Community CI Meeting Minutes - June 13, 2024

Community CI Meeting Minutes - June 13, 2024

[Patrick Robb]

#####################################################################
Attendees
1. Patrick Robb
2. Juraj Linkeš
3. Aaron Conole
4. Dean Marx
5. Jeremy Spewock
6. Manit Mahajan
7. Nicholas Pratte
8. Paul Szczepanek
9. Tomas Durovec

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

=====================================================================
General Announcements
* DPDK Summit in Montreal will be September 24-25:
https://www.dpdk.org/event/dpdk-summit-2024/
   * CFP closes July 21
* Next Wednesday is a federal holiday in the United States, so the UNH
folks other than Patrick (Jeremy, Nick, Dean) wouldn’t be able to
attend the DTS meeting
   * Rescheduled to 13:00 UTC next thursday
* Depends-on:
   * Adam from UNH has begun development for the patchwork server /
git-pw support for Depends-on. The project maintainer has approved his
approach and he addressed questions from a couple other community
members on the github issue.
      * Aaron requested that we start tracking which labs support
depends-on on the dpdk.org testing page, like how we do with retests.
https://patchwork.dpdk.org/project/web/patch/20240531220110.5159-1-probb@iol.unh.edu/

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

---------------------------------------------------------------------
UNH-IOL Community Lab
* Environments added for Compile/Unit tests:
   * Fedora 40
   * Rhel 9
* Template Engine:
   * New versions of python/pip do not allow for pip installs outside
of a virtual environment without an override, per pep 668
https://peps.python.org/pep-0668/. UNH team proposes to shift to using
–break-system-packages to allow for pip installs outside of venv, in
modern distros
      * For these ephemeral containers which only exist to run some
DPDK build, unit tests, etc. the concern of breaking system python
dependencies did not seem very salient
      * Flagging that this also adds a venv requirement to the
linux-setup.sh script, although many users were probably already
running from venv to begin with:
https://git.dpdk.org/dpdk/tree/.ci/linux-setup.sh
* Zlib vdev compression test is now running on an ARM tx2 server in
the lab, and reporting results. So, now we have coverage for both x86
and arm arch.
* SPDK compilation is running in CI:
https://mails.dpdk.org/archives/test-report/2024-June/692342.html
   * This is only for x86_64. We tried arm64 but spdk support is poor
on that side, and almost all distros compile with warnings.
* Dashboard updates:
   * Now posting a retest counter for a patchseries on the report
detail page (reminder, once you hit 3/3, you can no longer request
retests on a series)
* We introduced a bug to our reporting Monday afternoon which resulted
in some missing patchwork contexts. This was fixed yesterday and all
reports re-queued. It appears that we have backfilled all results now.
* Pending:
   * https://patchwork.dpdk.org/project/ci/patch/20240523215945.16468-1-probb@iol.unh.edu/
      * Ali acked v1, but Patrick Robbwill ping him for final
confirmation this is good
   * UNH has staged all the code internally to support this, so we
should be good to go
---------------------------------------------------------------------
Intel Lab
* None

---------------------------------------------------------------------
Github Actions
* No news - it seems to be stable after the fixes implemented a couple weeks ago
* New features like cirrus and retests v2 are forthcoming
* No missing results were reported from the downtime mentioned at the
last meeting

---------------------------------------------------------------------
Loongarch Lab
* Zhoumin indicated that he would be interested in supporting
depends-on for his CI project, once it is supported by the patchwork
project

=====================================================================
DTS Improvements & Test Development
* New Testsuites written at UNH:
   * Vlan
      * Tests vlan filtering, vlan insertion, vlan stripping
      * Filtering works normal for mellanox cx5. I40e and bnxt_en nics
are forwarding on packets with VLAN ids which differ from the vlan
assigned to the rx_port
   * Jumboframes
      * Runs fine
   * Dynamic_queue
      * Is working for rx side.
      * When you change testpmd to tx forwarding mode, it is supposed
to immediately send “a bunch of packets”. This happens the first time
for i40e NICs, but not on subsequent starts, but happens every time on
cx5.
      * Debugging this behavior, though it may necessitate a bugzilla ticket
   * Queue_start_stop
      * Cannot start/stop queues in vectorized RX mode. Need to
determine whether another burst mode is allowed for this NIC, or
whether start/stop is simply invalid on Mellanox.
* Juraj pinged Thomas about merging the 7 patches which are ready, and
he indicated he will be able to take a look at them on Monday.
   * update mypy and clean up:
https://patches.dpdk.org/project/dpdk/list/?series=32026
   * error and usage improvements:
https://patches.dpdk.org/project/dpdk/list/?series=32038
   * Methodology change for hugepage configuration:
https://patches.dpdk.org/project/dpdk/list/?series=32129
   * rename execution to test run:
https://patches.dpdk.org/project/dpdk/patch/20240607083858.58906-1-juraj.linkes@pantheon.tech/
   * node and inheritance improvements:
https://patches.dpdk.org/project/dpdk/list/?series=31808
   * clean up close in remote session:
https://patches.dpdk.org/project/dpdk/patch/20240423122004.75240-1-juraj.linkes@pantheon.tech/
   * testpmd show port info/stats:
https://patches.dpdk.org/project/dpdk/list/?series=32112
* Juraj has finished reviews for all framework patches. He is not
going to review new testsuites right now, so that he can focus
entirely on his capabilities patch.
   * The next main question is to check whether the capabilities are
conditional.
   * I40e NIC was reporting false for a capability (scatter), unless
MTU is increased, then it can report true
* Nick submitted the patch which splits out the execution and node
configurations into separate files
* Nick is also submitting a series which removes some unneeded conf
values, makes some other refactors
   * Refactoring the port configuration is a separate patch - Patrick
is looking at this

=====================================================================
Any other business
* Next Meeting: June 27, 2024