Networks use the same devices in many different roles, so configuration validation based on device type isn’t sufficient. There are some configuration things that can be checked across all the devices of a certain type in your network, such as checking the syslog and NTP configuration, or verifying that all the vtys in Cisco routers and switches have an appropriate ACL applied. But when it comes to other configuration factors like routing protocols or interface configuration specifics, there are other factors that need to be incorporated into deciding which configuration template to check.
Grouping devices by role makes it possible to create configuration policy validation checks that work across large numbers of devices. For example, checking all the OSPF backbone router configurations or checking the EIGRP stub router configurations needs to be done on only a subset of the devices in the network. Similarly, grouping of interfaces by functional role, like whether it is an uplink or an edge port, can allow configuration template comparisons against large numbers of interfaces.
The key to device and interface grouping is having an automated method for doing the grouping. Sure, you can do it manually, and that’s how a lot of systems force you to do it. Let’s say your network contains 600 network devices. Are you going to manually classify each device into the proper groups and maintain those groups over time? Keep in mind that if a device’s role is modified (e.g., it now does QoS classification and marking), you’ll need to add it to the QoS group. A device can be a member of multiple groups, so adding a new device may mean identifying which groups to which it should belong and editing those group lists. Manual processes like this are why network management systems turn into shelfware – the process is too intensive to be followed consistently.
If manual processes don’t work, what will? A rich grouping language must be used to classify devices and interfaces into different groups. A device that has OSPF, QoS, BGP, and is a core device may need to be a member of four groups. The device configuration, addressing, and neighbors, may need to be examined in order to classify the device. Interface classification follows similar rules in that its addressing, neighbors, or trunking may determine whether it is an edge interface or an infrastructure interconnection interface. Should the interface be configured with portfast and bpdu-guard? Should it be set to ‘duplex auto’, or fixed duplex, and what factors determine the setting?
Once the device and interface grouping exists, the configuration policy checking system needs the ability to apply group-specific configuration template analysis. Of course, your network needs to be designed in a way that results in a minimal number of overall configurations or the number of groups and templates will explode and become unmanagable.
Let me know what you’ve found that works well in your environment for configuration validation and how it handles differences in device and interface roles.
-Terry
_____________________________________________________________________________________________
Re-posted with Permission
NetCraftsmen would like to acknowledge Infoblox for their permission to re-post this article which originally appeared in the Applied Infrastructure blog under http://www.infoblox.com/en/communities/blogs.html
