Re: Re: Virtualization support
On Wed, Nov 10, 2010 at 9:22 AM, Scott M Ferguson <smferguson(a)gmail.com> wrote:
> Message: 3
> Date: Tue, 9 Nov 2010 08:29:27 -0500 (EST)
> From: Kamil Paral <kparal(a)redhat.com>
> Subject: Re: Virtualization support
> To: AutoQA development <autoqa-devel(a)lists.fedorahosted.org>
> Message-ID:
>
<1532790094.383731289309367710.JavaMail.root(a)zmail03.collab.prod.int.phx2.redhat.com>
>
> Content-Type: text/plain; charset=utf-8
>
> ----- "Scott M Ferguson" <smferguson(a)gmail.com> wrote:
>
>> Hey all,
>>
>> Joza was kinda enough to let me have a go at part of the
>> virtualization support: enabling 2-way communication between a
>> virt-host and virt-guests and I'm starting to dig into it. Per his
>> suggestion I'm working on a generic api that can be tuned based on
>> the
>> needs of the project and I wanted to bring it up to the list. A few
>> initial thoughts from Joza regarding communication were:
>>
>> Host->Guest - this needs to be able to select a certain virt guest
>> 1) Is a test actually running?
>> 2) Disconnect autotest-enabled eth
>> 3) run a command
>>
>> Guest->Host
>> 1) Test finished
>> 2) Test crashed
>> 3) Test running too long
>> 4) Destructive test finished, revert me to 'safe' snapshot
>>
>> He also noted a management interface on the server might be
>> interesting and James noted that it will need to play nice with
>> autotest.
>>
>> I know this is a long-term goal, but I'd love to hear everyone's
>> thoughts when time permits.
>
> Hey Scott,
>
> it absolutely great that you want to help us with this one. We really
> appreciate it. I don't know how many details jskladan provided to you,
> but here us some additional information about it:
>
> https://fedorahosted.org/pipermail/autoqa-devel/2010-June/000642.html
> https://fedorahosted.org/autoqa/milestone/Virtualization
>
> Don't hesitate to ask and consult anything related in here.
>
> Thanks,
> Kamil
>
Thanks for the additional info. I'll definitely keep the list updated
and likely have lots of questions.
Best,
Scott
Hey all. I have a rough proof-of-concept going for the client/server,
but in chatting it over with a friend a novel idea came up. What if
the server just opened 2 ssh sessions for each client. 1 to run the
tests, 1 to tail the log (grab the result). Then we wouldn't need a
client daemon. I'm not sure this is desirable since other projects use
a standard client/server model, but thought it was worth bringing up.
Best,
Scott