Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm with a small company that plans to sell a rack mounted network appliance which is configurable via a web interface (think of a router configuration page sort of deal), and I'm wondering in large data center like environments what the process usually is for the initial setup of such systems. The main question is, if the system is headless, how do you get initial remote access to it? Do companies usually first plug a server into a monitor/keyboard/mouse in order to configure the network settings before mounting it in a rack? How else would they know what the IP address of the machine was if DHCP (and it can't be hard coded because of IP conflict potential)?

share|improve this question
1  
If it should be deployed in a datacenter, please make it configurable by SNMP and add either VLAN or a separate configuration interface to allow for separation of user and management data. Failing that, serial console on RJ45 plug and an option to turn off management from the network. –  Simon Richter Oct 26 '12 at 6:42
add comment

4 Answers

up vote 3 down vote accepted

There are multiple approaches to this problem. Let me at least point you in the right direction.

Most SOHO routers come with a default IP address in the RFC 1918 address space. This configuration may be modified by connecting a host usually by ethernet cable and then pointing a browser to that default address. Sometimes the devices offer DHCP so that the admin client doesn't have to manually set an IP in the correct network to connect for the first time.

Another common approach is to connect out-of-band, usually over serial cable. Higher end routers and switches come with this cable and end users install the device, power it on, and connect to the serial port from a local machine. From there, they configure the IP to match their network.

One final approach would be to set the ethernet port on the device to request a DHCP address. You would then plug the device into a DHCP network such as a small office network or home network. You could use process of elimination to determine what the new device's IP address is and then connect to a web UI or similar interface from there.

share|improve this answer
add comment

Initial setup of such devices is often done first through a serial console. Another option is a static IP address on a common subnet. That forces a technician to either put the device on a staging subnet or directly connect to the device using the same subnet statically assigned to a laptop or other computer. A final option is acquiring a DHCP address, which is pretty lame and only reserved for home networking equipment.

share|improve this answer
    
I disagree with the idea of DHCP being used only by consumer grade equipment. I've seen enterprise networks that use static DHCP to centrally manage IP settings, and DNS name -> IP mappings. –  Kevin Soviero Oct 26 '12 at 4:22
    
@Kevin Yes, I've seen that too and even done it twice to the best of my recollection. I've never thought it was a good thing in general, but that's neither here nor there in this Q/A. The specific context of the question was how do appliances / devices get network settings out of the box, which makes reservations a moot point in this scenario. –  Wesley Oct 26 '12 at 7:36
    
Unless an administrator configured a DHCP server to only give out addresses to static DHCP entries, got the MAC address from the appliances network interface through some physical notation on the device, created the static entry in the DHCP server prior to putting the device on the enterprise network, and then through that gave the appliance its initial network configuration to then be able to access the web interface. Ew. I think I just threw up in my mouth. –  Wesley Oct 26 '12 at 7:39
add comment

The answer to this depends on what kind of role the device plays, its location in the network and the quantity of devices that need to be managed. A very small number of devices could do something like the default address approach (i.e. SOHO devices). If typical deployment calls for less than 50 or so boxes and there's a centralized location then some sort of basic serial console setup is fine. This approach can (and does) scale to large numbers of devices in the case of network hardware, but this is in some part a function of history.

If the intention is for many hundreds (or thousands) of devices deployed on a wide basis then working out some kind of mechanism for dynamically pushing configurations, firmware, etc can be a big win. Using DHCP to pull not only an address but also some measure of custom configuration (link to central config server, pull down image based on serial/MAC, etc) has been an approach that has been used to good effect by some vendors.

share|improve this answer
add comment

I've dealt with some specialty devices that have a configuration application, that uses broadcast messages to locate devices on the local subnet.

There are a number of layer-2 hacks you can use to implement this. However, it is required that you have a computer on the same subnet for initial configuration. This may or may not be tenable for your application.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.