Community CI Meeting Minutes - July 6, 2023

Community CI Meeting Minutes - July 6, 2023

[Patrick Robb]

[-- Attachment #1: Type: text/plain, Size: 6523 bytes --]

July 6, 2023

#####################################################################
Attendees
1. Patrick Robb
2. Manit Mahajan
3. Lincoln Lavoie
4. Aaron Conole
5. Ali Alnubani
6. Juraj Linkes
7. Jeremy Spewock

#####################################################################
Agenda
1. General Announcements
2. CI Status
3. DTS Improvements & Test Development
4. Any other business

#####################################################################
Minutes

=====================================================================
General Announcements
* Ali has created a dpdk-ci project on dpdk patchwork, allowing for better
tracking of incoming patch series

=====================================================================
CI Status

---------------------------------------------------------------------
UNH-IOL Community Lab
* On Monday a patch was merged into main which changes behavior relating to
the libnuma dependency for compiling dpdk. This tripped up our apply
patchset + compile pipeline, and as a consequence we began publishing build
error results for unrelated incoming patch series. We have made the libnuma
dependency change needed to return to normal CI, and we are working now to
backfill with reruns for all affected patch series.
   * Unfortunately this coincided with USA’s independence day, which is why
we couldn’t address this “same day”
   *
https://git.dpdk.org/dpdk/commit/?id=5afed6966da072e7f569871d86421e70c2326601
   * No results were published to patchwork from IOL due to it failing at
the apply patchset+compile stage failing. So, there may not have been
awareness of this matter from the developer/maintainer side.
      * This also makes submitters think they are still waiting on results
to come in
   * After applying patchset in our script, we do a sanity compile before
passing the patch series to our CI. This might be causing more harm than
good by preventing results from being published.
   * This connects to the conversation from the previous meeting regarding
setting up an automated process for checking that all expected results have
been published to patchwork
      * Should this be handled from the lab side, or the patchwork project
side?
      * IOL could send up an initial report (unh-iol-testing), which can
display a failure if a build error occurs, and can report a pass once we
see all (ci wide) results are up.
* Jeremy’s patch series for refreshing NIST json web tokens has been
approved and merged into upstream dpdk-ci
* We are continuing to lay the groundwork for correctly re-triggering
testing pipelines per email based retesting requests
* Adam is exploring the tsfactory testing framework by setting it up on our
development servers. This is a tech board request which may allow for
expanded ethdev related testing.
* We are internally testing usage of gcov/lcov to produce unit test
coverage reports. Initially this would just be tracking coverage related to
the “fast tests” meson unit test suite, but could possibly be expanded
later. Discussions regarding the most appropriate use of the tool are
ongoing, but we could run this regularly and publish results if desired.
Or, they may just serve as a helpful internal tool and reference point.
   * This can be regularly published to the dashboard (1x week/month or
something)
   * Can we have dedicated runs which cover all built in unit test suites?
   * Can we expand this to DTS, etc?
   * Gcov can “self-aggregate” by adding to an existing file
   * There are tools which can read these and show not only how much code
is being covered, but also which code
      * Lab will look into cobertura
* We’ve added a system to our CI dedicated for running containerized
report/saving type jobs, to avoid situations where heavy testing load means
all of our container runners are “busy” and results which are ready to be
reported have to wait in queue for executor slots
   * This is mostly a quality of life change reducing time from patch
submission to reporting, but it also helps prevent situations where our CI
“locks up”
* NIC hardware refresh:
   * 2 Intel E810s are in, DAC cabling has arrived, this NIC will be
installed in the to be donated Intel server
   * 1 of the 2 mlnx cx6 cards are in, the other is backordered
   * No news from Ericcson about the embedded snow board
   * Additional E810 has arrived to act as a traffic generator for Marvell,
still waiting on them to ship their hardware
   * Everything else is ordered

---------------------------------------------------------------------
Intel Lab
* None

---------------------------------------------------------------------
Loongarch Lab
* None

---------------------------------------------------------------------
Github Actions
* Aaron has some code which will scan comments for a reported context
   * Right now it scans on the open patches for comments - he is working on
refactoring this for the events API
   * Should there be a universal script for this which all labs use, or
should every lab write their own script?
      * Aaron is using https://github.com/ovsrobot/pw-ci currently for
other patchwork based projects. It may be possible for our script to be
integrated here, preventing some work overlap between labs.
* Also writing the script for kicking off the required retest according to
the result of calling the events API

=====================================================================
DTS Improvements & Test Development
* Some 23.07 DTS roadmap patches are still not merged.
   * Jeremy has been testing patches on UNH’s development test bed servers,
and is adding “tested-bys” for those series. That work will continue today.
* Release has been postponed a few days, we should have time until July 17th
* Juraj is sending his final patch submission soon, and has some comments
for Jeremy’s patches which he can send in today
   * A sync is needed regarding how to manage overlap between patches
* We will schedule a call between DTS folks to discuss final steps before
the 23.07 release

=====================================================================
Any other business
* Aaron still waiting for confirmation that the talk regarding making the
community lab a gating mechanism for patches
   * Also will just update the community regarding some of the challenges
the lab faces
* Next meeting is Jul 20, 2023

[-- Attachment #2: Type: text/html, Size: 6891 bytes --]