Uploaded image for project: 'OpenShift Virtualization'
  1. OpenShift Virtualization
  2. CNV-61263

[CLOSED LOOP for] CPU hotplug logic still going over the limits

XMLWordPrintable

    • Icon: Closed Loop Closed Loop
    • Resolution: Done
    • Icon: Critical Critical
    • None
    • None
    • CNV Virt-Cluster
    • None
    • Future Sustainability
    • False
    • Hide

      None

      Show
      None
    • False
    • None
    • CNV Virt-Cluster Sprint 273
    • None

      Description of problem:

      The issue we discussed at length in https://issues.redhat.com/browse/CNV-48124 is still not fixed. The RHEL side to automatically enable eim when the count is high is insufficient.
      
      The hotplug logic still does a x4 on the CPU count to allow for hoplug. Now instead of breaking at 255 with eim not automatically enabled, it can break at 710 (the hard limit of the machine type).
      
      
      

      Version-Release number of selected component (if applicable):

      4.18.0 (I can reproduce), 4.16 (customer)

      How reproducible:

      Always

      Steps to Reproduce:

      spec:
        domain:
          cpu:
            cores: 216
            sockets: 1
            threads: 1 

      Actual results:

      That same logic will do 216x4, and go over the limits.
      
      {"component":"virt-launcher","level":"error","msg":"unsupported configuration: Maximum CPUs greater than specified machine type limit 710","pos":"qemuValidateDomainVCpuTopology:739","subcomponent":"libvirt","thread":"27","timestamp":"2025-03-02T23:26:31.090000Z"}
      
      The XML generated by virt-launcher has tons of vCPUs:
      
      {"component":"virt-launcher","level":"info","msg":"\t\t<vcpu id=\"863\" enabled=\"no\" hotpluggable=\"yes\"></vcpu>","subcomponent":"libvirt","timestamp":"2025-03-02T23:26:31.078974Z"}

      Expected results:

      Fix the hotplug logic to not go over limits

      Additional info:

      Its the same problem as the previous issue, but now at 710 instead of 255 (eim), the libvirt fix just moved the goalpost. 
      
      See https://issues.redhat.com/browse/CNV-48124?focusedId=25546000&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-25546000.

              kbidarka@redhat.com Kedar Bidarkar
              dagur@redhat.com Daniel Gur
              Sibo Wang Sibo Wang
              Votes:
              0 Vote for this issue
              Watchers:
              5 Start watching this issue

                Created:
                Updated:
                Resolved: