Is there a right way to implement NPM?

In all the years I've deployed, updated, refreshed Solarwinds Orion using NPM I've always used it for networkoing gear only. Presently I'm on a contract with a company that sort of blends their server and network staff and they are using NPM to monitor all the NIC's on their servers along with the network gear, Not to be a Network snob, it seems a little off the mark to use a tool like NPM to monitor NICs (physical, virtual and vether's). If you are doing NIC teaming does an alert go out but the other NIC is areally working?? if you move a VM are those vnics really down?

Looking for some opinions, feedback or if you've seen NPM used like this before.

Thanks

Find more posts tagged with

network_monitoring

npm

Comments

superfly

We monitor servers with NPM. In the case of teaming, you will see 3 interfaces. NIC 1 & 2 plus the teaming interfaces. It would be handy to know if one of the NIC's has gone down as now you've lost redundancy. You could alternatively monitor the interfaces on the switch to see if they are up/down. We monitor both ends but you could achieve the same with just monitoring one end.

witrick

Okay that's one way to monitor things. My wonder is if a NIC goes down is it a systems problem or could it be a network problem? Do you get two alerts, one for the NIC and one for the network port?

LeBeauUK

What you can do is make NICs on the server Children of the Network Switch, so if the switch goes to the state of Down Node, the NICs on the Server becomes classed as unreachable nodes. But if the Server goes down the NICs on the Server / Server itself becomes classed as a node down.

witrick

Hmmm sounds like it works but there are sooooo many servers. What happens when the VM folks vmotion a bunch of servers? Trying to remember my VM stuff, cause if the vnic drops on one side picks up on the other a physical switch might report down when it's a normal operation.

superfly

Yes you would get an alert for the network switch port and the NIC on the server if both are being monitored. But in the case of a VM, there is no physical network switch port for it. The hosts are the only ones that have the physical connection.

superfly

It would be easier to just monitor one or the other. As when a server NIC goes down, the network port would go down as well.

NightAngel07

Tell me more about this and how to do it... Is it possible to do this for switches that sit behind other switches so I don't get 1000 alerts?

bobmarley

We monitor both the NIC as well as the Switchport, but we don't alert on when a NIC goes up/down. Only if the NIC utilization goes too high for an extended period. There has been a few times when someone decided to copy huge amounts of data from one place to another and filled up a WAN link. The Network Top 10 report will usually point the machine out right away as long as you are monitoring all the Servers NIC cards.

adam.beedell

Just +1 to "I've used NPM like this before"

NPM (in conjunction with SAM and maybe VMAN) is pretty good for servers tbh. I've always split out the alerting for network (cisco etc) stuff vs server (windows/linux/etc) stuff though, as the approaches to fixing it tend to be very different

Could well be worth splitting off a "teaming issue" alert for that scenario if the behaviour you're trying to elicit is different