On each start-up make several attempts to check that Ironic is available

README.rst

@@ -171,6 +171,7 @@ v0.2.2
 
 * ``/v1/discover`` now does some sync checks (`bug #3`_).
 * Actually able to start under Python 3.3.
+* On each start-up make several attempts to check that Ironic is available.
 
 .. _bug #3: https://github.com/Divius/ironic-discoverd/issues/3
 

example.conf

@@ -10,6 +10,11 @@
 ; Tenant name for accessing Ironic API.
 ;os_tenant_name =
 
+; Number of attempts to do when trying to connect to Ironic on start up.
+;ironic_retry_attempts = 5
+; Amount of time between attempts to connect to Ironic on start up.
+;ironic_retry_period = 5
+
 ; Interface on which dnsmasq listens, the default is for VM's.
 ;dnsmasq_interface = br-ctlplane
 ; Amount of time in seconds, after which repeat periodic update of firewall.

ironic_discoverd/discoverd.py

@@ -25,7 +25,9 @@ def init_conf():
                   'dnsmasq_interface': 'br-ctlplane',
                   'authenticate': 'true',
                   'firewall_update_period': '15',
-                  'ports_for_inactive_interfaces': 'false'})
+                  'ports_for_inactive_interfaces': 'false',
+                  'ironic_retry_attempts': '5',
+                  'ironic_retry_period': '5'})
 
 
 CONF = None

ironic_discoverd/main.py

@@ -74,6 +74,27 @@ def main():
 
     interface = discoverd.CONF.get('discoverd', 'dnsmasq_interface')
     firewall.init(interface)
+
+    # Before proceeding we try to make sure:
+    # 1. Keystone access is configured properly
+    # 2. Keystone has already started
+    # 3. Ironic has already started
+    attempts = discoverd.CONF.getint('discoverd', 'ironic_retry_attempts')
+    assert attempts >= 0
+    retry_period = discoverd.CONF.getint('discoverd', 'ironic_retry_period')
+    LOG.debug('Trying to connect to Ironic')
+    for i in range(attempts + 1):  # one attempt always required
+        try:
+            discoverd.get_client().driver.list()
+        except Exception as exc:
+            if i == attempts:
+                raise
+            LOG.error('Unable to connect to Ironic or Keystone, retrying %d '
+                      'times more: %s', attempts - i, exc)
+        else:
+            break
+        eventlet.greenthread.sleep(retry_period)
+
     period = discoverd.CONF.getint('discoverd', 'firewall_update_period')
     eventlet.greenthread.spawn_n(periodic_update, period)