I was wondering pretty much the same thing.. Getting Lustre out into the
open source community would be something that a lot of Clusterfs's
paying customers would like. I know DOE for one has mandates to support
open source projects.

Agreed, I work for a DOE funded lab that uses lustre and I know others
have asked for anonymous cvs access in the past and have just been
ignores by the cfs guys.

This is one thing I don't understand, I have experience with open
source projects and know full well the benefits of providing anonymous
scm access to the latest development trees that the developers are
working on. Latest development trees include side projects and
exploring potential forks in the code.

Providing the development trees to the customer means the support
teams for the clusters that are both in a production role and for
experimental use can more easily maintain their clusters.  We wouldn't
have to bounce back to the lustre developers or attempt to search
bugzilla for solutions to our problems, those avenues are available to
us but providing anonymous scm access is yet another way we can find
solutions without having some form of human initiated response with
the developers. I'm sure most maintainers of production systems know
what it means to have that production system go down and sitting on
your thumbs waiting from support for third party vendor to respond.

Also I'm sure the cfs guys have seen many more bugs and know what the
outputs look like than we do, and I know they've gotten repeat bugs
for versions of their software both in bugzilla and on the mailing
lists. Instead of answering the questions with the same patch over and
over you can provide an scm commit id (for whatever scm you are using)
and keep the emails short. Providing the commit id also means
consistency in the patch you provide that will fix our problems.

Also providing development trees freely to the customer provides the
customer with insight as to where you are going with your software.
Most of us are fairly well adept in writing C code and understanding
the bigger pictures. We also can provide feed back and inevitably code
for free to the head of the scm.  Which gives the cfs guys a better
picture of what their customers are looking for in the future, and a
little bit of free development.

Tracking also works a lot better when you provide anonymous scm
access.  Maintainers of the clusters have one centralized place to
look for code solutions to the clusters they manage. So providing an
scm commit id instead of patches saves us time when searching for
kernel panics or bugs. If we find the problem we are experiencing and
are given an scm commit id to look at we can then push the patches
around locally and then resubmit to the devel mailing list when we
make it work.

From my experience you either go open source all the way, or close
everything off, only going half way in between just frustrates people
more...

- David Brown

_______________________________________________
Lustre-discuss mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-discuss

Reply via email to