Re: Jenkins and distro choice

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Jan 5, 2022 at 9:39 AM Daniel Gryniewicz <dang@xxxxxxxxxx> wrote:
>
> We are getting ready to merge our first contributed Zipper Store, and
> we'd like to have it build by default in Jenkins, so that we don't
> accidentally break it all the time.  We have RPMs, but not DEBs.  This
> brought up the question for us:  Does Jenkins use both Centos and
> Ubuntu?  If so, how does it pick which one it's using?  Is there some
> way we can limit RGW-related builds to Centos, or do we need to get debs
> from our contributors?

I think I've missed something about Zipper. Why does it depend
directly on the distribution choice?

Okay, guess this is about PR https://github.com/ceph/ceph/pull/44379

I built a couple different CephFS plugins for other projects back in
the day and am trying to remember why this wasn't an issue for those.
Hadoop just decided they weren't taking any more external FS plugins
around then, so I guess they never even tried to build it. My guess is
Hypertable's testing was also pretty limited at that point. For
ourselves, we just pulled down a given Hadoop release to do our builds
and the patchy automated testing that happened.

Ceph distributes upstream packages for Ubuntu and CentOS, and there's
no plan to change that. In my opinion, anything which is part of Ceph
(which it needs to be, if it's getting tested in the Ceph
infrastructure) needs to be built and buildable for both those classes
of packaging.

That said, if the Zipper architecture requires modifying Ceph code to
run external plugins, that sounds like a world of hurt. I haven't seen
any of the implementation details before so this certainly isn't a
good time for the MOTR guys, but I'd encourage the RWG team to think
about alternative integration models there if you anticipate more
third-party backends. Something that requires dropping a shared
library into a folder and detects them as options and blindly passes
through configs based on a defined naming scheme, instead of needing
to maintain an in-tree if-else tree and config definitions, is a lot
more scalable/maintainable. Plus it keeps a good interface boundary
for things like testing concerns...
-Greg


>
> Thanks,
> Daniel
>
> _______________________________________________
> Dev mailing list -- dev@xxxxxxx
> To unsubscribe send an email to dev-leave@xxxxxxx
>

_______________________________________________
Dev mailing list -- dev@xxxxxxx
To unsubscribe send an email to dev-leave@xxxxxxx



[Index of Archives]     [CEPH Users]     [Ceph Devel]     [Ceph Large]     [Information on CEPH]     [Linux BTRFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux