refreshes the dhcp info on boot, and fixes a bug where a server would linger in the background if not enough disk available

Author: Kim Brugger

ehos/ehos.py

@@ -384,10 +384,15 @@ def create_execute_nodes( config:Munch,execute_config_file:str, nr:int=1):
  if ( 'scratch_size' in config.ehos and
  config.ehos.scratch_size is not None and
  config.ehos.scratch_size != 'None'):
- 
- volume_id = cloud.volume_create(size=config.ehos.scratch_size, name=node_name)
- cloud.attach_volume( node_id, volume_id=volume_id)
 
+ try:
+ volume_id = cloud.volume_create(size=config.ehos.scratch_size, name=node_name)
+ cloud.attach_volume( node_id, volume_id=volume_id)
+ except:
+ logger.warning("Could not create execute server, not enough disk available, deleting the instance.")
+ cloud.server_delete( node_id )
+
+ 
  instances.add_node( id=node_id, name=node_name, cloud=cloud_name, status='starting', state='booting')
  logger.debug("Execute server {}/{} is booting".format( node_id, node_name))
 

share/execute.yaml

@@ -75,6 +75,7 @@ write_files:
  permissions: '0644'
 
 runcmd:
+ - dhclient # because the DNS server at uh-cloud is sometimes broken...
  - systemctl enable condor
  - systemctl start condor
  - condor_store_cred -p {password} -f /var/lock/condor/pool_password

share/master.yaml

@@ -62,6 +62,7 @@ write_files:
 
 runcmd:
 
+ - dhclient # because the DNS server at uh-cloud is sometimes broken...
  - systemctl enable condor
  - systemctl start condor