14:03:27 #startmeeting FD.io CSIT Project Meeting 14:03:27 Meeting started Wed Oct 14 14:03:27 2020 UTC and is due to finish in 60 minutes. The chair is mackonstan. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:03:27 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:03:27 The meeting name has been set to 'fd_io_csit_project_meeting' 14:03:28 #info Jan Gelety 14:03:30 #info Tibor Frank 14:03:31 #chair 14:03:31 Current chairs: mackonstan 14:03:38 #info Dave Wallace 14:03:38 #info Vratko Polak. 14:04:17 #topic Agenda bashing 14:04:25 #topic Inputs from projects - VPP 14:05:53 #info Andrew, VPP v21.01 rls schedule published 14:05:56 #link https://wiki.fd.io/view/Projects/vpp/Release_Plans/Release_Plan_21.01 14:14:22 #info Vratko, it makes sense to keep CSIT-2101 schedule with VPP v21.01 14:14:44 #info Vratko, it will give CSIT time to re-test v2009 in the updated CSIT environment 14:16:33 #info Andrew, Maciek, per discussion on the VPP call yesterday, CSIT env will be moving from 2MB to 1GB hugepages to accommodate VPP physical memoryt alloc improvements done by Damjan 14:16:45 #info Dave, daily auto-gen VPP rls notes 14:16:53 #link https://github.com/vpp-dev/vpp-release-notes/blob/master/RELEASE.md 14:17:08 #link https://github.com/vpp-dev/vpp-release-notes 14:17:24 #topic Inputs from projects - VSAP 14:17:28 #info no inputs 14:17:36 #topic Inputs from projects - TSC 14:17:58 #info Vratko, main thing for CSIT is status of GPLv2/Scapy 14:18:24 #info Awaiting two contributors to give their permission 14:19:52 #topic Releases - CSIT-2009 14:20:30 #link https://git.fd.io/csit/commit/?id=4bafd0aa54fa78c4c3a6a32281cfa551e61bcafc <-- Contribution from Lucian (Enea) to be replaced by a GPL-able contribution. 14:21:05 #info Tibor, most testbeds done, 3n-skx coverage tests are behind 14:25:48 #info Tibor, CSIT-2009 report is camera ready, few formatting nits will be generatedshortly 14:26:00 #link https://docs.fd.io/csit/rls2009/report/index.html 14:27:08 #info Tibor, target is to add following test data next week: nat44ed CPS UDP and TCP tests, more of 3n-skx coverage tests 14:27:27 #info Vratko, nat44ed pps/throughput tests most likely in 2 weeks time 14:29:37 #info Tibor, Andrew, Maciek: highlight the most important things, incl. multi-core scaling with 100GE mlx NICs 14:30:13 #info Tibor also RCA will be coming in the next weeks 14:31:27 #info Tibor also RCA will be coming in the next weeks 14:31:51 #info Tibor also RCA for regressions/progressions of v20.09 vw. v20.05 will be coming in the next weeks 14:32:05 #topic Physical Infrastructure 14:33:05 #info Tibor, 2n-dnv, 3n-dnv dpdk tests failing, will be looked at by Yulong/Intel 14:33:51 #info Juraj, thunderx2 new vpp_device setup is almost done 14:36:09 #info Keesang, all bios and ucode are aligned on both servers 14:36:13 #topic AMD EPYC 7532 zen2 server settings in FD.io CSIT labs 14:36:13 #info amd 2n-zn2 BIOS settings 14:36:13 #link https://gerrit.fd.io/r/gitweb?p=csit.git;f=docs/lab/testbeds_sm_zn2_hw_bios_cfg.md;hb=refs/changes/33/29033/2#l483 14:36:13 #info zen2 servers firmware inventory 14:36:13 #link https://gerrit.fd.io/r/gitweb?p=csit.git;f=docs/lab/testbeds_sm_zn2_hw_bios_cfg.md;hb=refs/changes/33/29033/2#l627 14:36:13 #info amd zen2 GRUB cmd line: 14:36:13 #info GRUB_CMDLINE_LINUX="isolcpus=0-15,16-31,32-47,48-63 rcu_nocbs=0-15,16-31,32-47,48-63 transparent_hugepages=never iommu=pt amd_iommu=on nohz=off numa_balancing=disable nmi_watchdog=0 audit=0 nosoftlockup processor.max_cstate=0" 14:38:42 #info Correct link: 14:38:47 #link https://git.fd.io/csit/tree/docs/lab/testbeds_sm_zn2_hw_bios_cfg.md#n628 14:38:59 #link https://git.fd.io/csit/tree/docs/lab/testbeds_sm_zn2_hw_bios_cfg.md#n484 14:39:52 #info Peter next steps: finish applying GRUB cmd line 14:40:43 #info Peter, Maciek next is to run baseline tests l2, ip4, ip6 14:42:01 #info Peter, job spec is already there, c&p from 2n-clx 14:43:13 #info Keesang, prefers to execute the full set of iterative tests ASAP 14:47:39 #info Keesang, to make a recommendation about reading core frequency during tests, as it is not set to baseline 14:53:23 #info Maciek, proposal for AMD onboarding: run a set of baseline and scale tests for NIC2NIC tests, MRR, NDRPDR - l2patch, l2xc, l2bd, ip4, ip6; 2 runs; review the results; if okay then schedule few iterative runs, if consistent, schedule csit2-2009 run and publish as initial/to-be-reviewed set of results. Must be accompanied by clearly document clock frequencies during the tests. 14:53:43 #topic Virtual Infrastructure, CI/CD 14:55:51 #info Dave, ci-mgmt patches: cleaning docker scripts, new CSIT docker image patch, … 14:56:02 #topic Environment, Framework 14:56:25 #info MAciek, Peter, plan for Hugepages change 2MB to 1GB 14:56:59 #topic Performance Tests 15:01:48 #info Peter, new VPP GSO tests, currently coding iPerf3 … already getting early performance results with and without GSO, continue to tune the tests for affinity/resource allocation, and iPerf3 input params. 15:05:04 #info Vratko, nat44ed astf update, hit another 2 big issues with PPS / throughput: i) negative loss due to TRex server instances interactions, ii) issue with splitting bi-dir to uni-dir value interpretation, optimistic this is it 15:05:32 #info Vratko, CPS patch is (almost) ready for review, early comments welcome 15:06:07 #endmeeting