====================================== #fdio-csit: FD.io CSIT Project Meeting ====================================== Meeting started by mackonstan at 14:01:22 UTC. The full logs are available at http://ircbot.wl.linuxfoundation.org/meetings/fdio-csit/2021/fd_io_csit_project_meeting/fdio-csit-fd_io_csit_project_meeting.2021-04-14-14.01.log.html . Meeting summary --------------- * Tibor Frank (tifrank, 14:02:08) * Agenda bashing (mackonstan, 14:02:39) * Juraj Linkeš (jlinkes, 14:03:06) * VPP updates (mackonstan, 14:03:12) * Andrew, Ole is preparing a patch with crc recalc for APIs, to discuss later in the call (mackonstan, 14:04:08) * Andrew, long discussion about the source tree placement for vpp make test, no consensus yet, discussion to continue on the next vpp call in 1 month (mackonstan, 14:05:11) * VSAP (mackonstan, 14:05:24) * Dave, VSAP project patches arriving based on feedback (mackonstan, 14:07:44) * Dave, developing tests for UDP on the back of existing TCP tests, for VPP hoststack, with iperf3 (mackonstan, 14:07:44) * Releases 2106 (mackonstan, 14:08:32) * Release 2101 (mackonstan, 14:10:00) * Physical Infrastructure (mackonstan, 14:10:19) * Dave, encountered Gerrit fetch failure yesterday, from .32.8 subnet, no signature pointing towards ingress issue (mackonstan, 14:12:51) * Peter, MTL1 to YUL1 move, stage-2 server list provided to Vexxhost/Farid, Nomad servers are put out-of-prod, Farid found an inventory inconsistency fixing it now. (mackonstan, 14:16:35) * Maciek, expecting from Vexxhost/ logical network diagram covering all http://fd.io/ subnets before and during/after move MTL1 > YUL1, with annotations where each of the segments live and if any traffic is expected to cross WAN (it shouldn’t) (mackonstan, 14:17:07) * Agreed on a call that we are ready to go ahead with stage-2 move this week. (mackonstan, 14:17:26) * Coordination of decisions and work thru a dedicated Vexxhost - CSIT slack channel owned by Vexxhost. (mackonstan, 14:18:22) * FD.io lab extension in 2021 (mackonstan, 14:18:45) * TG server recommendations provided to FD.io TSC (mackonstan, 14:19:51) * Intel's Xeon Icelake announced last week, expecting Intel to donate some of new Xeon's to the lab for TGs and SUTs (mackonstan, 14:22:24) * Juraj, will keep checking re new Arm perf testbeds e.g. based on Ampere/N1 (mackonstan, 14:22:46) * Maciek, to check with AMD if any new SUT HW expected to be donated in 2021 into FD.io labs (mackonstan, 14:23:05) * Virtual Infrastructure, CI/CD (mackonstan, 14:23:18) * Dave, finished cleanup of logging storage move from nexus to S3, for CSIT no longer archive directory (mackonstan, 14:28:01) * Dave, unclear if the actual path to logs.fd.io/production/.. is going to change or not … (mackonstan, 14:28:26) * Dave, also observe ccache mods per below patch (mackonstan, 14:32:00) * LINK: https://gerrit.fd.io/r/c/ci-management/+/31914 (mackonstan, 14:32:06) * Environment, Framework (mackonstan, 14:33:00) * Maciek, re addressing papi performance for CSIT scale tests, and email thread by Vratko been going on for a while, few months, no conclusion yet, will add this to the agenda and also drive focus consultation with vpp-dev to conclude here (mackonstan, 14:36:16) * Andrew, api msg crc calc change patch, https://gerrit.fd.io/r/c/vpp/+/30345 : vppapigen: remove flag day code, merge into VPP master planned for 21-Apr (mackonstan, 14:39:47) * LINK: https://gerrit.fd.io/r/c/vpp/+/30345 (mackonstan, 14:39:55) * see https://jenkins.fd.io/job/vpp-csit-verify-api-crc-master-ubuntu2004-x86_64/482/ : FAILURE (mackonstan, 14:40:21) * Maciek, Andrew, Vratko to coordinate with Paul, Ole on the handling of the flad day to ensure smooth VPP and CSIT CI/CD transition (mackonstan, 14:41:33) * Performance Tests (mackonstan, 14:42:00) * Tomas, trending failing tests incl. nat44ed multi-worker, ipsec qat tests, other issues got resolved (mackonstan, 14:43:05) * FD.io TSC Evolution update (mackonstan, 14:43:32) * Maciek walks thru the TSC evolution (mackonstan, 14:48:24) * LINK: https://docs.google.com/presentation/d/1kJ0uI3S6Ka5JdtFQh9XYwGq9fWRaQH4RHkzKf55TTgc/edit#slide=id.gca88b1a448_0_186 (mackonstan, 14:48:42) * Dave, checking the status of "Faulty VF initialization during DPDK startup when multiple DPDK instances use different VFs with the same PF", Maciek case open and with DPDK/Intel team per Intel case open (mackonstan, 14:54:36) * Andrew, some make test changes to watch, 31950 : tests: cpus awareness, by Klement (mackonstan, 14:58:58) * LINK: https://gerrit.fd.io/r/c/vpp/+/31950 (mackonstan, 14:59:05) * Andrew, this type of core allocation may be causing make test failures (mackonstan, 14:59:26) Meeting ended at 14:59:51 UTC. People present (lines said) --------------------------- * mackonstan (42) * collab-meetbot (3) * tifrank (1) * jlinkes (1) Generated by `MeetBot`_ 0.1.4