Community CI Meeting Minutes - February 2, 2023

Community CI Meeting Minutes - February 2, 2023

[Patrick Robb]

[-- Attachment #1: Type: text/plain, Size: 5330 bytes --]

February 2, 2023

#####################################################################
Attendees

1. Patrick Robb
2. Juraj Linkeš
3. Lincoln Lavoie
4. Owen Hilyard
5. Ali Alnubani
6. Aaron Conole

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

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

=====================================================================
General Announcements
* N/A

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

---------------------------------------------------------------------
UNH-IOL Community Lab
* We have begun reporting results for FIPS testing for all new patches and
LTS branches. Currently we are just using the crypto_aesni_mb cryptodev and
AES-CBC test vector. Ideally we will expand to cover other test vectors,
but we are running into some complications:
   * Most test vectors cause a “test vector unsupported” output when
running on the crypto_aesni_mb cyptodev, although the dpdk page for that
documentation suggests it should support other test vectors.
   * AES-GCM test vector will run through and return a cipher text error.
We can explore using different cryptodevs than the one in the DPDK FIPS
sample app user guide, but our understanding is that this cryptodev should
be validating with more than just the AES-CBC test vector - some
documentation may be dated or development needed on the DPDK side.
   * We have documented our problems with using the sample application
HERE. The biggest issue seems the documentation for the sample application
usage is very outdated, the build requirements refer to outdated /
nonexistent commits, and what algorithms are supported by each cryptodev.
We are still exploring this situation but if there are indeed issues with
the sample application itself then FIPS testing coverage will remain
limited in the short term.
* We have set up Broadcom 100G nics for DTS testing on the Altra server
(ARM), and the Broadcom servers (x86), currently publishing DTS functional
tests only. Both test beds have issues with inadequate PCIe slots. These
cards ideally run on Gen 4 x16 slots, however we are not able to provide
adequate PCIe slots with the servers we are running on currently:
   * The Altra server appears to be fine as a DUT, but the tester server is
using gen 3 x8 (form factor x16) pcie slots, which is not adequate for full
performance testing with these nics
   * Broadcom has the same issue except it’s for both tester and server DUTs
   * Two possible solutions include new servers or performance testing with
limited throughput requirements.
   * We have run into a similar problem with enabling the Mellanox 100G
card recently on the Altra server as well.
   * We will continue to review this situation with the aim of discussing
more concrete proposals or solutions in the next CI meeting.
* Work continues with setting up dedicated arm systems for unit testing on
the ThunderX2 servers. We have run into some more complications with
maintaining multi-arch container image manifests in our container registry.
Most images work fine but we continue to bugfix for a few remaining distro
builds. This is the blocker for utilizing these systems right now.
* ZUC/Snow3G: The container images needed to run these unit tests appear to
build fine using our new system, so we plan to bring this testing online
today or tomorrow
* Broadcom-25G instability: We noticed some instability with our
performance results on the Broadcom-25G NIC, and made some CPU
configuration changes including isolating cores for DTS. We are now seeing
much more consistent behavior from this test bed.
* It was a very bad week of weather around UNH last week (January 23) and
we had multiple power blips and outages. Apologies for the inconvenience -
with our migration to a different server rack in the near future with a
battery UPS, these interruptions should be less frequent.
* Mailing list: automated emails to test-report email list may have a
problem with new UNH IT policies, where the lab will be required to enable
DMARC enforcement in the future.  Ali will check on this for the dpdk.org
list server.

---------------------------------------------------------------------
Intel Lab
* N/A

---------------------------------------------------------------------
Loongarch Lab
* N/A

---------------------------------------------------------------------
Github Actions
* Aaron is changing some of the patchwork polling scripts to poll the
events API instead of the series API directly - happening this week.

=====================================================================
DTS Improvements & Test Development
* UNH Lab: the new DTS Hello World patch and Dockerfile patch were run on
community lab servers without issue - further discussion in DTS WG section

=====================================================================
Any other business
* Next meeting is February 16th

-- 

Patrick Robb

Technical Service Manager

UNH InterOperability Laboratory

21 Madbury Rd, Suite 100, Durham, NH 03824

www.iol.unh.edu

[-- Attachment #2: Type: text/html, Size: 7386 bytes --]