kube-dns fails to start with - FailedCreatePodSandBox error #587
Comments
|
Seems like you haven't install any pod network plugin. |
|
I'm seeing very similar behavior as of a few hours ago. I can confirm that it seems to occur with flanneld container up and running applied from the standard recommended pod-network plugin yaml. This is a bare-metal install on a blade system. (The restarts are normal for us, as pulling images often takes several tries for some reason on our net). |
|
Figured this out. Flanneld requires the portmap binary in when using the default plugin yaml. A pull request is in at utf18/ansible-kubeadm#3 To work around the issue you can grab portmap and place it at /opt/cni/bin/portmap set to 0755. |
|
@Inevitable, thanks. in the mean time. I got this running with I am curious as to why there are 6 options in the install instructions. More specifically, there is no default choice or a clear reason why I (blissfully ignorant of the nuances of networking) would choose one over the other. Would it make sense to have an officially supported version, or an initial recommendation? |
|
@Inevitable - how did you figure out this was related to Portmap? I want to confirm this solved the issue for me to, and that the same issue was causing similar behaviour |
|
Bit of search-fu based on the flanneld pod log lead me to flannel-io/flannel#890 From there just a simple test to see if my situation was the same. |
|
Thanks @Inevitable for the clear solution! |
|
I think this is still a bug. It is happening for me with 1.8.4 in GKE... I've tried deleting the host node to get it on a new one and it has the same problem over and over. Deployment config: ---
# ØMQ forwarder
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
annotations:
maintainers: "Peter Novotnak <[email protected]>,Jane Doe <[email protected]>"
source: https://github.com/myorg/myproj
labels:
name: myprojorwhatever
tier: backend
name: zmq-fwd
spec:
replicas: 1
strategy:
rollingUpdate:
maxSurge: 1
maxUnavailable: 1
type: RollingUpdate
template:
metadata:
labels:
app: zmq
spec:
containers:
- name: zmq-fwd
image: gcr.io/myproj/zmq-dev
command:
- invoke
- zmq-forwarder
env:
- name: ZMQ_FORWARDER_PULL
value: 'tcp://*:5556'
- name: ZMQ_FORWARDER_PUB
value: 'tcp://*:5557'
ports:
- containerPort: 5556
name: zmq-fwd-pull
protocol: TCP
- containerPort: 5557
name: zmq-fwd-pub
protocol: TCP
resources:
requests:
cpu: "1"
memory: "100m"
limits:
cpu: "1"
memory: "300m"Associated service: apiVersion: v1
kind: Service
metadata:
name: zmq-fwd
spec:
ports:
- name: zmq-fwd-pull
port: 5556
protocol: TCP
targetPort: zmq-fwd-pull
- name: zmq-fwd-pub
port: 5557
protocol: TCP
targetPort: zmq-fwd-pub
selector:
name: zmq-fwd
sessionAffinity: None
type: ClusterIP
status:
loadBalancer: {}The result of a |
|
I don't know if I can provide access to our cluster but I can provide uncensored logs to anyone looking into this. |
|
Ah, in my case this is because the requests/ limits I have configured are written incorrectly. |
|
Closing, this is a heavily validated area and is typically a conf or setup issue. |
What keywords did you search in kubeadm issues before filing this one?
I was suggested to create a new issue from #507
Is this a BUG REPORT or FEATURE REQUEST?
BUG REPORT
Versions
kubeadm version (use
kubeadm version):Environment:
kubectl version):ubuntu/xenialLinux ubuntu-xenial 4.4.0-101-generic #124-Ubuntu SMP Fri Nov 10 18:29:59 UTC 2017 x86_64 x86_64 x86_64 GNU/LinuxWhat happened?
kube-dns pod is failing to start when creating a cluster with kubeadm
What you expected to happen?
How to reproduce it (as minimally and precisely as possible)?
This gist contains the scripts I use to create the cluster :
To repro, you can just run:
Then to ssh into the machine run
vagrant sshAnything else we need to know?
kubelet logs:
The text was updated successfully, but these errors were encountered: