Le 25 juil. 2018 à 13:42, Edward Haas <ehaas@redhat.com> a écrit :



On Tue, Jul 24, 2018 at 1:08 PM, Fabrice Bacchella <fabrice.bacchella@orange.fr> wrote:


Le 24 juil. 2018 à 11:50, Dominik Holler <dholler@redhat.com> a écrit :

On Tue, 24 Jul 2018 11:04:58 +0200
Fabrice Bacchella <fabrice.bacchella@orange.fr> wrote:

To monitoring the network interfaces, I have a script that check if
ifAdminStatus and ifOperStatus values matches in snmp.

But with oVirt it fails on a server with 4 physical interfaces, but
only two connected, and return an error:


You want that eth0 and eth1 are UP, and eth2 and eth3 are DOWN?

Yes.


snmptable XXX IF-MIB::ifTable | less
SNMP table: IF-MIB::ifTable

ifIndex     ifDescr ifAdminStatus ifOperStatus
      1          lo            up           up
      2        eth0            up           up
      3        eth1            up           up
      4        eth2            up         down
      5        eth3            up         down
     24 ;vdsmdummy;          down         down
     25       vnet0            up           up


And indeed on the server:

ip link show eth2
4: eth2: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc mq state
DOWN mode DEFAULT group default qlen 1000 link/ether
40:a8:f0:30:81:1a brd ff:ff:ff:ff:ff:ff


looks like eth2 is DOWN, as expected.

It's in state DOWN, but marked UP anyway.

A really DOWN interface is shown as (on another server, not an ovirt host):

4: eth2: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether a0:d3:c1:fa:8c:8a brd ff:ff:ff:ff:ff:ff



eth2 seems to be in state DOWN, which seems to be reflected in
ifOperStatus.

Yes it match. The state is reflected in the ifOperStatus. The ifAdminStatus match the UP in the <...>


Is the issue that ifAdminStatus is up for eth2 and eth3, but you want
it to be down?

That's it. I never ask it to be in such state.


If eth2 and eth3 are not defined under oVirt control, I see no reason for the system to touch it.
Perhaps, you machine has these interfaces under NetworkManager control (you can do "nmcli device" to check it),
in that case, NM will keep the admin state up and monitor it.
If you want it down, mark the interfaces as unmanaged (by NM) and perform an ifdown on them.

Let us know if it helped.
 

I added:

diff --git a/NetworkManager/NetworkManager.conf b/NetworkManager/NetworkManager.conf
index 1979ea6..420aba5 100644
--- a/NetworkManager/NetworkManager.conf
+++ b/NetworkManager/NetworkManager.conf
@@ -48,3 +48,5 @@
 #
 #level=TRACE
 #domains=ALL
+[keyfile]
+unmanaged-devices=eth2;eth3


And indeed it solves my problem with no unwanted side effects.