Uploaded image for project: 'Red Hat OpenStack Services on OpenShift'
  1. Red Hat OpenStack Services on OpenShift
  2. OSPRH-6665

EDPM fails to set kernel args in deployed compute in RHEL 9.4 [Beta Blocker]

XMLWordPrintable

    • False
    • Hide

      None

      Show
      None
    • False
    • ?
    • No Docs Impact
    • Committed
    • ?
    • None
    • Release Note Not Required
    • Approved
    • Critical

      When deploying dataplane it seems like kernel args are not properly set on compute which causes the OVS to fail with DPDK configuration because it is unable to find the hugepages that were expected to be ready.

      [root@tigon01 ~]# cat /proc/cmdline 
      BOOT_IMAGE=(hd0,gpt3)/vmlinuz-5.14.0-427.13.1.el9_4.x86_64 root=LABEL=img-rootfs ro console=tty0 console=ttyS0,115200n8 no_timer_check console=ttyS0,115200 no_timer_check memtest=0 boot=LABEL=mkfs_boot skew_tick=1 tsc=reliable rcupdate.rcu_normal_after_boot=1 nohz=on nohz_full=2-19,22-39 rcu_nocbs=2-19,22-39 tuned.non_isolcpus=00300003 intel_pstate=passive nosoftlockup
       

      while the configuration is:

      edpm_kernel_args: "default_hugepagesz=1GB hugepagesz=1G hugepages=64 iommu=pt intel_iommu=on tsx=off isolcpus=2-19,22-39" 

      this causes the "configure-ovs-dpdk-edpm-deployment-openstack-edpm" pod to fail due to ovs service failing to start (on the first time) but it finishes on the second try although the ovs service fails to start and then fails in the network configuration stage due to ovs not started.

      ovs service status:

      [root@tigon01 ~]# systemctl status openvswitch
      ○ openvswitch.service - Open vSwitch
           Loaded: loaded (/usr/lib/systemd/system/openvswitch.service; enabled; preset: disabled)
           Active: inactive (dead)Apr 30 05:56:33 tigon01 systemd[1]: Dependency failed for Open vSwitch.
      Apr 30 05:56:33 tigon01 systemd[1]: openvswitch.service: Job openvswitch.service/start failed with result 'dependency'. 

      ovs log:

      2024-04-30T09:56:35.402Z|00019|dpdk|ERR|EAL: Cannot get hugepage information.
      2024-04-30T09:56:35.402Z|00020|dpdk|EMER|Unable to initialize DPDK: Permission denied
      2024-04-30T09:56:35.404Z|00002|daemon_unix|ERR|fork child died before signaling startup (killed (Aborted))
      2024-04-30T09:56:35.404Z|00003|daemon_unix|EMER|could not detach from foreground session 

      this happens although this configuration exists in the grub.cfg:

      if [ -z "${kernelopts}" ]; then
        set kernelopts="root=LABEL=img-rootfs ro console=tty0 console=ttyS0,115200n8 no_timer_check   default_hugepagesz=1GB hugepagesz=1G hugepages=64 iommu=pt intel_iommu=on tsx=off isolcpus=2-19,22-39   console=ttyS0,115200 no_timer_check memtest=0 boot=LABEL=mkfs_boot"
      fi
       

      inside the /etc/default/grub I saw this is set:

      GRUB_EDPM_KERNEL_ARGS=" default_hugepagesz=1GB hugepagesz=1G hugepages=64 iommu=pt intel_iommu=on tsx=off isolcpus=2-19,22-39 "
      GRUB_CMDLINE_LINUX="${GRUB_CMDLINE_LINUX:+$GRUB_CMDLINE_LINUX }${GRUB_EDPM_KERNEL_ARGS}"
      
      

      when running grubby I can see that the kargs are not set

      [root@tigon01 ~]# cat /boot/loader/entries/65566f5501474383980967083db85801-5.14.0-427.13.1.el9_4.x86_64.conf 
      title Red Hat Enterprise Linux (5.14.0-427.13.1.el9_4.x86_64) 9.4 (Plow)
      version 5.14.0-427.13.1.el9_4.x86_64
      linux /vmlinuz-5.14.0-427.13.1.el9_4.x86_64
      initrd /initramfs-5.14.0-427.13.1.el9_4.x86_64.img $tuned_initrd
      options root=LABEL=img-rootfs ro console=tty0 console=ttyS0,115200n8 no_timer_check   console=ttyS0,115200 no_timer_check memtest=0 boot=LABEL=mkfs_boot $tuned_params
      grub_users $grub_users
      grub_arg --unrestricted
      grub_class rhel  

      when running the following command and rebooting the issue is solved and openvswitch start correctly:

      grub2-mkconfig -o /boot/grub2/grub.cfg --update-bls-cmdline

       

              eshulman Ella Shulman
              eshulman Ella Shulman
              rhos-dfg-nfv
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Created:
                Updated:
                Resolved: