DPDK CI discussions
 help / color / mirror / Atom feed
From: Patrick Robb <probb@iol.unh.edu>
To: ci@dpdk.org
Cc: dev <dev@dpdk.org>, dts@dpdk.org
Subject: Community CI Meeting Minutes - October 3, 2024
Date: Wed, 23 Oct 2024 01:50:36 -0400	[thread overview]
Message-ID: <CAJvnSUBvkjtz2OkagasBggs01_enFH807sy-Y2D6KQREtvcqTw@mail.gmail.com> (raw)

#####################################################################
October 03, 2024
Attendees
1. Patrick Robb
2. Luca Vizzarro
3. Aaron Conole
4. Ali Alnubani
5. Paul Szczepanek
6. Adam Hassick

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

=====================================================================
General Announcements
* DPDK Summit Montreal 2024 has concluded
   * There was some good engagement on CI topics following the CI
Testing talk, with particular interest in:
      * Creating code coverage reports more frequently, posting
coverage in an embedded fashion or in a better location
      * Patchwork: Maxime requested that since Adam is working on
updating the patchwork project to include DPDK’s depends-on
“protocol,” he could also add support for the “series-reviewed-by” tag
which is used by some DPDK maintainers, such that patch files inherit
the reviewed by tag when you download the mbox.
         * Thomas suggested Patrick write a blog post about how the
DPDK project is utilizing patchwork for CI checks
      * The new cryptodev testing at the lab
      * There was good interest from community members for getting
more hardware into the lab, both existing gold members and entities
who may have an interest in getting dpdk membership
   * A new DPDK project, grout (a graph router based on dpdk) was
announced and demonstrated. It is used for simulating a router for
VNF/CNF functions for testing.
      * It is worth it for the CI group to engage with David Marchand
and Robin Jarry to see what level of interest they have in CI lab
engagement for this project.
   * Luca gave a talk about how to write testsuites / run DTS, which
was helpful for people in the community
      * Some updates for the README and dts .rst are needed
      * It would be a benefit to write some blogs / youtube videos for
setting up and running DTS
         * Probably we should just screencapture and go through the
whole process
* Aaron Conole, Thomas Monjalon, and Ben Thomas visited the UNH Lab on 9/26
   * There was a lot of good discussion for the group here, including
topics like:
      * How to effectively engage with hyperscalers to convince them
to stand up patchwork reporting CI test labs for DPDK
      * How to resolve ambiguities in DPDK policy/specs when writing a
testsuite (the example used was MTU/max packet len mismatch between
PMDs
      * Thomas suggested an initiative - maintaining a dynamic page
under dpdk.org/testing which maintains information regarding lab
coverage and downtime
         * MVP would include only mainline
         * LTS testing could be considered

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

---------------------------------------------------------------------
UNH-IOL Community Lab
* NVIDIA:
   * ConnectX6 was restored to CI testing for performance testing
   * ConnectX7 was added to production for the first time, now that
the single core forwarding results table is 1:1 with NVIDIA’s report
      * The “issue” that IOL was running into came from the
numa_balancing=disable kernel option which AMD recommends. It appears
to cause some initial performance oscillations in the short term (at
least) which dramatically affects the performance test. This was
removed and so the problem was “solved.”
         * It’s not clear whether this would/would not be resolved by
simply testing for longer, dropping more of the highest/lowest
measurements before calculating the average. I.e. maybe these
oscillations are only in the beginning tens of seconds after
forwarding starts.
      * In any case, this is now online on the new AMD servers
      * Patrick and Ali will pick up this conversation and loop in AMD
people if needed
* Nathan and Patrick send AMD legal some clarifications regarding the
performance testing at UNH and the server donations are now approved
* Manit is adding some patches to the CI mailing list
   * The first (send yesterday) adds support for the dts tag in the
patch_parser.py script. This is needed for conditionally running the
dts check format script and reporting a result to patchwork
      * Patrick added a tag - this should be good to merge
   * The second updates create_series_artifact.py and artifact.yml to
allow for the user to optionally provide a devtools/check_meson.py
patch to artifact.yml. If they do so, that script will be run in the
create artifact script and a check_meson pass/fail value will be added
to the .properties file which is output by the create_series_artifact
script.
      * Will submit this today or tomorrow
* Legacy DTS:
   * Action items for Patrick:
      * Update the vfio modprobe options patch for the cryptodev test
      * Add the CX7 card to DTS inventory (the locally running version
at UNH has this change) - need to upstream this
      * Submit the patches which were needed to add Octeon CN10K support
* Patrick needs to prepare a retrospective and start working on the SOW process
   * Honnappa has expressed some interest in working on DTS
proposals/SOW items soon

---------------------------------------------------------------------
Intel Lab
* Patrick can share his current Intel lab contacts with Aaron

---------------------------------------------------------------------
Github Actions
* Aaron is going to export his monitoring flow to the pw-ci project
   * These are the scripts which monitor patchwork, includes some
particular behavior for DPDK patchwork, OvS patchwork, etc.
* Cirrus CI support is ready to be merged into mainline
   * Can then work on the cirrus yaml to add support for freeBSD, etc.
   * Is not very flexible for per-patch testing.
      * Aaron is thinking of some workarounds like sending/processing
patches slowly with some gap in between
      * Can do series testing for now

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

=====================================================================
DTS Improvements & Test Development
* Juraj merged some patches on his final day working on DTS/maintainer
   * Capabilities patch series
   * XML-RPC server replacement
* UNH students have been rebasing their series on next-dts - Patrick
needs to start reviewing more aggressively and pushing what is ready
to the remote repo
* Let’s coordinate roadmap review effort:
   * status DTS 24.11 Roadmap
   * Patrick needs to add links to series
* Patrick will update the MAINTAINERS file to add
* Patrick should share the 2024 DTS SOW with Paul

=====================================================================
Any other business
* Next Meeting Oct 17, 2024
   * Aaron will not be around for this meeting due to vacation time,
but will send an update

                 reply	other threads:[~2024-10-23  5:52 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAJvnSUBvkjtz2OkagasBggs01_enFH807sy-Y2D6KQREtvcqTw@mail.gmail.com \
    --to=probb@iol.unh.edu \
    --cc=ci@dpdk.org \
    --cc=dev@dpdk.org \
    --cc=dts@dpdk.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).