[SSSD] [PATCH v2] CI: Allow disabling distro-(in)dependent tests

Nikolai Kondrashov Nikolai.Kondrashov at redhat.com
Fri Sep 12 08:23:48 UTC 2014


On 09/12/2014 10:55 AM, Lukas Slebodnik wrote:
> On (11/09/14 23:32), Nikolai Kondrashov wrote:
>> These options would not be useful much to anyone running CI locally and nobody
>> would need to bother how CI server uses them, so there will be little loss in
>> removing them later.
>>
> The purpose of this patch was to speed up builds. I checked state of our
> consumption of RAM in lab and we are not limited at the moment.
> The current limitation is storage.
>
> The most significant improvement for CI execution will be to use mock tmpfs
> plugin. Mock build root for fedora 21 takes up to 1GiB (818M in my case)
> We can increase RAM for fedaor and rhel build machines from 512 MiB to 2GiB.
> I was told we can use even more.

Ah, it's good to hear that we can use more RAM. That could speed up the CI
overall as well.

>> I promise to fix that ticket some time afterwards, this way or another.
>>
>> I just think that --essential/--moderate/--rigorous would be bearable for the
>> start and it is more important to make server CI available sooner. I might be
>> wrong, though.
>>
>> What do you think?
> So I have very good news for you. With current version of CI script and mock
> tmpfs plugin, we will meets the reasonable requirements for test execution.
>
> So you will have enough time for implementing Michal's proposal (ticket #2425)
> We can start discussion in another thread.

If I understood you right, you're proposing that I should make another patch
instead, which will run mock builds on RHEL and Fedora only, with
corresponding native chroots, and then increase physical RAM available for
corresponding CI VMs. Please correct me, if I'm wrong.

That would mean that to run both mock builds locally developers would have to
run CI both on Fedora and RHEL, or we'd need to provide options controlling
that.

That solution would also still leave Clang analyzer (taking 5-9 minutes)
running on all CI hosts and wouldn't accomodate for separating further
distro-independent tasks.

Lastly, another solution will take time to implement, test and review. Whereas
this solution and server setup for it is ready, tested and can go live within
15 minutes.

All-in-all, I think that going with the presented patch is best.

Nick



More information about the sssd-devel mailing list