15:02:06 <mackonstan> #startmeeting FD.io CSIT Project Meeting
15:02:06 <collab-meetbot> Meeting started Wed Oct 28 15:02:06 2020 UTC and is due to finish in 60 minutes.  The chair is mackonstan. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:06 <collab-meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:02:06 <collab-meetbot> The meeting name has been set to 'fd_io_csit_project_meeting'
15:02:08 <jgelety> #info Jan Gelety
15:02:10 <pmikus> #info Peter Mikus
15:02:10 <tifrank> #info Tibor Frank
15:03:44 <vrpolak> #info Vratko Polak.
15:04:50 <mackonstan> #topic Agenda bashing
15:05:34 <mackonstan> #info Juraj, updating the kernels and associated issues discovered
15:05:48 <mackonstan> #info Juraj, seeing one failure in vpp_device tests, would like to discuss
15:06:02 <mackonstan> #topic Inputs from projects - VPP
15:07:13 <dwallacelf> #info Dave Wallace
15:07:21 <mackonstan> #info Dave not much new, VPP 21.01 plan published, need to publish CSIT-2101 plan aligned with VPP's
15:08:41 <mackonstan> #info Maciek, expecting more patches in the memory/heap mgmt area, may be impacting startup.conf and feature configurations esp. related to higher memory settings for scale tests
15:09:10 <mackonstan> #topic Inputs from projects - VSAP
15:09:16 <mackonstan> #info Dave, no updates
15:09:24 <mackonstan> #topic Inputs from projects - TSC
15:10:07 <mackonstan> #info Vratko, GPLv2 relicense, waiting for the last CSIT contributor to sign. Maciek confirmed it's done.
15:10:24 <dwallacelf> #info Damjan working on adding Link Time Optimization to build & Dave W working on updating debian-9 build executor to support this.
15:12:10 <mackonstan> #info Vratko, new project onboarding, may introduce refactor of csit job definitions - WHAAAT?
15:14:45 <mackonstan> #info Dave, the problem is related to the change of the name of code branch: s/master/main/
15:16:25 <mackonstan> #info will be discussed on the TSC call tomorrow
15:18:48 <mackonstan> #topic Releases - CSIT2009
15:19:25 <mackonstan> #info updated report is out, added more test cata
15:19:31 <mackonstan> #info s/cata/data
15:19:34 <mackonstan> #link https://docs.fd.io/csit/rls2009/report/introduction/report_history.html
15:20:39 <mackonstan> #info Tibor, following this week, expecting few more updates in the upcoming weeks, mainly test data incl. nat44ed astf stateful tests
15:20:58 <mackonstan> #info Vrako, RCAs, not much to investigate, main one is vpp l2patch dpdk, wip
15:22:07 <mackonstan> #info Vratko, having some fun with bisecting tooling
15:22:55 <mackonstan> #info RCA is small, so we may not have it mentioned in the report
15:23:31 <mackonstan> #info Vratko, nat44ed udp&tcp cps and pps tests, code is reviewed +1s from Jan and Tibor already
15:24:04 <mackonstan> #info Vratko, waiting for the final verify test runs in rls2009 branch before merging, ETA tonight
15:25:15 <mackonstan> #info Tibor, if all goes well, nat44ed astf test results will be in w45 maintenance version of the report next week
15:25:33 <mackonstan> #topic Releases CSIT-2101
15:25:47 <mackonstan> #link https://wiki.fd.io/view/CSIT/csit2101_plan
15:26:18 <mackonstan> #topic Physical Infrastructure
15:27:04 <mackonstan> #info Peter, Xeon, all in clear.
15:27:42 <mackonstan> #info Tibor, Denvertons, still missing test data for dpdk apps, action is with Yulong/Intel.
15:28:24 <mackonstan> #info Juraj, Thunderx2 x2, vpp_device testbed install completed.
15:28:37 <mackonstan> #info Juraj, tests are passing.
15:29:51 <jlinkes> 4.15.0-55-generic
15:30:00 <jlinkes> 4.15.0-118-generic
15:30:26 <mackonstan> #info Juraj, had to revert to older kernel (4.15.0-55-generic) from the latest (4.15.0-118-generic) with ubuntu 18.04
15:35:04 <mackonstan> #info Juraj, issue is with i40e VF dot1q vlan tag (aka internal tag) handling behaviour changed in the latest kernel, will send packet trace for both working and broken cases.
15:36:21 <mackonstan> #info Juraj, discussed last week in csit call, if csit should periodically update the kernel, the right platform for this is vpp_device
15:38:26 <mackonstan> #info Dave, Juraj, taking action to start a thread about the last item as topic for TWS calls
15:42:22 <mackonstan> #info Juraj, sometimes one random test (1 in 100 in arm platform) fails due to not getting the right info from VPP about the interfaces.
15:44:05 <mackonstan> #info Juraj, will create two csit tickets and send emails
15:45:55 <mackonstan> #info Juraj, will be reusing "old" thunderX for perf tests, reusing existing x86 TG for that
15:48:00 <mackonstan> #info Peter, amd 2n-zn2 testbeds, executing more baseline runs, need core frequency under load to compare, and if all tests passing, however number of iterative tests are failing, need to spend time to categorize the failures, see how much work to identify the issues, or alternatively decide to go ahead with just working tests for csit-2009 test  cycles
15:49:26 <mackonstan> #topic Virtual Infrastructure, CI/CD
15:50:06 <mackonstan> #topic Environment, Framework
15:50:18 <mackonstan> #topic Performance Tests
15:53:58 <mackonstan> #info Tibor, trending jobs still running at low "report mode" frequency, time to change back to normal frequency (except 3n-skx, 3n-tsh, …)
15:55:08 <mackonstan> #info Maciek, some failures observed in csit-report
15:55:17 <mackonstan> #info Vratko, some tests got fixed on vpp side
15:55:37 <mackonstan> #info Vratko, some test failures got addressed on the vpp side
15:55:53 <mackonstan> #topic Functional Tests
15:56:29 <mackonstan> #endmeeting