XCP-ng
    • Categories
    • Recent
    • Tags
    • Popular
    • Users
    • Groups
    • Register
    • Login

    Gather CPU utilization of host as variable for prometheus exporter

    Scheduled Pinned Locked Moved Compute
    13 Posts 4 Posters 3.4k Views 5 Watching
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • P Offline
      petr.bena @olivierlambert
      last edited by

      @olivierlambert but does that really export CPU utilization? From what I saw it only shows memory utilization. Also how does orchestra get these metrics? There must be some low level way to accomplish this, or is xsconsole also using the API for that?

      1 Reply Last reply Reply Quote 0
      • olivierlambertO Online
        olivierlambert Vates 🪐 Co-Founder CEO
        last edited by

        Everything is already exported via RRDs (see https://wiki.xenproject.org/wiki/XAPI_RRDs). That's what we use in XO.

        But this is far from being nice because:

        • Those RRDs are encapsulated in XML, which is costly to parse
        • XAPI itself is hammering Xen with requests every 5 secs

        So ideally, doing a metrics redesign from Xen to XAPI first, will be better.

        1 Reply Last reply Reply Quote 0
        • P Offline
          petr.bena
          last edited by

          how does API get these metrics?

          1 Reply Last reply Reply Quote 0
          • olivierlambertO Online
            olivierlambert Vates 🪐 Co-Founder CEO
            last edited by

            You mean how XAPI is fetching these metrics?

            P 1 Reply Last reply Reply Quote 0
            • P Offline
              petr.bena @olivierlambert
              last edited by

              @olivierlambert yes

              1 Reply Last reply Reply Quote 0
              • P Offline
                petr.bena
                last edited by

                there must be some low level way to do that from dom0

                1 Reply Last reply Reply Quote 0
                • olivierlambertO Online
                  olivierlambert Vates 🪐 Co-Founder CEO
                  last edited by

                  That's the plan, yes, there's some Xen stats backend, but it requires some work to modify Xen for a better system.

                  1 Reply Last reply Reply Quote 0
                  • M Offline
                    MikeDombo
                    last edited by

                    I just installed XCP-ng on a new server and wanted Prometheus metrics too, so I wrote this: https://github.com/MikeDombo/xen-exporter.

                    It is a simple Python script (which can run in Docker) that collects the metrics from XCP-ng's RRD API. These are the same metrics that you'd see in Xen Orchestra or in the XenServer Console application. In addition to the exporter, I've also created a Grafana dashboard to visualize these metrics too.

                    Example dashboard

                    1 Reply Last reply Reply Quote 1
                    • olivierlambertO Online
                      olivierlambert Vates 🪐 Co-Founder CEO
                      last edited by

                      Ah nice!

                      Ideally, we'd like to avoid to use the complicated XAPI RRD + XML system. See https://xcp-ng.org/blog/2022/05/25/fetching-metrics-in-xen-why-the-method-matters/

                      Thanks a lot to bring this, I'll promote it!

                      S 1 Reply Last reply Reply Quote 1
                      • S Offline
                        ScarfAntennae @olivierlambert
                        last edited by

                        I was about to link Mike's work here, but wanted to say how we desperately need a metrics system that would be exposed from a centralized location. There's no way in hell I'm giving a docker host credential access to the hypervisor it resides on (and all the other hypervisors) - on the other hand, if the metrics were exposed only via XO and limited in terms of where these metrics could be accessed from, now that would be a sold implementation.

                        S 1 Reply Last reply Reply Quote 0
                        • S Offline
                          ScarfAntennae @ScarfAntennae
                          last edited by

                          Related: https://xcp-ng.org/forum/topic/9160/xen-orchestra-prometheus-backup-metrics/10

                          1 Reply Last reply Reply Quote 0
                          • First post
                            Last post