helm based contrail deployment failing on kubernetes

[tnithya]

Hi,
I have followed the install procedures needed to bring up opencontrail on top of kubernetes cluster.
I wanted to have a setup that is a nested contrail controller as shown in Rudra's talk https://www.openstack.org/videos/boston-2017/juniper-opencontrail-for-kubernetes-openstack-and-nested-kubernetes

Following are what I assumed are the steps needed:

1. Get kubernetes cluster up (without any CNI plugin configured on the master)
2. Use Helm install to install CNI plugin and deploy contrail cluster on top of the kube cluster.
   This step I assumed was performed via the steps provided
   https://github.com/Juniper/contrail-docker/wiki/Helm-based-contrail-provisioning-for-Kubernetes-orchestration
3. Use Helm to install openstack cluster on top of kubernetes cluster with networking handled by opencontrail.
4. Deploy VMs or VNFs on top of this that are controlled using opencontrail.

I'm stuck at step 2.

After several attempts I managed to get the helm install to complete successfully.
However, using the script

for pod_name in `kubectl get pods -n kube-system -o wide | grep contrail | awk '{print $1}'`; do kubectl exec -it  $pod_name -n kube-system -- contrail-status ; done 

I find most of the components of contrail are shown inactive and disabled on boot.

Is there a detailed installation guide that can be used for getting this setup?

Following are the nodes I have as part of my kubernetes cluster:

kubectl get nodes
NAME STATUS ROLES AGE VERSION
r12u NotReady 3d v1.9.0
r16u NotReady 3d v1.9.0
r233u Ready master 3d v1.9.0
r238u NotReady 3d v1.9.0

I have disabled CNI on the master node by changing as specified in another document.

Any help is greatly appreciated.

[kranthikirang]

Are the pods status showing running? Seems weird that worker nodes are in NotReady state.
kubectl -n kube-system get pods -o wide

I have seen issues with finding the physical interface to crease vhost0 in compute nodes. So, basically we need to tweak the helm chart a bit. Which version are you using? Also, I have seen issues contrail services won't come up because of analytics DB doesn't come up. We have to start that service manually and restart the other services in analytics and collector container.

[kranthikirang]

Also the documentation is not up-to-date

[tnithya]

@kranthikirang Thanks for your response. The pods are not in ready state as I have disabled the CNI plugin in the master as required by one of the documentations.
I'm not sure as to how I can enable contrail CNI plugin for K8s using helm. I followed the documentation and expected this to be enabled automatically as mentioned in step 2 of my earlier message in this thread.
The version of helm charts is 2.5.1

The issue Im finding is that the vrouter agent pod is not getting deployed on the kubernetes worker nodes 192.168.xx.116,117,118 as shown below in bold

NAME	READY	STATUS	RESTARTS	AGE	IP	NODE
contrail-analytics-854nm	1/1	Running	0	5d	192.168.61.114	r233u
contrail-analyticsdb-cb7hg	1/1	Running	0	5d	192.168.61.114	r233u
contrail-controller-w67nm	1/1	Running	0	5d	192.168.61.114	r233u
contrail-kube-manager-ttnkm	1/1	Running	0	5d	192.168.61.114	r233u
contrail-vrouter-agent-7zlkl	1/1	Running	0	5d	192.168.61.114	r233u
contrail-vrouter-agent-hs7pj	0/1	Init:0/1	0	5d	192.168.61.116	r16u
contrail-vrouter-agent-jt7l9	0/1	Init:0/1	0	5d	192.168.61.117	r238u
contrail-vrouter-agent-mkwv2	0/1	Init:0/1	0	5d	192.168.61.118	r12u
etcd-r233u	1/1	Running	0	6d	192.168.61.114	r233u
kube-apiserver-r233u	1/1	Running	0	6d	192.168.61.114	r233u
kube-controller-manager-r233u	1/1	Running	0	6d	192.168.61.114	r233u
kube-scheduler-r233u	1/1	Running	0	6d	192.168.61.114	r233u
tiller-ds-dpzsk	1/1	Running	0	6d	192.168.61.114	r233u

[kranthikirang]

Hi @tnithya

I didn't use contrail for my k8s CNI. You may have to follow this guide. I generally use contrail for my tenant network needs

https://github.com/Juniper/contrail-docker/wiki/Provision-Contrail-CNI-for-Kubernetes