> Another problem with the dashboard API check in its current form
> is wasted lab resources [1].  Perhaps someone from the dashboard
> team could assist David there.

Ilya, when we asked David to enable this Jenkins job on all PRs, the first thing we did was to check with David and ensure we had enough lab resources to cope with running this for all PRs.

That said, we'll provide all support needed to improve the build workflow (either by saving workspaces/using pipelines/running from packages/whatever).

Coming back to the topic of the stability of the API test, this is the 60-day chart of build statuses for 'make check' vs. Ceph API test.

image.png

image.png

As you may see, the failure rate & build count is lower for API test job, and that considering that build failures at "make check" will result in API test failures.

While "make check" errors are usually more informative for the PR author, API test failures can also be useful for detecting & improving Ceph-wide issues (as Kefu did recently: https://github.com/ceph/ceph/pull/35871).

Nevertheless, we'll try to do our best to improve the test reliability. As Lenz already pointed out, please feel free to contribute with ideas & suggestions for improving API tests: https://pad.ceph.com/p/dashboard-api-test-improvements

Kind Regards,

Ernesto


On Thu, Jul 9, 2020 at 11:15 AM Ilya Dryomov <idryomov@gmail.com> wrote:
On Thu, Jul 9, 2020 at 11:07 AM Lenz Grimmer <lgrimmer@suse.com> wrote:
>
> Hi Dan,
>
> On 7/8/20 9:58 PM, Dan Mick wrote:
>
> > I've temporarily removed the requirement the above check, as it's
> > failing often on unrelated pull requests.  Once it gets back healthy we
> > can reenable it (it still runs, it just won't cause the pull request to
> > be blocked).
>
> Thanks for the information and sorry if this test was causing issues.
> Have these failures been reported on the tracker? It would be really
> helpful if we could address these failures to make the tests more stable
> and keep them enabled.
>
> The night tests that perform the API tests on master have been passing,
> so we should take a closer look at why they're failing on these pull
> requests:
>
> https://jenkins.ceph.com/view/all/job/ceph-api-nightly-master-backend/
>
> In general, please reach out to the dashboard developers if you're
> facing issues with the test, we're glad to help to get them resolved.
>
> Please keep in mind that while it's currently called "Dashboard API",
> the intention is to establish the dashboard backend as the official Ceph
> REST API (as a successor of the "restful" module), to be used for
> managing and monitoring all aspects of Ceph (with or without the
> Dashboard UI). We'll rename the test on the pull requests to "Ceph API"
> to make this more clear.
>
> See https://pad.ceph.com/p/rest-api-stability-requirements and
> https://pad.ceph.com/p/dashboard-api-test-improvements for some
> background information and https://tracker.ceph.com/issues/40907 for an
> overview of all ongoing work around the Ceph API.

Another problem with the dashboard API check in its current form
is wasted lab resources [1].  Perhaps someone from the dashboard
team could assist David there.

[1] https://lists.ceph.io/hyperkitty/list/dev@ceph.io/thread/KMV3HZE6IPEGWY5WP4XXPEKZ225RB3XH/

Thanks,

                Ilya
_______________________________________________
Dev mailing list -- dev@ceph.io
To unsubscribe send an email to dev-leave@ceph.io