!

[Jens Albrecht]

It is best practice and strongly recommended to separate Guest Wifi access from your users. You need to configure the new Guest vlan on all switches and the router and the connections from the access-points to the switch, between switches and from the switches to the router need to be configured as trunks. The Ubiquiti access-points will tag the Guest Wifi traffic with the Guest vlan so all devices need to know this vlan and understand the tag. Of course, you also need to define a new DHCP scope for the Guest network. This is the first step to establish connectivity.

The next and really important step is to separate the Guest Vlan from your users. I assume all routing is done on the RV345P so you need to configure firewall rules to prevent the Guest Vlan from talking to your users. You definitely want to prevent Guests from accessing any resources at your school. On your RV345P you need to go to "Firewall" --> "Rules" and create rules to DENY any traffic between your Guest Vlan and the regular user Vlan.

HTH and please mark posts as Helpful/Solution if your questions got answered. Thanks!

[tech4shcs]

Sorry, another question.  The single vlan that exists now (vlan1) has not ports tagged.  When I create the new vlan (vlan2) which ports should be tagged, untagged, or excluded?  In terms of the tagged, untagged, or excluded, does that affect what equipment is plugged into which ports, or am I thinking about that wrong?  Also, if vlan2 is tagged, that would mean that vlan2 would be tagged across all swithces, access points, etc.

Thank you for your assistance with this!

[Jens Albrecht]

If I understand your original post correctly, then all switches simply have a direct connection to the RV345P router, right?

In this case all ports that connect to your APs, all ports that connect your switches to the router and the ports on the router that connect to the switches need to transport Vlan 2 as tagged. The config always has to match between both ends of a trunk connection.

On all these ports you need to transport 2 vlans, i.e. vlan 1 and vlan 2. This is only possible if at least one vlan is tagged. Since vlan 1 is currently untagged on all ports, the easiest way is to tag vlan 2 to minimize the configuration overhead.

HTH!

[tech4shcs]

Ok, that makes perfect sense, and yes all switches connect back directly to the router.  So to make sure I have everything right in my mind...  vlan1 (default) untagged
             vlan2 (guest network) tagged
             any additional vlans (i.e. voip) tagged

Will the RV345P provide dhcp server for each vlan configured, even if different ip addresses/masks?

For statically assigned addresses, should I have those addresses excluded from the dhcp server range?

Finally, should the statically assigned address be assigned at each device, on the static-dhcp configuration list, or on both?

Thank you again for the assistance!  This is tremendously helpful!

[Jens Albrecht]

---

@tech4shcs wrote:

Ok, that makes perfect sense, and yes all switches connect back directly to the router.  So to make sure I have everything right in my mind...  vlan1 (default) untagged
             vlan2 (guest network) tagged
             any additional vlans (i.e. voip) tagged

Correct. Vlan 2 and any additional vlans need to be tagged.

---

Will the RV345P provide dhcp server for each vlan configured, even if different ip addresses/masks?

Yes. The RV345P has IP interfaces in each vlan. So if it receives a DHCP request for vlan 2 it will reach the IP interface for vlan 2 telling the router which network is asking for the IP address.

---

For statically assigned addresses, should I have those addresses excluded from the dhcp server range?

Yes. The router does send pings to an IP address to check whether it is already in use, before handing it out. BUT this only works if the devices with static IP addresses are online. If not, the router would hand those addresses out and you get problems when the devices with static IP's go online later.

---

Finally, should the statically assigned address be assigned at each device, on the static-dhcp configuration list, or on both?

No need to configure the static-dhcp configuration list. This is used if you do NOT assign the IP address directly on the device but want to always use DHCP. However, as previously discussed your purpose is to reduce the load on the DHCP server. Therefore, configure them statically on the device, exclude these addresses from the DHCP server range but do NOT configure the static-dhcp configuration list.

HTH!

[tech4shcs]

Ok, after much investigation and following the suggestions provided, it seems that I’m having an issue with the “dnsmasq”.  The list of errors in my log file shows errors related to this every few seconds.  I’ve attached the log file, if that’s helpful.  I’m looking to do a factory reset and reconfigure the router.  Is this the way to go? Is there a better option?  Does this point to a hardware issue that reconfiguration won’t solve?

thank you again for your assistance!

[Jens Albrecht · Accepted solution]

Looks like the problem is caused by a corrupted DHCP lease file leading dnsmasq to crash on start.

This router has reached EOL quite some time ago so if this is a bug, it will not be fixed anymore. I assume you run the last firmware release 1.0.03.29, right?

I don't know what exactly you changed/configured but the first action should be a reload of the router. As mentioned previously this model is not designed to handle large amounts of concurrent DHCP requests. The memory might also limit the maximum size of the DHCP lease database. How many lease entries do you see?

If a reload does not help, you can of course try to reset and reconfigure the router. However, if the router is running out of resources, then this won't help. Nevertheless it is worth a try.

HTH!