OS-specific known issues

The following issues were encountered when testing Antrea on different OSes, or reported by Antrea users. When possible we try to provide a workaround.

CoreOS

Issues
#626

CoreOS Container Linux will reach its end-of-life on May 26, 2020 and will no longer receive updates. It is recommended to migrate to another Operating System as soon as possible.

CoreOS uses networkd for network configuration. By default, all interfaces are managed by networkd because of the configuration files that ship with CoreOS. Unfortunately, that includes the gateway interface created by Antrea (antrea-gw0 by default). Most of the time, this is not an issue, but if networkd is restarted for any reason, it will cause the interface to lose its IP configuration, and all the routes associated with the interface will be deleted. To avoid this issue, we recommend that you create the following configuration files:

  1. # /etc/systemd/network/90-antrea-ovs.network
  2. [Match]
  3. # use the correct name for the gateway if you changed the Antrea configuration
  4. Name=antrea-gw0 ovs-system
  5. Driver=openvswitch
  7. [Link]
  8. Unmanaged=yes
  1. # /etc/systemd/network/90-antrea-veth.network
  2. # may be redundant with 50-docker-veth.network (name may differ based on CoreOS version), which should not be an issue
  3. [Match]
  4. Driver=veth
  6. [Link]
  7. Unmanaged=yes
  1. # /etc/systemd/network/90-antrea-tun.network
  2. [Match]
  3. Name=genev_sys_* vxlan_sys_* gre_sys stt_sys_*
  5. [Link]
  6. Unmanaged=yes

Note that this fix requires a version of CoreOS >= 1262.0.0 (Dec 2016), as the networkd Unmanaged option was not supported before that. See CoreOS release notes.

Photon OS 3.0

Issues
#591

If your K8s Nodes are running Photon OS 3.0, you may see error messages in the antrea-agent logs like this one: "Received bundle error msg: [...]". These messages indicate that some flow entries could not be added to the OVS bridge. This usually indicates that the Kernel was not compiled with the CONFIG_NF_CONNTRACK_ZONES option, as this option was only enabled recently in Photon OS. This option is required by the Antrea OVS datapath. To confirm that this is indeed the issue, you can run the following command on one of your Nodes:

  1. grep CONFIG_NF_CONNTRACK_ZONES= /boot/config-`uname -r`

If you do not see the following output, then it confirms that your Kernel is indeed missing this option:

  1. CONFIG_NF_CONNTRACK_ZONES=y

To fix this issue and be able to run Antrea on your Photon OS Nodes, you will need to upgrade to a more recent version: >= 4.19.87-4 (Jan 2020). You can achieve this by running tdnf upgrade linux-esx on all your Nodes.

After this fix, all the Antrea Agents should be running correctly. If you still experience connectivity issues, it may be because of Photon’s default firewall rules, which are quite strict by default. The easiest workaround is to accept all traffic on the gateway interface created by Antrea (antrea-gw0 by default), which enables traffic to flow between the Node and the Pod network: iptables -A INPUT -i antrea-gw0 -j ACCEPT