I’m working on network management requirements for several customers and keep running into the same requirements. One of the key requirements is that of event reporting. Events typically take two forms: syslog and SNMP traps. In both cases, they are asynchronous notification of something happening on the network. Sometimes the notification is not important, like an unimportant edge link going up/down. But other notifications are critical to early awareness of significant network problems. The problem with event reporting is that the event logs can often be quite long and identifying the one or two key things out of a log that is thousands of lines long is literally like finding the needle in the haystack.
At some sites, the network staff watches the tail of the log file, often ‘tail -f ‘ on a Unix/Linux log server. If you’re watching the log file when something important occurs and you happen to notice, then that’s good. If you’re not watching or don’t notice it, then that’s not so good.
Other sites get a daily report on network events. The summary report shows a count of all the events for the day in the upper section. The lower section shows the count of the events, the device that sent the event, and the event name (see below). A quick scan of the upper section shows what events have happened and the lower section lets you identify which devices you should investigate, based on the criticality of the message, or perhaps the volume of the message. In the example below, device test1.com and test4.com are experiencing a large number of EIGRP neighbor change events, which would indicate a link problem on each of these devices. Together, they comprise the majority of the EIGRP neighbor change events.
Summary of GNS Cisco syslog Messages on Wed Jan 17 23:59:00 EST 2007 Cisco Messages: 437 DUAL-5-NBRCHANGE 353 LINEPROTO-5-UPDOWN 114 CRYPTO-6-IKMP_MODE_FAILURE ... Messages sorted by frequency and source device: 346 test1.com DUAL-5-NBRCHANGE 114 test2.com CRYPTO-6-IKMP_MODE_FAILURE 84 test3.com LINEPROTO-5-UPDOWN Tunnel119 67 test4.com DUAL-5-NBRCHANGE
Other events, like fan or power supply failures, environmental alerts, link up/down on core links, unplanned device reboots, and security notifications are important for many organizations. As I noted above, watching a log file isn’t very valuable.
NetMRI has an add-on NetMRI Event Analysis appliance that can collect events (including SNMP traps) and has an exception notification system that creates NetMRI issues whenever an important event is detected. The underlying engine is Splunk, so it has excellent pattern matching and filtering capabilities. A great addition in the NetMRI implementation is the ability to use device and interface grouping to filter events. So an interface up/down event on a core link generates an issue while the same event on an unimportant edge interface does not generate an issue.
Having issues generated allows the network staff to reduce the number of screens that need to be watched or monitored. NetMRI also allows important events to be emailed, so a message to your phone can alert you to an important network event.
By integrating event log analysis with other network analysis, tools like NetMRI can take a big load off the network staff and that’s what’s important to a smoothly running network.
-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