Samba testing on CentOS 8
abartlet at samba.org
Tue Dec 15 05:27:13 UTC 2020
On Tue, 2020-12-15 at 15:40 +1100, Martin Schwenke wrote:
> On Sat, 12 Dec 2020 11:25:35 +1100, Martin Schwenke via samba-
> <samba-technical at lists.samba.org> wrote:
> > On Sat, 12 Dec 2020 09:45:44 +1300, Andrew Bartlett
> > <abartlet at samba.org> wrote:
> > > One advantage of the current bootstrap system is that as long as
> > > nobody
> > > deletes the CI images that are built, the existing images remain
> > > a
> > > stable snapshot. It is 'just' the bootstrap scripts that break
> > > for
> > > others, and of course building a new image now becomes more
> > > difficult.
> > Thanks for hammering that observation home! I found it slightly
> > annoying when I had to build new images to test my branch on GitLab
> > but
> > I now realise that the stability is a core feature.
> Of course this is also a bit dangerous. The worst case is very
> unlikely, but if a change accidentally depends on a 3rd party libray
> bug and all of the test platforms later fix the bug then our test
> framework won't notice the breakage because we're using "stable"
> (i.e. out-of-date) images.
> It makes sense to update our CI images every so often to make sure
> tests still pass on current versions. I'm not sure how we would do
> that. The technical part is quite clear
> but the process part is more interesting. For example, how do
> we do out-of-band CI image builds to ensure that new images are OK
> before the CI system defaults to using them.
So we verify the images are OK with the same task that we deploy on
them, being samba-o3. So you can just rebuild them any time you want
and if they pass (and so will pass the next master build) they will
overwrite the existing images.
> I guess the easiest way would be to add a comment to
> bootstrap/.gitlab-ci.yml like:
> # Rebuild: 1
> (or make it actual YAML data) and increment this for every rebuild.
> That would change the sha1sum so that old images are not overwritten
> and the new images can be confirmed OK before pushing the change with
> the incremented rebuild number.
This would indeed not trigger the overwrite. But we do check.
At the time this was built I really did think the bootstrap system was
over-engineered, but it has been really awesome to be able to rely on
it so comprehensively.
Thanks Joe and Metze!
Andrew Bartlett (he/him) https://samba.org/~abartlet/
Samba Team Member (since 2001) https://samba.org
Samba Team Lead, Catalyst IT https://catalyst.net.nz/services/samba
More information about the samba-technical