[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

is it recommended to disable ipv6

I am running OCP 3.7.54.

I can only find notes indicating IPv6 is not supported, but no indication in the prereqs that it should be disabled on the nodes. In my case IPv6 is not specifically configured so there are only link local addresses present. I suspect it is causing me problems.

While investigating timeouts during container creation I noticed lots of messages like this:

Jul 19 14:00:38 ose-prod-node-09.example.com kernel: IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
Jul 19 14:00:38 ose-prod-node-09.example.com kernel: IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready

This is odd because I have no eth0 physical interface, ifcfg-eth0 file, or udev rules referencing it.

[root ose-prod-node-09 ~]# nmcli dev
DEVICE          TYPE         STATE         CONNECTION
team0           team         connected     mlag
docker0         bridge       connected     docker0
eno1            ethernet     connected     team0-eno1
eno2            ethernet     connected     team0-eno2
eno3            ethernet     disconnected  --
eno4            ethernet     disconnected  --
vxlan_sys_4789  vxlan        disconnected  --
eth0            ethernet     unmanaged     --
veth00853eea    ethernet     unmanaged     --

While looking at the system a couple hours later there was suddenly a second default route added with a lower metric that made the node unreachable. I wonder if this was caused by NetworkManager reacting to a change in the unexplained eth0 interface. Oddly, the default route points to the IP of the tun0 interface as next hop and is directed out this previously nonexistant eth0 interface:

[root ose-prod-node-09 ~]# ip route ls
default via dev eth0     <------ !
default via dev team0 proto dhcp metric 350 dev tun0 scope link dev eth0 proto kernel scope link src dev team0 proto kernel scope link src metric 350 dev docker0 proto kernel scope link src dev tun0 dev eth0

[root ose-prod-node-09 ~]# ip -4 -o a
1: lo    inet scope host lo\       valid_lft forever preferred_lft forever
6: team0    inet brd scope global noprefixroute dynamic team0\       valid_lft 70821sec preferred_lft 70821sec
11: docker0    inet scope global docker0\       valid_lft forever preferred_lft forever
13: tun0    inet brd scope global tun0\       valid_lft forever preferred_lft forever
17845: eth0    inet brd scope global eth0\       valid_lft forever preferred_lft forever   <---- !

Any idea how or why this eth0 would have been created by NetworkManager? Could it be caused by IPv6 being enabled and triggering all those link state events?

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]