Welcome to the Linux Foundation Forum!

CP looses internet connection after WORKER node JOIN.

Options

Hi all,

I need help again.

I succeeded into creation of CP node an WORKER one, but some seconds after join CP node is no more able to contact external world (example 'ping www.google.com') and there's no way to download ngnix image to complete pod startup.

Communication between nodes runs without issues.

They are running under QEMU hypervisor and after JOIN host is no more able to connect to guests using SSH.

So I have to fallback to QEMU shell access and manage hosts.

I've dedicated some time to investigate this network issue and found that on WORKER node annotations say:

... ... ... ...
projectcalico.org/IPv4Address: 192.168.122.90/24
projectcalico.org/IPv4IPIPTunnelAddr: 192.168.68.0
... ... ... ...

On CP node annotations say:

... ... ... ...
projectcalico.org/IPv4Address: 192.168.122.74/24
projectcalico.org/IPv4IPIPTunnelAddr: 192.168.242.64
... ... ... ...

Is difference between IPv4IPIPTunnelAddr values the cause of my trouble?
If not, where I have to look at?

Full describe of nodes in attachment.

Thanks for support as ever.

Massmiliano

Answers

  • I can add some more informations about kube-system pods in attachment.

    Almost CALICO pods under kube-system namespace are running, but I can see warning about their state and an ugly error I can't manage:

    error adding pod ********* to CNI network "k8s-pod-network": error getting ClusterInformation: Get "https://10.96.0.1:443/apis/crd.projectcalico.org/v1/clusterinformations/default": dial tcp 10.96.0.1:443: connect: connection refused

  • chrispokorni
    Options

    Hi @MassimilianoGullusci,

    Most networking issues are encountered during cluster installations on local VMs, with hypervisors like KVM or VirtualBox. I would recommend watching the set up videos for AWS and/or GCE, paying close attention to the networking recommendations. Then match those steps on your hypervisor. Overall ensure that you have a bridged network managed by the hypervisor, one network interface per VM, VM IP addresses not overlapping the Calico default subnet 192.168.0.0/16, no firewalls, all traffic allowed to/from your VMs, all protocols allowed, to all ports.

    Regards,
    -Chris

  • chrispokorni
    Options

    Hi @MassimilianoGullusci,

    The duplicate Discussion post has been removed.

    Regards,
    -Chris

Categories

Upcoming Training