Hi Ales,

Le mer. 16 mars 2022 à 07:11, Ales Musil <amusil@redhat.com> a écrit :

[../..]

I am trying to add a new ovirt-node-ng 4.4.10 node to my cluster managed by an ovirt-engine version 4.4.10.

My cluster is composed of other ovirt-node-ng which have been successively updated from version 4.4.4 to version 4.4.10 without any problem.

This new node is integrated normally in the cluster, however when I look at the status of the network part in the tab "Network interface" I see that all interfaces are "down".

Did you try to call "Refresh Capabilities"? It might be the case that the engine presents a different state that is on the host after upgrade.

I tried but and I show the pull in the vdsm.log on my faulty node. But the bond/interfaces states still "down". I tried a fresh install several time  the node with "ovirt-node-ng-installer-4.4.10-2022030308.el8.iso" but the issue still there.

 

I have a paperclip at the "bond0" interface that says: "Bond state; NONE"

I compared the content of "/etc/sysconfig/network-script" between an hypervisor which works and the one which has the problem and I notice that a whole bunch of files are missing and in particular the "ifup/ifdown...." files. The folder contains only the cluster specific files + the "ovirtmgmt" interface.

Since 4.4 in general we don't use initscripts anymore, so those files are really not a good indicator of anything. We are using nmstate + NetworkManager, if the connection are correctly presented here
everything should be fine.
 

networkManager show interfaces and bond up and running from the node perspective
nmcli con show --active
NAME       UUID                                  TYPE      DEVICE    
ovirtmgmt  6b08c819-6091-44de-9546-X  bridge    ovirtmgmt
virbr0     91cb9d5c-b64d-4655-ac2a-X  bridge    virbr0    
bond0      ad33d8b0-1f7b-cab9-9447-X  bond      bond0    
eno1       abf4c85b-57cc-4484-4fa9-X  ethernet  eno1      
eno2       b186f945-cc80-911d-668c-X  ethernet  eno2 
 
nmstatectl show return correct states

- name: bond0
  type: bond
  state: up
  accept-all-mac-addresses: false
  ethtool:
    feature:
        [../..]
  ipv4:
    enabled: false
    address: []
    dhcp: false
  ipv6:
    enabled: false
    address: []
    autoconf: false
    dhcp: false
  link-aggregation:
    mode: active-backup
    options:
      all_slaves_active: dropped
      arp_all_targets: any
      arp_interval: 0
      arp_validate: none
      downdelay: 0
      fail_over_mac: none
      miimon: 100
      num_grat_arp: 1
      num_unsol_na: 1
      primary: eno1
      primary_reselect: always
      resend_igmp: 1
      updelay: 0
      use_carrier: true
    port:
    - eno1
    - eno2
  lldp:
    enabled: false
  mac-address: X
  mtu: 1500

The state for eno1 and eno2 is "up".


The hypervisor which has the problem seems to be perfectly functional, ovirt-engine does not raise any problem.

This really sounds like something that a simple call to "Refresh Capabilities" could fix.

I did it several times. Everything is fetched (I checked in the logs), but the states are still down for all interfaces... If I do a fresh install in 4.4.4, the states shown by rhevm are OK, if I reinstall in 4.4.10 the WebUI Hosts/<node>/Network Interfaces is KO.