Improve vagrant reliablility, fix race condition with openvswitch and docker

cluster/saltbase/salt/kubelet/init.sls

@@ -72,7 +72,3 @@ kubelet:
       - file: /etc/init.d/kubelet
 {% endif %}
       - file: /var/lib/kubelet/kubernetes_auth
-{% if grains.network_mode is defined and grains.network_mode == 'openvswitch' %}
-      - sls: sdn
-{% endif %}
-

cluster/saltbase/salt/sdn/init.sls

@@ -1,15 +1,8 @@
 {% if grains.network_mode is defined and grains.network_mode == 'openvswitch' %}
 
-openvswitch:
-  pkg:
-    - installed
-  service.running:
-    - enable: True
-
 sdn:
   cmd.wait:
     - name: /kubernetes-vagrant/network_closure.sh
     - watch:
-      - pkg: docker-io
-      - pkg: openvswitch
+      - sls: docker
 {% endif %}

cluster/saltbase/salt/top.sls

@@ -37,7 +37,9 @@ base:
     - monit
     - nginx
     - kube-client-tools
+{% if grains['cloud'] is defined and grains['cloud'] != 'vagrant' %}
     - logrotate
+{% endif %}
     - kube-addons
 {% if grains['cloud'] is defined and grains['cloud'] == 'azure' %}
     - openvpn

cluster/vagrant/provision-master.sh

@@ -60,6 +60,14 @@ done
 mkdir -p /etc/salt/minion.d
 cat <<EOF >/etc/salt/minion.d/master.conf
 master: '$(echo "$MASTER_NAME" | sed -e "s/'/''/g")'
+master: '$(echo "$MASTER_NAME" | sed -e "s/'/''/g")'
+auth_timeout: 10
+auth_tries: 2
+auth_safemode: True
+ping_interval: 1
+random_reauth_delay: 3
+state_aggregrate:
+  - pkg
 EOF
 
 cat <<EOF >/etc/salt/minion.d/grains.conf

cluster/vagrant/provision-minion.sh

@@ -34,9 +34,26 @@ for (( i=0; i<${#MINION_NAMES[@]}; i++)); do
 done
 
 # Let the minion know who its master is
+# Recover the salt-minion if the salt-master network changes
+## auth_timeout - how long we want to wait for a time out
+## auth_tries - how many times we will retry before restarting salt-minion
+## auth_safemode - if our cert is rejected, we will restart salt minion
+## ping_interval - restart the minion if we cannot ping the master after 1 minute
+## random_reauth_delay - wait 0-3 seconds when reauthenticating
+## recon_default - how long to wait before reconnecting
+## recon_max - how long you will wait upper bound
+## state_aggregrate - try to do a single yum command to install all referenced packages where possible at once, should improve startup times
+##
 mkdir -p /etc/salt/minion.d
 cat <<EOF >/etc/salt/minion.d/master.conf
 master: '$(echo "$MASTER_NAME" | sed -e "s/'/''/g")'
+auth_timeout: 10
+auth_tries: 2
+auth_safemode: True
+ping_interval: 1
+random_reauth_delay: 3
+state_aggregrate:
+  - pkg
 EOF
 
 cat <<EOF >/etc/salt/minion.d/log-level-debug.conf

cluster/vagrant/provision-network.sh

@@ -39,6 +39,11 @@ grep -q kbr0 /etc/sysconfig/docker || {
   # Stop docker before making these updates
   systemctl stop docker
 
+  # Install openvswitch
+  yum install -y openvswitch
+  systemctl enable openvswitch
+  systemctl start openvswitch
+
   # create new docker bridge
   ip link set dev ${DOCKER_BRIDGE} down || true
   brctl delbr ${DOCKER_BRIDGE} || true
@@ -85,6 +90,7 @@ grep -q kbr0 /etc/sysconfig/docker || {
   echo "OPTIONS='-b=kbr0 --selinux-enabled ${DOCKER_OPTS}'" >/etc/sysconfig/docker
   systemctl daemon-reload
   systemctl start docker
+
 }
 EOF