On 02/07/2012 07:02 AM, Lucas Meneghel Rodrigues wrote:
> On 02/07/2012 05:47 AM, tangchen wrote:
>> Hi~
>>
>> On 02/07/2012 01:52 AM, Chris Evich wrote:
>>> Rather than lock the code into the implementation details of particular
>>> virsh options, just code for the minimum necessary. However, expose
>>> a set of free-form variables tests can use to exercize more specific
>>> options.
>>>
>> OK, that will solve the compatibility problems.
>>
>>> + # Rely on test-code to verify guest state on receiving-end
>>> + # Assume success unless proven otherwise
>> But here, migrate command needs users to input a password when
>> connecting to
>> destination host, such as a ssh password.
>> Do we need to implement it as an auto-login function ?
>
> I'd just document somewhere that an ssh key setup needs to exist between
> the machine running the test and the destination machine.
>

All,

Yes, an ssh-pubkey setup is required (for now).  The issue is, we 
support passing the -c option to virsh, but it only supports pubkey 
right now.  However, libvirt supports other auth types like TLS certs, 
and manually entering a password.  These should be supported (so they 
can be tested), but for now using ssh pubkey is quick and easy.

Lucas,

Along the lines of supporting other types, I found a "bug" if 
~/.ssh/known_hosts doesn't contain the destination host.  This results 
in the "Are you sure you want to add the host key yes/no" prompt, which 
blocks on input.

What do you think is the best way to solve this, and possibly automated 
entry of passwords to virsh?  i.e. Should virsh_cmd doing something 
similar to vm.login() instead of just tossing a utils.run() and preying 
for the best?

-- 
Chris Evich, RHCA, RHCE, RHCDS, RHCSS
Quality Assurance Engineer
e-mail: cevich + `@' + redhat.com o: 1-888-RED-HAT1 x44214
_______________________________________________
Autotest mailing list
[email protected]
http://test.kernel.org/cgi-bin/mailman/listinfo/autotest

Reply via email to