Hello everyone,

I am trying to run the ISC DHCP server in 4094 VLANs simultaneously for a big network of virtual machines. The problem is that the server segfaults when I put it in more than 1370 VLANs. I have tried with several versions (3.11, 4.0.2, 4.1.1).
Have someone tried before to do this and succeeded? Do you think it is a good idea to have just one process with more than 4000 open socket or it will be a big performance problem?

If you want more information just ask. I also accept any idea of how to solve this problem.

Thank you

Jeez! "4000" of *anything* in one process ... sounds like a big *design* problem!

What exactly are you trying to accomplish?

And how big is your lab? Like, you're not trying to do all this on one PC, are you? Or worse, a single Celeron-class 32-bit PC with 1GB or less of RAM and an old version of Redhat Linux, are you?

What's your goal, and do you have sufficient hardware to support this goal?

Hehe, first of all let me tell you that I am only assessing the options.
This is for an application that will run in lots of VMs in a lot of nodes (I'm thinking about maybe thousands).
The point is that the DHCP server will have to be remotely controlled from a master application, thus I'd prefer if I had to communicate with just one process. I could put a DHCP server in each VLAN but that would mean that the controller application would have to be in all the VLANs and control 4000 DHCP servers.
At fist I thought it would not be such a problem. In principle a DHCP server only exchanges dhcprequest-dhcpresponse messages which does not seem to require a lot of computational resources. Obviously I would have to put a failover DHCP server or something.
In the tests I have done, although the DHCP server crashes, Linux seems to be able to handle 4000 VLANS (= 4000 virtual interfaces) without any problem.
I don't know, probably I will try with 4 servers with 1000 VLANs each or something and see how it behaves...

If someone has any experience or suggestion please leave a message.

I do hope you are not trying to use ISC DHCPD with tagged WLAN's to assign two IP addresses to the same MAC address on two different networks. That won't work (known bug).

Hi, thank you for your answer. I suppose you mean tagged VLANs, and yes it was an option to reuse MACs. Do you have a link to the bug? I'd like to know more about it. It sounds strange that the server behaves different depending on whether the frames are tagged or not.

Thank you very much for your help

A working configuration example

Server ip : 172.27.0.10
Swtich ip / gateway ip of server : 172.27.0.100
Option domain –name servers : 172.27.0.6

Please note default vlan 172.27.0.x in this case

Switch and server need to be in default lan for communication / or else we need to trunk in case other vlan connected , we need to configure vconfig on server to communicate


For biggner I would recommend to go for the default vlan connectivity


Example :

ddns-update-style interim;
ignore client-updates;
default-lease-time 43200;
max-lease-time 43200;
authoritative;


#-----subnet mask-- broadcast-- gateway-#

option subnet-mask 255.255.255.0;
#option broadcast-address 192.168.1.1;
option routers 172.27.0.100;
option domain-name-servers 172.27.0.6;


# ----------- Server Scope and vlan1 with switches and server--------------#
subnet 172.27.0.0 netmask 255.255.255.0 {
#range 172.27.0.0 172.27.0.50;
option routers 172.27.0.1;
#option subnet-mask 255.255.255.0;
option broadcast-address 172.27.0.255;
#option domain-name-servers 172.27.0.5;
}


# ----------- clients1 Scope-vlan2-------------#
subnet 172.27.1.0 netmask 255.255.255.0 {
range 172.27.1.3 172.27.1.200;
option routers 172.27.1.1;
option subnet-mask 255.255.255.0;
option broadcast-address 172.27.1.255;
#option domain-name-servers 192.168.1.1 #
}


Save the file

Chkconfig –list dhcpd on
{ For enabling the service on all run levels }
Service dhcpd start
In case , dhcp server faild to start , check with log messages

/var/log/messages

Check for the ip helper address in layer 3 swtich , which act as dhcp-relay
Which has to be configured as 172.27.0.10


Troubleshooting Dhcp server start up error :

1) Run the Dhcp service in debug mode
2) Check the ip configuration
3) Check for syntax errors in dhcpd.conf file
4) Check for the right location of the dhcpd file
5) Ping test between the switch and server
6) Possible conflict of other dhcp server may be the issue

dhcpd in the foreground in debug mode with /usr/sbin/dhcpd -d –f

Hopefully, a DHCP server like the one we’ll be configuring will respond. Running tcpdump shows a dhcp request looks like:
17:26:02.003956 00:00:00:00:00:00 > ff:ff:ff:ff:ff:ff, ethertype IPv4 (0×0800), length 342: 0.0.0.0.68 > 255.255.255.255.67: BOOTP/DHCP, Request, length 300
You should notice DHCP running in the process (ps) list. Any problems, check syslog

Congratulations you have finally configured In easy steps