This discussion has been locked. The information referenced herein may be inaccurate due to age, software updates, or external references.
You can no longer post new replies to this discussion. If you have a similar question you can start a new discussion in this forum.

creating alerts in NPM for 100% packet loss

I am trying to create an alert to trigger on 100% packet loss and I have no idea what the criteria should be to make it work.

I have the device names, but what fields/events do I need to make it work?

  • Isn't 100% packet loss a Node down alert?

  • Well,

    Not sure I follow. Are you saying that if the node goes down then the 100% packet loss alarm will trigger? What if the node is up, will it still trigger if packet loss is 100%?

    Rebecca Murphy | Systems Engineer I | MISM| Service Delivery Systems - OSS/NMS

    | 314-543-5625 (O) | 314-452-4968 (C)

    12405 Powerscourt Drive, St. Louis, MO 63131

  • What i'm saying is if packet loss is 100% the node-down alarm will trigger

    There is no way for NPM to mark a node as being up with 100% packet loss

    If every packet NPM sends at a node is lost, how will it consider it as being 'up'?

    the NPM node ICMP status poller pings a node each Default Node Poll Interval to determine its status AND its packet loss.

    If a node does not reply in ICMP Timeout then it goes into a rapid-ping mode and pings it again (three times) . if it gets no reply then the node is flagged as being down.

    -> packet loss is now 100%

    Note: the packet loss probably did not grow slowly to 100%, but instead it went from 0 to 100% in about 120 + 4*2.5s = 130seconds

    when a node starts to respond again each ping-cycle the packet loss falls by (I think) ~20% each time, and after 4 ping cycles the node is considered up.

  • out phone switches reply to ICMP

  • The field you're looking for is called "Percent Loss" under the Node. You can set it to anything you'd like, including 100%.

    You can also include this under the reset condition, allowing you to (for example) alert you of the issue for only 100% packet loss, but only reset the condition if it improves to less than 10% packet loss, not when it goes back to 90%.

  • I asked the question -

    Will there ever be a time where packet loss is 100% other than when the node is down?

    Here is what the response was -

    It is possible.

    In past history, I have seen packet loss go to 100% when servers have multiple hung processes and possibly very high CPU load.

  • I guess in that case the server can still be contacted through WMI but pings are failing.

    However I think it would be extremely rare to have 100% packet loss and the node not be marked as down. I do have an alert for packet loss on all my WAN links. It will send an email if it detects packet loss over 10% for a continuous 15 mins. This will filter out if the link just happens to be really busy and pings get dropped.

  • FormerMember
    0 FormerMember

    I have seen a couple of times where a device that I know is offline and should be shown as "Down" is shown to have "High Packet Loss".  It had actually shown that the HPL was for a duration of about 10 minutes, but the node was never listed as down.

  • I've also had this happen. Node should have been listed as down and it took forever to finally alert. Had another with like 30ms latency and never got an alert.