Re: Staging website at cassandra.staged.apache.org

[Murukesh Mohanan <mu...@gmail.com>]

For clarification, when you say "this would clean the master
branch history", would the content directory be removed from past commits
using bfg or similar tools?

(I'm fine with either way; just curious.)

On Fri, 1 May 2020, 07:12 Anthony Grasso, <an...@gmail.com> wrote:

> Hi everyone,
>
> Thanks to hard work by Mick every push to the cassandra-website *src/*
> directory in master is now automatically deployed to
> https://cassandra.staged.apache.org/. The automation is carried out by a
> Jenkins job at https://ci-cassandra.apache.org/job/cassandra-website/.
> This
> is really useful because it allows us to preview our changes in the staging
> site before we push them to the production site!
>
> With the above change in place, the process to deploy a change to the
> production site is.
> 1. Commit your change to the *src/* directory in the master branch. Jenkins
> will now automatically generate the (HTML) site pages in *content/*
> directory on the asf-staging branch.
> 2. Preview your changes on the staging site:
> https://cassandra.staged.apache.org/.
> 3. If the changes look good, then merge the *content/* directory on the
> asf-staging branch to the master branch using:
>
>
> $ git merge asf-staging
> $ git push
>
>
> An issue with the above process is the master branch history is now going
> to get polluted with the auto-generated content commits. Even without the
> Jenkins automation, the process to publish to the production site still had
> the issue where commits to the master branch included both the *src/* and
> *content/* directory changes. A small one line change to the *src/*
> directory, could result in changes to hundreds of pages in the *content/*
> directory.
>
> Rather than serving the production site from the *content/* directory on
> the master branch, is there any objection to serving it from the asf-site
> branch? This would mean that the last step in the above process would be
> performed on the asf-site branch rather than the master branch. In
> addition, there would be no need to have a *content/* directory on the
> master branch. The *content/* directory from which the production site is
> served would live in the asf-site branch. This would clean the master
> branch history, hiding the generated *content/* directory and the unwieldy
> content changes generated by Jenkins user.
>
> Let me know what you think about the proposed change.
>
> Regards,
> Anthony
>
> On Tue, 21 Apr 2020 at 16:40, Mick Semb Wever <mc...@apache.org> wrote:
>
> > For our cassandra-website repository, any changes to our website can now
> > first be staged at https://cassandra.staged.apache.org/
> >
> > The staged website comes from the content/ directory on the `asf-staging`
> > branch.
> >
> > regards,
> > Mick
> >
>

Re: Staging website at cassandra.staged.apache.org

[Anthony Grasso <an...@gmail.com>]

Hi Murukesh,

That is a good question. To clarify, I meant that if we changed the site to
be serviced from the 'asf-site' branch the master branch would be cleaner
from that point forward.

However, now that you mention it, I do like the idea of cleaning the entire
master branch. Given that this alters the repository history, we might want
to do such a change in a phased approach. e.g.

1. Change production site to be served from 'asf-site' branch.
2. Make a copy of the master branch called master_archive (or something
similar) - this can be done sometime later after Step 1.
3. Clean master branch using bfg.
4. Keep the master_archive branch until we think it is no longer needed -
we could keep the branch anywhere from a few days to a few years.

Let me know what you think.

Regards,
Anthony

On Fri, 1 May 2020 at 18:18, Murukesh Mohanan <mu...@gmail.com>
wrote:

> For clarification, when you say "this would clean the master
> branch history", would the content directory be removed from past commits
> using bfg or similar tools?
>
> (I'm fine with either way; just curious.)
>
> On Fri, 1 May 2020, 07:12 Anthony Grasso, <an...@gmail.com>
> wrote:
>
> > Hi everyone,
> >
> > Thanks to hard work by Mick every push to the cassandra-website *src/*
> > directory in master is now automatically deployed to
> > https://cassandra.staged.apache.org/. The automation is carried out by a
> > Jenkins job at https://ci-cassandra.apache.org/job/cassandra-website/.
> > This
> > is really useful because it allows us to preview our changes in the
> staging
> > site before we push them to the production site!
> >
> > With the above change in place, the process to deploy a change to the
> > production site is.
> > 1. Commit your change to the *src/* directory in the master branch.
> Jenkins
> > will now automatically generate the (HTML) site pages in *content/*
> > directory on the asf-staging branch.
> > 2. Preview your changes on the staging site:
> > https://cassandra.staged.apache.org/.
> > 3. If the changes look good, then merge the *content/* directory on the
> > asf-staging branch to the master branch using:
> >
> >
> > $ git merge asf-staging
> > $ git push
> >
> >
> > An issue with the above process is the master branch history is now going
> > to get polluted with the auto-generated content commits. Even without the
> > Jenkins automation, the process to publish to the production site still
> had
> > the issue where commits to the master branch included both the *src/* and
> > *content/* directory changes. A small one line change to the *src/*
> > directory, could result in changes to hundreds of pages in the *content/*
> > directory.
> >
> > Rather than serving the production site from the *content/* directory on
> > the master branch, is there any objection to serving it from the asf-site
> > branch? This would mean that the last step in the above process would be
> > performed on the asf-site branch rather than the master branch. In
> > addition, there would be no need to have a *content/* directory on the
> > master branch. The *content/* directory from which the production site is
> > served would live in the asf-site branch. This would clean the master
> > branch history, hiding the generated *content/* directory and the
> unwieldy
> > content changes generated by Jenkins user.
> >
> > Let me know what you think about the proposed change.
> >
> > Regards,
> > Anthony
> >
> > On Tue, 21 Apr 2020 at 16:40, Mick Semb Wever <mc...@apache.org> wrote:
> >
> > > For our cassandra-website repository, any changes to our website can
> now
> > > first be staged at https://cassandra.staged.apache.org/
> > >
> > > The staged website comes from the content/ directory on the
> `asf-staging`
> > > branch.
> > >
> > > regards,
> > > Mick
> > >
> >
>