Skip to content

FAQ

Kylin ARM system cross-host container access intermittently fails

Behavior

There is a problem with Kylin ARM system and some NIC offload, which can cause intermittent container network failure.

Use netstat to identify the problem:

# netstat -us
IcmpMsg:
    InType0: 22
    InType3: 24
    InType8: 117852
    OutType0: 117852
    OutType3: 29
    OutType8: 22
Udp:
    3040636 packets received
    0 packets to unknown port received.
    4 packet receive errors
    602 packets sent
    0 receive buffer errors
    0 send buffer errors
    InCsumErrors: 4
UdpLite:
IpExt:
    InBcastPkts: 10244
    InOctets: 4446320361
    OutOctets: 1496815600
    InBcastOctets: 3095950
    InNoECTPkts: 7683903

If InCsumErrors is present and increases with netwoork failures, you can confirm that this is the problem.

Solution

The fundamental solution requires communication with Kylin and the corresponding network card manufacturer to update the system and drivers. A temporary solution would be to turn off tx offload on the physical NIC, but this would cause a significant degradation in tcp performance.

ethtool -K eth0 tx off

From the community feedback, the problem can be solved by the 4.19.90-25.16.v2101 kernel.

Pod can not Access Service

Behavior

Pod can not access Service, and dmesg show errors:

netlink:Unknown conntrack attr (type=6, max=5)
openvswitch: netlink: Flow actions may not be safe on all matching packets.

This log indicates that the in-kernel OVS version is too low to support the corresponding NAT operation.

Solution

  1. Upgrade the kernel module or compile the OVS kernel module manually.
  2. If you are using an Overlay network you can change the kube-ovn-controller args, setting --enable-lb=false to disable the OVN LB to use kube-proxy for service forwarding.

Frequent leader selection occurs in ovn-central

Behavior

Starting from the v1.11.x version, in a cluster with 1w Pod or more, if OVN NB or SB frequently elects the master, the possible reason is that Kube-OVN periodically performs the ovsdb-server/compact action, which affects the master selection logic.

Solution

You can configure the environment variables for ovn-central as follows and turn off compact:

- name: ENABLE_COMPACT
  value: "false"

微信群 Slack Twitter Support Meeting

Comments