6 Replies Latest reply on Apr 23, 2009 7:01 PM by rgward

    Monitor Polling Engines feature requests

    rgward

      1. Provide ability to balance pollers by custom property, nodeid in addiiton to the present nodename and ip address.

      2. Provide a status area on the screen showing, at a minimum, a running count of nodes assigned to each poller as load balancing changes are made.  It would be even better to show interfaces and volumes as well.  Today, the only way of knowing how many nodes have been assigned to a poller is to restart the NPM services to get the numbers.  If they aren't distributed evenly or where you want them to be you have to stop the services again and repeat the cycle.  Very much a pain and time consuming especially since our NPM takes about 10 minutes to shutdown.

      3.  As an extention of #2, it would be great to have the Recommended Maximum Node and Interface Status Polls and Maximum Statistics Collection values calculated and posted for each poller as is calculated by the Poles Per Second tuning tool.  This would also allow the ability to adjust the Polls Per Second values at the same time without having to start the pollers again only to have to stop/restart the services again (taking another outage) to adust them.

        • Re: Monitor Polling Engines feature requests
          denny.lecompte

          We are planning to add load-balancing to our poller.  The working plan is more dynamic balancing so that, for the most part, specific nodes are not assigned to specific pollers.  The most available poller at any moment would do the poll.  We would leave the option for affiliating specific nodes to specific pollers, but that would probably we the exception.

          Any comments on the plan?

            • Re: Monitor Polling Engines feature requests
              brian_duvall

              We tend to prefer statically setting the poller for one main reason.  If you are a global company and have pollers spread across the globe, if nodes are constantly moving around from poller to poller, response time metrics will be thrown off.  For instance if I have a poller in Maryland and a poller in Singapore I want to poll all my US nodes from Maryland and all my Asia nodes from Singapore.  If dynamically they move...response time reports will look like something bad has happened on the network when really its just getting polled from farther away.

              • Re: Monitor Polling Engines feature requests
                rgward

                I concur with Brian.  At least at the moment, I'm not favorable to the "most available poller at any moment" concept not knowing or having control over which poller polls what devices.   Our pollers are not geographically dispersed but devices are assigned to pollers based on Organization.  I like the dynamic capability but would prefer to also have the flexibility to configure "poller profiles" based on one or more text strings using nodename, DNS name, custom property, subnet or supernet.  For instance, our network devices begin with a 3 or 4 letter organization identifier.  Our server names don't exactly follow the same convention but the organization identfier is embedded in the nodename.  It would be nice to be able to configure pollerA to poll all devices with subnet 10.10.x.x and nodenames with a 'BBB' prefix and pollerB to poll all devices with subnet 10.20.x.x, nodenames with a 'CCC' prefix and 'CCC' as an embedded text string.