Hi -
> On Oct 29, 2021, at 12:02 PM, Matteo Merli <[email protected]> wrote:
>
> The Pulsar website is getting published through a CI job that updates
> the generated HTML files and commits them in the Pulsar repo, in a
> separate branch ('asf-site'). From there the site is immediately
> visible on the web.
>
> One of the issues with this process is that we have a lot of updates
> of generated HTML files that are growing the size of the Pulsar Git
> repo. Each time we clone, the entire repo has to be fetched by
> developers and users.
>
> This is somewhat made worse by having daily updates in many HTML files
> to update timestamps. I just merged a fix for that
> https://github.com/apache/pulsar/pull/12538 .
>
> The size of the clone git repo is already at 1.4 GB. 90% of this size
> is due to the 'asf-site' branch.
>
> Ideally, we should try to find a solution to use an ad-hoc repo for
> the website deployment, outside the main Pulsar repo.
We can have as many apache/pulsar-* repos as the PMC wants
If we create a pulsar-site repos we can publish from multiple branches.
See GitHub.com/apache/openjpa-site
The main branch could contain website sources.
The asf-site branch would have the built website.
.asf.yaml
publish:
profile: ~
whoami: asf-site
A builds branch could have api docs that seldom change. OpenJPA keeps every
release…
.asf.yaml
publish:
profile: ~
subdir: output/builds
whoami: builds
>
> In the meantime, I propose to truncate the history of the "asf-site"
> branch and squash all commits into a single one, in order to reduce
> the repo size.
+1
>
> Let me know what you think.
>
> Matteo
>
> --
> Matteo Merli
> <[email protected]>
