4 Replies Latest reply on Jul 27, 2010 3:59 AM by mezdem

    Node polling vrs. Interface polling

      I have a Cisco 37xx Catalyst switch (actually, many such switches) that has approximately 200 interfaces. Most of these interfaces have PC's or Printers plugged in to them. I am NOT interested in these ports. Instead, I am only interested in the SMLT ports that connect the switch to the main distribution frame. Typically, these switches look like the following screen-shot in Orion's System Manager.

      I have configured Custom Pollers to track, the following Cisco MLT Errors:

      Carrier Transitions

      Aborted Input Packets

      CRC Errors

      Misaligned Frame Errors

      Ignored Inbound Packets

      Inbound Queue Drops

      Outbound Queue Drops

      Interface Resets

      (Additionally I am monitoring several MIB II variable)

      For the sake of discussion, let's only examine the CRC Errors. I have created a custom poller (See, below).

      I have also created a chart to display the data collected from this poller (see below).

      As can be immediately seen, even though I have configured two of the interfaces on this switch, ALL of the interfaces are being polled. Of course, for purposes of presentation, I do have the ability to select ONLY the two interfaces that I am interested in to display (see, below).

      Before selection:

      After selection:

      The chart then looks like this:

      My first point is to question, if only two interfaces are configured for a specific node, why does the poller poll ALL interfaces? This seems to be a significant squandering of resources.

      Of course, it can be pointed out that instead of configuring the node polling type, the overhead can be managed by configuring an interface polling type. Although this may appear to be true on a superficial level, it really is not practical. In this scenario, the number of configured pollers dramatically rises, especially considering that more than a dozen of the switches that I wish to monitor in this manner may have 10 or more interfaces that I am interested in amongst the roughly 200 interfaces available. All in all, where eight Node pollers were initially configured, I estimate that 900 interface pollers would be required to perform this function. This just seems to be an administrative nightmare.

      There is also a secondary issue I would like to address.

      In addition to the charting, I also will be creating thresholds and alerts for all of the eight variables I wish to poll. Immediately it is apparent that instead of the eight alerts that would be created when Node polling is used, the use of interface polling would also require roughly 900 separate alerts. But let's leave that point aside, for the time being, so that we can pursue the problems at hand using Node Polling.

      Using alerts with threshold that are dependent upon Node Polling creates a circumstance where it is likely that an alert will be raised from one of the PC or printer ports that we are not interested in.

      Has anyone else encountered any similar issues when monitoring switch interfaces (or router interfaces, for that matter).



        • Re: Node polling vrs. Interface polling

          Hi , sorry i cant answer your question,, but could you please tell me how you created your custom chart?

            • Re: Node polling vrs. Interface polling


              I will be happy to provide my method, but I can't promise exactly when I can provide it.  Reasonably, it should be before Wednesday of next week (JUly 28).

              • Re: Node polling vrs. Interface polling

                Creating a Custom Chart in Orion


                Before even beginning the process of creating a custom chart in Orion, one needs to determine the managed view of the node from which the chart will be created.  For instance, we monitor several types of switches, Nortel 8600 series switches, Nortel 470 series switches, Nortel 4550 series switches, Nortel 5500 series switches and Cisco Catalyst 37xx series switches.  When these devices are discovered, Orion recognizes them as falling under three device type classes: Cisco Catalyst 37xx, Rapid City Communication (the 8600 series) and Synoptics (the 470's, the 4550's and the 5500's).


                When Orion is first installed, if one looks at Admin ==> Views ==> Views by device type, these three device type classes will be set to (default).  Changing these setting is, in my view, an important, but not required step that will make the organization of one's Orion Web presentation significantly more meaningful.


                The next step is to create a Managed View for the device(s) that one wishes to custom chart.  Continuing with our example of switches, and for purposes of discussion, let's use the Cisco Catalyst 37xx switches for our example.  Creating the  Managed View by navigating Select Admin ==> Views ==> Managed views. 



                Then select "Add".



                Use a name that makes sense to you and your organization and choose "Node Details" for the Type of View then click Submit.



                Every organization has their preferences on  what a "basic" view should look like.  My own preference for switches is the following.




                Once the "basic" view has been created, click "Done" and navigate to Admin ==> Views ==> Views by Device Type.




                Now, when you select the drop down menu for the device you wish to create the charts for, the name of Managed View that you just created will be available as a selection.  Choose it and Submit.


                We are now - FINALLY - ready to begin the process of setting up the poller(s) and creating the chart(s).  On the Orion Server, open the Universal Device Poller and select "New Universal Device Poller" (or select an existing poller, then choose Duplicate poller.  If you have chosen "New Universal Device Poller", your screen will look like:




                For organizational purposes, change the Default Group to a name of your choosing, then select "Browse MIB Tree".




                Navigate to the OID that you wish to gather data for your chart, then select the Vendor and a test device.  Select "Test" and if a valid OID had been chosen, you should see the "Test Successful" prompt.  Click the Select button.




                Ensure that "Counter", "GET TABLE" and "Node" are selected.  (This is the condition for creating charts that are based on counter OID's.  The Counter is actually an accumulation of data on an ongoing basis or a running total since the last time the counter was reset).  With these settings, Orion will perform a differential, comparing the most recent poll to the previous poll and calculating the difference. 


                Click the "Next>" button.




                Select the nodes that you wish to monitor, then select "Next>".




                In most instances (but not in the instance behind the reason of my initial post) the "Use interface names from Orion" is sufficient.  Again, click "Next>".




                Ensure that "Yes" is selected under "Do you want to display results on your Orion Website?"  Drag the Scrollbar of views all the way to the bottom.  Note that the Managed View that was created earlier is the bottom selection.  It wouldn't have been there had we not performed the preliminaries,  If you want the poller data to be presented as a Chart, select it.  If you want the data in Table form, select table.  You can also select both.  Click the "Finish" button.


                Back on your browser, select Admin ==> Views ==> Managed views.  Select the view that you created and select "edit".  Note that the name of the poller you just created is at the bottom of the left column.




                Now is a good time to designate the "View Limitation" with the type of device you are polling.


                Next, select the "Preview" button to see the results.


                It will look similar to the following.




                You can customize the Title and other items by selecting the "Edit" button.

                Upon reviewing this, I see a few areas where I could have been more clear.  If you need help, let me know.