Implement parallel node provisioning

examples/provisioning/provision.py

@@ -13,9 +13,7 @@ from solar.events.api import add_event
 
 discovery_service = 'http://0.0.0.0:8881'
 
-# DEBUG use a single node
-nodes_list = [requests.get(discovery_service).json()[0]]
-
+nodes_list = requests.get(discovery_service).json()
 
 # Create slave node resources
 node_resources = vr.create('nodes', 'templates/not_provisioned_nodes.yaml', {'nodes': nodes_list})

examples/provisioning/provision.sh

@@ -4,6 +4,10 @@ set -eux
 
 DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
 
+# Remove generated pxe exclude files
+sudo rm -f /etc/dnsmasq.d/no_pxe_*.conf
+sudo service dnsmasq restart
+
 solar resource clear_all
 python "${DIR}"/provision.py
 
@@ -11,4 +15,3 @@ solar changes stage
 solar changes process
 solar orch run-once last
 watch --color -n1 'solar orch report last'
-

resources/dnsmasq/actions/exclude_mac_pxe.yaml

@@ -3,4 +3,15 @@
 
   tasks:
     - lineinfile: create=yes dest=/etc/dnsmasq.d/no_pxe_{{exclude_mac_pxe | replace(':', '_')}}.conf line="dhcp-host={{exclude_mac_pxe}},set:nopxe"
+      # FIXME: currently there is no way to specify
+      # policy not to run several tasks in parallel,
+      # so when we deploy several nodes in parallel
+      # it causes the problems when two tasks try
+      # to restart supervisor at the same time, and
+      # fail to do it.
+    - command: service dnsmasq status
+      register: log
+      until: log.stdout.find('running') > -1
+      retries: 5
+      delay: 2
     - shell: service dnsmasq restart

resources/not_provisioned_node/actions/run.sh

@@ -4,3 +4,5 @@ set -eux
 
 # Fake run action which is required in order to make
 # dependency `run` -> `provision`
+
+exit 0