SNMP Monitoring

Discussion in 'Proxmox VE: Networking and Firewall' started by John Allison, Mar 21, 2018.

  1. John Allison

    John Allison New Member

    Joined:
    Feb 1, 2018
    Messages:
    9
    Likes Received:
    0
    I'm currently evaluating proxmox and so far am really impressed. One thing that seems strangely missing tho is performance monitoring.

    I've installed snmp on our 3 test nodes and can collect node related data, but it would be fantastic if i could collect data relating to the virtual machines running on the node.

    I found a livirt-snmp project on 'wiki.libvirt.org/page/Libvirt-snmp' but cant get it to work.

    Any ideas or suggestions on how i can monitor virtual machines running on Proxmox via SNMP?

    Also, anyone think it would be a cool idea to have extensive vm monitoring built in?
     
  2. Alwin

    Alwin Proxmox Staff Member
    Staff Member

    Joined:
    Aug 1, 2017
    Messages:
    1,675
    Likes Received:
    144
    We are not using libvirt.

    Why SNMP, modern monitoring systems usually have their own agent or checks to get data. You could use the rrd data already gathered and send it to influxdb or graphite. https://pve.proxmox.com/wiki/External_Metric_Server
     
    Stop hovering to collapse... Click to collapse... Hover to expand... Click to expand...
  3. guletz

    guletz Active Member

    Joined:
    Apr 19, 2017
    Messages:
    641
    Likes Received:
    95
    .... or you can install snmp on any VM you want, and then you can get what you want.
    HINT: librenms is a great tool, and they have VM image, ready to use!
     
  4. John Allison

    John Allison New Member

    Joined:
    Feb 1, 2018
    Messages:
    9
    Likes Received:
    0
    Well SNMP is pretty standard, and we currently use PRTG. I've just started looking at the external metric server stuff, but seems a bit overkill.
     
  5. John Allison

    John Allison New Member

    Joined:
    Feb 1, 2018
    Messages:
    9
    Likes Received:
    0
    Yeah but we fire up / destroy VMs on a regular basis (software dev company) and so I want to be able to monitor the VM's via the hypervisor and not from each VM.
     
  6. guletz

    guletz Active Member

    Joined:
    Apr 19, 2017
    Messages:
    641
    Likes Received:
    95
    .... PRTG? Maybe MRTG ?
     
  7. guletz

    guletz Active Member

    Joined:
    Apr 19, 2017
    Messages:
    641
    Likes Received:
    95
    Ansible can install / config snmp on any(linux) VM!
     
  8. John Allison

    John Allison New Member

    Joined:
    Feb 1, 2018
    Messages:
    9
    Likes Received:
    0
    True but would that not just give me stats from the VM's perspective, and not show me the reality of things like actual CPU use

    Its definitely PRTG, not sure what MRTG is?
     
  9. John Allison

    John Allison New Member

    Joined:
    Feb 1, 2018
    Messages:
    9
    Likes Received:
    0
    Is there any information on how to quickly setup something like Proxmox monitoring via Graphite? I found a page in the help files, but it doesnt say much...
     
  10. guletz

    guletz Active Member

    Joined:
    Apr 19, 2017
    Messages:
    641
    Likes Received:
    95
    You are wrong!
     
  11. John Allison

    John Allison New Member

    Joined:
    Feb 1, 2018
    Messages:
    9
    Likes Received:
    0
    Are you sure?

    If you monitor the physical CPU's on a hypervisor with 40 cores running 10 VM's each with different CPU allocations. You will get stats showing how the hypervisor is handling the load but you wont be able to see what percentage of CPU time is being allocated to each VM. Now if you install SNMP directly on each VM, then the stats you get will only reflect what that VM see's of the world.It wont know about any of the other VM's or the Hypervisor. So it might only have 1 CPU allocated which could be running at 100% utilization but you cannot see what percentage of physical CPU the VM is using, only the Hypervisor could tell you this. Which is why im asking if theres something like the livirt-snmp project which might work with proxmox.

    I suggest you google livirt-snmp and then ask yourself why this might be needed. :)

    You might also want to take a quick look at www.logicmonitor.com/blog/a-tale-of-two-metrics-windows-cpu-or-vcenter-vm-cpu/ - there's many other similar articles too which talk about how VM's measure CPU :)
     
  12. guletz

    guletz Active Member

    Joined:
    Apr 19, 2017
    Messages:
    641
    Likes Received:
    95
    If you have start from the beginning with this details, I have understood what you need. Now it is clear what you want to get and yoy are wright!
    So sorry for my fault ..... ;)
     
  13. Nhoague

    Nhoague Member

    Joined:
    Sep 29, 2012
    Messages:
    68
    Likes Received:
    0
    I am interested in something very similar. I also use PRTG to monitor my entire network including the virtual hosts, but I want to find a way to monitor the customer status. To make sure in a machine hasn’t lost quorum. Any ideas?
     
  14. nicolaiB

    nicolaiB Member
    Proxmox VE Subscriber

    Joined:
    Dec 7, 2016
    Messages:
    34
    Likes Received:
    2
    Dont't know how to integrate custom checks into PRTG, but there are some in the Icinga / Nagios universe which you might want to have a look at. I've written for example a check which queries the PVE API for host and vm metrics including the cluster status.:

    https://github.com/nbuchwitz/check_pve
     
  15. John Allison

    John Allison New Member

    Joined:
    Feb 1, 2018
    Messages:
    9
    Likes Received:
    0
    Should be easy enough, just run a shell script from the PRTG server to view the output of pvecm status command.
     
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice