Geoffroy -
When I read the project outline, I was thinking about more of the management
aspect of it. Also, making it easily manageable and user-configurable sort
of paves the way for an automated FT system, as opposed to implementing the
automated system first and the management interface later. I built an OSCAR
cluster for our fluid dynamics lab, and everything went very smoothly. OSCAR
really is a great system, which is why I chose your organization to apply
with. I would like to help, and with GSoC I would have the resources to be
able to work on helping out the OSCAR community full-time over the next
several months.
That being said, I would really appreciate any help you all can provide. I
am still in the process of doing research on the IPMI spec and trying to
figure out the best approach to implementing this idea before I write my
proposal. If anyone has any thoughts, suggestions, etc, all are welcome.
I look forward to working on this project. I have a few smaller ideas that I
would like to see put into practice in OSCAR as well, and hopefully this
will mark the beginning of me becoming a regular contributor to the
organization.
- Travis W.
On Tue, Mar 25, 2008 at 12:00 AM, <[EMAIL PROTECTED]> wrote:
> Hi all,
>
> I second what was said: there is no point to try to rewrite what is
> already done
> by someone. I also would like to go further in the discussion: what do we
> have
> IMPI for? is it for the implementation of a pseudo fault tolerance (FT)
> mechanism (we detect that a failure failed, we assume we have a checkpoint
> of
> the running apps, we restart the node w/ IPMI, and then we restart the
> node)? or
> is it for management purpose (i know the node is down, i want to put it be
> online in order to check the logs and the system)?
>
> I think the ultimate goal will change deeply what needs to be done. In
> fact i
> think that if we do not have such an ultimate goal, the GSoC project will
> never
> be really used by the OSCAR community. And it is also clear that if you
> place
> the IPMI in the context of such a goal, a lot has to be done (but i think
> it
> will be a very interesting project).
>
> For example:
> - if we decide to implement some kind of FT mechanism, we need to design a
> global architecture to deal with any kind of failures (even if everything
> won't
> need to be done in that project). We also have to see how that will
> interact
> with HA-OSCAR,
> - if we decide to focus only on the management aspect, we will have to
> come up
> with an idea of what means management in OSCAR (we actually spoke about
> that
> during the OSCAR meeting few weeks ago, i think we have some ideas on the
> topic). The current OSCAR architecture is very naive to deal with the
> management
> issues.
>
> Personally i like the management issues and if someone is selected to work
> on
> that i will be happy to participate in the mentoring effort. And according
> to
> me, all IPMI project will be a difficult but interesting task because IMPI
> by
> itself is not interesting; the way we are using IPMI is interesting.
>
> My 2 cents,
>
> Selon Paul Greidanus <[EMAIL PROTECTED]>:
>
> > abshnasko wrote:
> > > Hi, I am a student interested in undertaking your IPMI feature
> > > implementation. I have done my research on this topic and I have
> experience
> > > with OSCAR, but before I begin the proposal I have a question:
> > >
> > > Am I allowed to use other freely available code or software products
> to
> > > accomplish this project? For example, a good GNU IPMI framework called
> > > FreeIPMI is available, and my current plan is to base my
> implementation off
> > > of their framework. Is this allowed, or does my final product need to
> be
> > > 100% original? Let me know.
> >
> > Hi Travis,
> >
> > Every piece of Oscar is imported from other projects, with OSCAR being
> > the glue to put them together. In light of this statement, I'm pretty
> > sure that using ipmitool, or freeimpi would be a very good place to
> > work. There is enough other work involved with getting it integrated
> > "properly" into OSCAR, that rewriting the ipmi tool would be a waste of
> > time. For example, adding the MAC addresses for the IMPI cards into the
> > oscar database, and setting up username/password pairs for them all.
> >
> > However, making it compartmentalized enough so that someone can choose
> > between impitool and freeimpi would be useful.
> >
> > One thing that isn't in the SoC description, is possibly setting up
> > watchdog timers on the BMC/IPMI cards if they are available..
> >
> > Paul
> >
> >
> -------------------------------------------------------------------------
> > This SF.net email is sponsored by: Microsoft
> > Defy all challenges. Microsoft(R) Visual Studio 2008.
> > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> > _______________________________________________
> > Oscar-devel mailing list
> > [email protected]
> > https://lists.sourceforge.net/lists/listinfo/oscar-devel
> >
>
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Oscar-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/oscar-devel
>
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Oscar-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/oscar-devel
