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.

NPM 10.1 UCS Monitoring - Adding Blades, SNMP Possible?

Hi,

great new feature! This bings big value to this product!
Since the documentation about monitoring UCS is not very extensive....

I have to add the Blades of my UCS, but the SNMP credentials of the fabrics doesn't seem to work here.
Do you know if I can query them via SNMP, or are they just pingable?

thanks,
Herwig

  • Thank you for the feedback!

    So you need to add the VIP or Primary Fabric Interconnect IP to Orion to get the UCS data with SNMP and the UCS API credentials.  To monitor whats on the blades, you will need to add them like you do any other node into Orion, very similar to what you do with VMWare ESX in Orion

  • I just set this up on my UCS.  As bshopp mentioned, don't use the individual FI IP.  I set my unit up with SNMP v3 on Friday and it appears to work fine.  SNMP is off by default, so you'll need to go into UCSM to enable it.

    In case you're interested, the SNMP location doesn't appear to be configurable and that's a bit of a drag.  The SNMP contact is actually pulled from the Call Home configuration (strange, eh?).

    After you get it working, would you mind adding a virtual ethernet interface to your monitoring?  I'm seeing 1000's of percent utilization on those and I want to see if it's an issue for everyone.

  • Yes I added the Primary Fabric Interconnect VIP and the physical nodes via IP to Orion without problems.

    The issue I have with adding the blades is, I am not able to find a configuration option for SNMP settings in the UCSM for each server/blade - and the credentials I used for the Interconnects do not work for the blades...

    This is the global UCSM configuration setting for SNMP, with this settings the Interconnects (VIP and physical) are SNMP readable:

     

    But when I try to add the several UCS Blades (in my case 4) with the same credentials --> failed:

     

    So the question is: Are UCS Blades only pingable from outside, or is there an option to enable SNMP get for them?
    I think the Blades report CPU and Memory stats- monitoring them is also important...

    By the way: I also miss CPU and Memory Gauges and Graphs for the UCS - I see the resources in "Customize View", but the views do not appear on the UCS Node Detail Page:



  • After you get it working, would you mind adding a virtual ethernet interface to your monitoring?  I'm seeing 1000's of percent utilization on those and I want to see if it's an issue for everyone.




    I added a screenshot to your initial posting, don't recognise your issue in my system:
    ">

  • Are you running a direct OS on the server or some sort of virtualization like VMWare?

    If just an OS I don't think there is any way for UCS to know what the IP Address is in order to link it up with the server.  

    Can you see the IP Address of the server OS in UCS Manager or no?

  • There isn't much that you might be able to get from the blade itself other than maybe model information if that access were available.  I don't believe UCS allows direct SNMP access to the blades.  The blade IPs are for the BMC (what you'd KVM into).  I checked the SNMP guide for UCS and see no reference for the BMC.  CPU and memory would be retrieved from the guest OS.  It would be impossible for the BMC to have that level of introspection.

  • I do believe you are correct here.  How we can do this with VMWare ESX is you must have the VMWare client tools loaded to tell ESX what the OS IP is.  I don't believe UCS has anything like this.

  • What you can do as well is add the OS on the blades as a node like you have and put them all in a Dynamic Service group with the UCS chassis and you can get rollup status there.



  • There isn't much that you might be able to get from the blade itself other than maybe model information if that access were available.  I don't believe UCS allows direct SNMP access to the blades.  The blade IPs are for the BMC (what you'd KVM into).  I checked the SNMP guide for UCS and see no reference for the BMC.  CPU and memory would be retrieved from the guest OS.  It would be impossible for the BMC to have that level of introspection.




    Thanks both of you for all of your inputs.

    Yes, what I found out you are right- because the several blades have IP addresses from a mgmt pool....
    My wrong assumtion was that its possible to query them via SNMP from outside the system to get additional Info directly from the blade hardware environment like CPU/Memory/Fans a.s.o. - seems like this is indeed not possible.

    Ok, I can monitor CPU/Memory of a blade on a VMWare level, but I still miss some important information from the UCS itself like Power Supplies and Fan stats which are of couse important to monitor....?

  • If you drill down on the blade I believe, you can get the status of PSU/Fans