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

Re: Preliminary launch of ci.rocm.debian.net



Hi all,

TL;DR ci.rocm.debian.net now has most of the core functionality needed!

On 2023-08-02 00:23, Christian Kastner wrote:
> This is more of an alpha release. While performing end-to-end tests over
> the last two days, I noticed the following issues that still need to be
> addressed:
> 
>   (1) There's no automatic scheduling yet. Jobs can only be submitted
>       manually. This will be addressed very soon.

There's now automatic scheduling of tests for all updated (by upload or
migration) ROCm packages AND their reverse dependencies.

For example, an new upload of rocm-hipamd to unstable will trigger
autopkgtests for basically all other ROCm packages, if they have tests.

Triggering our autopkgtests on dependency changes (like libstdc++6) are
much trickier. I've postponed these for now. To me, it was important to
be able to automatically discover what our own uploads break. For now,
we can discover what breaks us simply by manual scheduling of tests.

>   (2) There seems to be a RabbitMQ issue in bookworm; both readers and
>       writers seem to block occasionally.

I still haven't fully figured this out and it's annoying.

>   (3) I've underestimated the resources requirements. Last Sunday,
>       I naively scheduled jobs for unstable+testing for all packages
>       with autopkgtests. Well, some tests run for hours, they even
>       timed out (3h). I need to increase the limit.

We should be fine. Unless I'm mistaken, only changes of rocm-hipamd,
libc, and gcc will trigger large batches of tests. Most other packages
are leaf packages. And on my end, the full batch currently takes about
half a day. (That'll go up once more packages have tests.)

>   (4) I've not looked into getting tests done in experimental, which
>       debci treats as unstable + an extra APT source.

Packages in experimental and bookworm-backports now get tracked and
tested, too!

>   (5) I've deliberately postponed stats (munin) and self-service (API)
>       to a later point in time.

Still postponed.


Next steps
==========

I'll do uploads to unstable and experimental, to verify that these
actually get picked up by ci.r.d.n.

The next big TODO for me is to add autopkgtests to rocm-hipamd, as it's
such a vital dependency of other packages.

The next big issue in general is to solicit more test workers. I have
gfx1030 already running, gfx90c on my desk, and aiming for another one.
But I think it'll soon be time for a much broader call.

The UI on ci.r.d.n needs some tweaking. It's optimized for navigating
tens of thousands of packages, we'd be better served with a more compact
representation. If there are Ruby folks out there interested in helping,
do speak up :)

Best,
Christian

PS: TTBOMK, everything has been pushed to the various salsa repos.


Reply to: