search

kubernetes-sigs / cluster-api

Home for Cluster API, a subproject of sig-cluster-lifecycle
https://cluster-api.sigs.k8s.io
Apache License 2.0
3.57k stars 1.31k forks source link

Weave assigns conflicting IPs to Pods #83

Closed rsdcastro closed 6 years ago

rsdcastro commented 6 years ago

From @maisem on March 9, 2018 3:41

cc @medinatiger @jessicaochen @rsdcastro

Weave is assigning conflicting IPs to Pods and they also seem to be from different cidrs.

$ k get po --all-namespaces -o wide 
NAMESPACE     NAME                                       READY     STATUS    RESTARTS   AGE       IP           NODE
default       clusterapi-58c564d46b-bdc9g                2/2       Running   0          4m        10.32.0.3    gce-master-test1
default       etcd-clusterapi-0                          1/1       Running   0          4m        10.32.0.2    gce-master-test1
kube-system   etcd-gce-master-test1                      1/1       Running   0          4m        10.128.0.2   gce-master-test1
kube-system   kube-apiserver-gce-master-test1            1/1       Running   0          4m        10.128.0.2   gce-master-test1
kube-system   kube-controller-manager-gce-master-test1   1/1       Running   0          4m        10.128.0.2   gce-master-test1
kube-system   kube-dns-545bc4bfd4-g8mm4                  3/3       Running   0          5m        10.32.0.4    gce-master-test1
kube-system   kube-proxy-j8jpq                           1/1       Running   0          1m        10.128.0.3   gce-node-hwdx7
kube-system   kube-proxy-jtczr                           1/1       Running   0          5m        10.128.0.2   gce-master-test1
kube-system   kube-scheduler-gce-master-test1            1/1       Running   0          4m        10.128.0.2   gce-master-test1
kube-system   weave-net-7hkzx                            2/2       Running   0          1m        10.128.0.3   gce-node-hwdx7
kube-system   weave-net-hzgd8                            2/2       Running   1          5m        10.128.0.2   gce-master-test1

$ k get no
NAME               STATUS    ROLES     AGE       VERSION
gce-master-test1   Ready     master    5m        v1.8.3
gce-node-hwdx7     Ready     <none>    56s       v1.8.3

Repro steps:

1. ./generate-yaml.sh
1. ./gcp-deployer create -c cluster.yaml -m machines.yaml

Copied from original issue: kubernetes/kube-deploy#645

rsdcastro commented 6 years ago

Thank you @maisem!

This is a high priority bug that needs to be fixed.

rsdcastro commented 6 years ago

From @krousey on April 16, 2018 16:8

@kcoronado is working on fixing this. I believe it's a follow up to the configuration work she already has out. Assigning to myself to track since she's not in the k8s org.

/assign @krousey