EC2 backend will kill any not-yet-ready VM when the SessionBasedScript will exit

What steps will reproduce the problem?
1. Run any SessionBasedScript without -C option on a cloud resource and 
increased verbose
2. Check the output.

What is the expected output? What do you see instead?

A VM will be created, then all the jobs will fail to submit because the VM is 
probably not ready yet, but when the script will finish (after one iteration), 
then the resource is removed and since we don't have yet access to the 
associated resource the VM is terminated.

This is an example output:

gc3.gc3libs: DEBUG: Connecting to host 
'server-6530e9bf-218e-4e34-8bf5-880e5163ef53' as user 'gc3-user' via SSH ...
gc3.gc3libs: ERROR: Could not create ssh connection to 
server-6530e9bf-218e-4e34-8bf5-880e5163ef53: gaierror: [Errno -2] Name or 
service not known
gc3.gc3libs: WARNING: Error while updating EC2 subresource 
server-6530e9bf-218e-4e34-8bf5-880e5163ef53@hobbes: Failed while connecting to 
remote host 'server-6530e9bf-218e-4e34-8bf5-880e5163ef53': [Errno -2] Name or 
service not known. Turning off associated VM.
gc3.gc3libs: INFO: VM instance i-000003d2 at 
server-6530e9bf-218e-4e34-8bf5-880e5163ef53 is no longer needed. Terminating.

Original issue reported on code.google.com by antonio....@gmail.com on 6 Feb 2013 at 3:57

goerz / gc3pie

EC2 backend will kill any not-yet-ready VM when the SessionBasedScript will exit #392