[xcat-user] New xCat user experience

Sun Nov 1 07:07:55 MST 2009

Stuart,  the top doc is suppose to be your start point and give you links 
to the important documentation available.   Unfortunately,   like most 
documents it has grown over time,  but is still a good place to start and 
should link  you to the other documents such as  the  document 
specifically on setting up  iDataplex system .      A great source of 
information is the mailing list.    As Vallard notes,  we are constantly 
evaluating our documentation,  so thanks for the comments. 

Lissa K. Valletta
414/3-8
Poughkeepsie, NY 12601
(tie 293) 433-3102

From:
Vallard Benincosa <vallard at benincosa.com>
To:
xCAT Users Mailing list <xcat-user at lists.xcat.org>
Date:
10/31/2009 02:40 AM
Subject:
Re: [xcat-user] New xCat user experience
Sent by:
xcat-user-bounces at lists.xcat.org

Stuart,
Great email.  I think most your points require people to address them.  I 
will try to address what I can.  

The documentation issue is something I hope we can address before the end 
of the year.  I know it is one of my top priorities.  One reason I redid 
the web page was to make it more clear.  (as you can tell, I'm a wannabe 
web designer)  I think its better, but still needs work, and I hope to 
improve upon that.  I added the "Getting Started" link with the hopes to 
get things running better.  I don't want to make it too big, but perhaps a 
brief discussion on the architecture is warranted.  

I think all the developers need to know this as well:  Nobody wants to 
spend a lot of time learning how to use your product.  If you can't make 
it work quickly and automatically with sound documentation no one is going 
to use it.  

With that said, I hope that once you start using xCAT you can give more 
feedback as to what else needs improvement. 

1.  We were going back and forth about the links today.  I mentioned to 
the team that the https weren't working for me and they said the http 
links weren't working for them.  Sometimes it would, sometimes it 
wouldn't.  I'll investigate more.  But thanks for bringing that up.

2.  The mime/types I'm not sure how to do in html, do you know what I 
should put in the tag to make it come up correctly? 

Thanks for taking the time to write down your thoughts.  Let us know how 
your testing goes and if you need any support.

On Fri, Oct 30, 2009 at 3:46 PM, Stuart Barkley <stuartb at 4gh.net> wrote:
We have recently ordered an IBM iDataplex system with 252 nodes and
are expecting to use xCat with it.  It's still in manufacturing and
won't be delivered for a while yet, but its time for me to begin
getting familiar with xCat.

I plan to setup a small environment for testing purposes.

As a new user there seems to be a large initial step to understanding
xCat.  These notes are some of my initial observations about the
available xCat information.  Hopefully, these comments can help
improve the available information.

The IBM Redbooks spend a lot of time talking about moving from CSM and
using AIX.  Although there is some good information in them, they
don't translate very well to a new user planning a Linux based
implementation.

I've noticed some web page updates over the past few days.  An earlier
problem today with https vs http urls on the main pages seems to have
been corrected.  (The https versions required a login which I don't
currently have or want.)

The .pdf files on the web site don't get correct mime-type information
(they should be application/pdf).  For example, references from the
main page http://xcat.sourceforge.net/#3 the files referred to in "See
the official in depth xCAT Documentation Repository" are incorrectly
tagged as application/octet-stream.  "Top Document" is also
incorrectly tagged but for some reason "xCAT Advanced" gets tagged as
text/plain which works better with my Firefox browser.

The documentation is missing introductory information and basic
architecture information.  It appears that xCat requires or at least
assumes a complex network model with segregated networks.  We will run
some form of that for our HPC, but I was really hoping to use xCat to
handle all of our system provisioning needs, including systems on a
more flat network model.  My initial preferred test environment would
be a flat network with existing DNS and DHCP servers.

The documentation is missing basic security architecture information.
There are several incomplete sections at:

http://sourceforge.net/apps/mediawiki/xcat/index.php?title=XCAT_2_Security

There seems to be an internal CA used for some purpose.  We need
information about the purpose of this CA and how to integrate xCat
into our own CA.  There appears to be some form of user authentication
based upon this CA, but at this point in time I don't see what this
additional complication adds over standard unix and ssh
authentication.

There also appears to be some automated ssh key management.  I'm
hopeful for a good ssh key management scheme, but don't see any design
information for what xCat provides and uses.

The mailing list archives have a large number of messages which have
been stripped of multipart/alternative information making the archives
largely unusable.  In particular, it looks like Lissa Valletta has
lots of good information which is not in the archives.  This stripping
was also performed in the gziped mbox formatted archives
(http://xcat.org/pipermail/xcat-user/).  It would be very good if
proper mbox formatted archives could be retrieved from any underlying
mailman archives.  It looks like today there has been some work
towards a new mailing list archiving system.  Please keep good full
mbox format archives available.  Being able to download historic email
archives is important to not repeat common questions on the mailing
list.

The mailing list archives show a large number of "out of office"
messages.  I estimated about 5% of the messages for the past year are
of that form.  Most of them seem to come from the same employer.  Not
sending "vacation" messages to mailing lists was solved years ago, it
would be good for that employer to update their mail system to
something which works correctly with mailing lists.

Build and release distribution is not up to what I normally expect.
I prefer to inspect source code prior to installation.  I also prefer
to build from released source.  There are no source tar balls.  These
are important in providing reproducibility and traceability
functionality.  There are not even release tags in the subversion
sources.  To find the sources for a particular release would appear to
require guessing about a specific subversion revision number.

I'm less comfortable with multiple remote repositories.  I also am a
little troubled by the apparent need for xCat to provide its own
versions of some tools.  I'm sure there is some reason a different
ipmitool is needed.  I haven't explored this area fully yet and will
just use the modified versions for testing.  I would like to see
better integration before doing a production install.

Stuart Barkley
_______________________________________________
xcat-user mailing list
xcat-user at lists.xcat.org
http://www.xcat.org/mailman/listinfo/xcat-user
_______________________________________________
xcat-user mailing list
xcat-user at lists.xcat.org
http://www.xcat.org/mailman/listinfo/xcat-user

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.xcat.org/pipermail/xcat-user/attachments/20091101/23ca5539/attachment.htm