enhanced reset for contiv

This commit is contained in:
Erwan Miran 2018-09-17 16:46:19 +02:00
parent bcaf2f9ea3
commit afa2a5f1c4
5 changed files with 220 additions and 0 deletions

View file

@ -0,0 +1,10 @@
#!/bin/bash
set -e
echo "Starting cleanup"
ovs-vsctl list-br | grep contiv | xargs -I % ovs-vsctl del-br %
for p in $(ifconfig | grep vport | awk '{print $1}');
do
ip link delete $p type veth
done
touch /tmp/cleanup.done
sleep 60

View file

@ -0,0 +1,66 @@
---
- name: reset | Check that kubectl is still here
stat:
path: "{{ bin_dir }}/kubectl"
register: contiv_kubectl
- name: reset | Delete contiv netplugin and netmaster daemonsets
kube:
name: "{{ item }}"
namespace: "kube-system"
kubectl: "{{ bin_dir }}/kubectl"
resource: "ds"
state: absent
with_items:
- contiv-netplugin
- contiv-netmaster
register: contiv_cleanup_deletion
tags:
- network
when:
- contiv_kubectl.stat.exists
- inventory_hostname == groups['kube-master'][0]
- name: reset | Copy contiv temporary cleanup script
copy:
src: ../files/contiv-cleanup.sh # Not in role_path so we must trick...
dest: /opt/cni/bin/cleanup
owner: root
group: root
mode: 0750
when:
- contiv_kubectl.stat.exists
- name: reset | Lay down contiv cleanup template
template:
src: ../templates/contiv-cleanup.yml.j2 # Not in role_path so we must trick...
dest: "{{ kube_config_dir }}/contiv-cleanup.yml" # kube_config_dir is used here as contiv_config_dir is not necessarily set at reset
register: contiv_cleanup_manifest
when:
- contiv_kubectl.stat.exists
- inventory_hostname == groups['kube-master'][0]
- name: reset | Start contiv cleanup resources
kube:
name: "contiv-cleanup"
namespace: "kube-system"
kubectl: "{{ bin_dir }}/kubectl"
resource: "ds"
state: latest
filename: "{{ kube_config_dir }}/contiv-cleanup.yml"
when:
- contiv_kubectl.stat.exists
- inventory_hostname == groups['kube-master'][0]
ignore_errors: true
- name: reset | Wait until contiv cleanup is done
command: "{{ bin_dir }}/kubectl -n kube-system get ds contiv-cleanup -o jsonpath='{.status.numberReady}'"
register: cleanup_done_all_nodes
until: cleanup_done_all_nodes.stdout|int == groups['k8s-cluster']|length
retries: 5
delay: 5
ignore_errors: true
changed_when: false
when:
- contiv_kubectl.stat.exists
- inventory_hostname == groups['kube-master'][0]

View file

@ -0,0 +1,57 @@
---
kind: DaemonSet
apiVersion: extensions/v1beta1
metadata:
name: contiv-cleanup
namespace: kube-system
labels:
k8s-app: contiv-cleanup
spec:
selector:
matchLabels:
k8s-app: contiv-cleanup
template:
metadata:
labels:
k8s-app: contiv-cleanup
spec:
hostNetwork: true
hostPID: true
tolerations:
- key: node-role.kubernetes.io/master
effect: NoSchedule
serviceAccountName: contiv-netplugin
containers:
- name: contiv-ovs-cleanup
image: {{ contiv_ovs_image_repo }}:{{ contiv_ovs_image_tag }}
command: ["/opt/cni/bin/cleanup"]
securityContext:
privileged: true
volumeMounts:
- mountPath: /etc/openvswitch
name: etc-openvswitch
readOnly: false
- mountPath: /var/run
name: var-run
readOnly: false
- mountPath: /opt/cni/bin
name: cni-bin-dir
readOnly: false
readinessProbe:
exec:
command:
- cat
- /tmp/cleanup.done
initialDelaySeconds: 3
periodSeconds: 3
successThreshold: 1
volumes:
- name: etc-openvswitch
hostPath:
path: /etc/openvswitch
- name: var-run
hostPath:
path: /var/run
- name: cni-bin-dir
hostPath:
path: /opt/cni/bin

View file

@ -0,0 +1,80 @@
---
apiVersion: apps/v1
# This manifest deploys the contiv-ovs pod.
kind: DaemonSet
apiVersion: extensions/v1beta1
metadata:
name: contiv-ovs
namespace: kube-system
labels:
k8s-app: contiv-ovs
spec:
selector:
matchLabels:
k8s-app: contiv-ovs
template:
metadata:
labels:
k8s-app: contiv-ovs
annotations:
scheduler.alpha.kubernetes.io/critical-pod: ''
spec:
hostNetwork: true
hostPID: true
tolerations:
- key: node-role.kubernetes.io/master
effect: NoSchedule
containers:
# Runs ovs containers on each Kubernetes node.
- name: contiv-ovsdb-server
image: {{ contiv_ovs_image_repo }}:{{ contiv_ovs_image_tag }}
command: ["/scripts/start-ovsdb-server.sh"]
securityContext:
privileged: false
# Won't work until https://github.com/contiv/ovs-docker/pull/4 is merged and image is built again
env:
- name: OVSDBSERVER_EXTRA_FLAGS
valueFrom:
configMapKeyRef:
name: contiv-config
key: contiv_ovsdb_server_extra_flags
volumeMounts:
- mountPath: /etc/openvswitch
name: etc-openvswitch
readOnly: false
- mountPath: /var/run
name: var-run
readOnly: false
- name: contiv-ovs-vswitchd
image: {{ contiv_ovs_image_repo }}:{{ contiv_ovs_image_tag }}
command: ["/scripts/start-ovs-vswitchd.sh"]
securityContext:
privileged: true
# Won't work until https://github.com/contiv/ovs-docker/pull/4 is merged and image is built again
env:
- name: OVSVSWITCHD_EXTRA_FLAGS
valueFrom:
configMapKeyRef:
name: contiv-config
key: contiv_ovs_vswitchd_extra_flags
volumeMounts:
- mountPath: /etc/openvswitch
name: etc-openvswitch
readOnly: false
- mountPath: /lib/modules
name: lib-modules
readOnly: true
- mountPath: /var/run
name: var-run
readOnly: false
volumes:
# Used by contiv-ovs
- name: etc-openvswitch
hostPath:
path: /etc/openvswitch
- name: lib-modules
hostPath:
path: /lib/modules
- name: var-run
hostPath:
path: /var/run

View file

@ -1,5 +1,12 @@
---
- name: reset | include file with pre-reset tasks specific to the network_plugin if exists
include_tasks: "{{ (role_path + '/../network_plugin/' + kube_network_plugin + '/tasks/pre-reset.yml') | realpath }}"
when:
- kube_network_plugin in ['contiv']
tags:
- network
- name: reset | stop services
service:
name: "{{ item }}"