Closed issmirnov closed 11 months ago
Seeing this, too, in a similar environment (Ubuntu/rke2/rancher/hardened-calico:v3.24.1-build20221011)
I thought it might be this bug which was introduced in v3.24.3 and fixed in v3.24.4: https://github.com/projectcalico/calico/issues/6927
But since @cureforoptimism is seeing this in v3.24.1, sounds like it might be something different.
2022-11-05 02:30:46.466 [ERROR][2619] felix/route_table.go 1035: Failed to get link attributes error=interface not present ifaceRegex="^vxlan.calico$" ipVersion=0x4 tableIndex=0
Does the vxlan.calico
interface exist on your node if you run ip link show
?
Not on my node - just the calibfooos
How can we best assist with debugging?
We're eager to get this launched, and will happily deploy some eng resources to generate more logs or run dev builds.
Happy to help in any way possible once this reaches the top of your queue.
Thank you in advance!
Setting spec.calicoNetwork.nodeAddressAutodetectionV4
in the Installation
resource to something other than firstFound
should rule out #6927 for certain.
https://projectcalico.docs.tigera.io/reference/installation/api#operator.tigera.io/v1.NodeAddressAutodetection
Is there anything else running that might be trying to manage interfaces? e.g. network manager, etc
I also have the same problem, as my hosts are arm64 and calico also fails to gather ipset list
while executing:
$ sudo -E calicoctl node diags
Collecting diagnostics
Using temp dir: /tmp/calico3738869463
Dumping netstat
Dumping routes (IPv4)
Dumping routes (IPv6)
Dumping interface info (IPv4)
Dumping interface info (IPv6)
Dumping iptables (IPv4)
Dumping iptables (IPv6)
Dumping ipsets
Failed to run command: ipset list
Error:
Dumping ipsets (container)
Failed to run command: docker run --rm --privileged --net=host calico/node ipset list
Error: ipset v7.1: Kernel and userspace incompatible: settype hash:net with revision 7 not supported by userspace.
Copying journal for calico-node.service
Dumping felix stats
Failed to run command: pkill -SIGUSR1 felix
Error:
Copying Calico logs
Error creating log directory: mkdir /tmp/calico3738869463/diagnostics/logs: file exists
Diags saved to /tmp/calico3738869463/diags-20221205_064915.tar.gz
If required, you can upload the diagnostics bundle to a file sharing service.
As we can see here it runs
docker run --rm --privileged --net=host calico/node ipset list
But, if you run
sudo docker run --rm --privileged --platform=aarch64 --net=host calico/node:v3.24.5-arm64 ipset list
You'll see proper output, so, is that somehow related to some inner ctl bug?
And that's a bit strange because latest and v3.24.5-arm64
differs only by Variant
property (version doesnt have that).
UPD full difference between the versions is here
There are some differences in ContainerConfig
and Config
props.
I am attempting to install Calico on 3x k3s nodes. The
Expected Behavior
Calico nodes should come up and work, as per https://projectcalico.docs.tigera.io/getting-started/kubernetes/k3s/multi-node-install.
Current Behavior
Calico does not come up.
Output of
kubectl get -n calico-system all
Logs of calico pod:
Output of
sudo kubectl calico node diags
Output of
ifconfig
on the master node:Ipset version:
Output of
ipset list
Possible Solution
It seems that calico can't create the network interfaces. Per https://github.com/projectcalico/calico/issues/5717 this was fixed in v3.22.4, and my version of calico is newer. I'm not sure what's wrong.
Perhaps it's related to https://github.com/rancher/rancher/issues/38017 ?
Steps to Reproduce (for bugs)
tigera-operator.yaml
andcustom-resources.yaml
files.kubectl get -n calico-system all
and observer the failed pods.Context
I am attempting to deploy k3s, replacing flannel with calico for our workloads.
Your Environment