====================================== #fdio-csit: FD.io CSIT project meeting ====================================== Meeting started by mackonstan at 15:06:25 UTC. The full logs are available at http://ircbot.wl.linuxfoundation.org/meetings/fdio-csit/2020/fdio-csit.2020-01-15-15.06.log.html . Meeting summary --------------- * Tibor Frank (tifrank, 15:06:29) * Dave Wallace (dwallacelf, 15:06:31) * Jan Gelety (jgelety, 15:06:35) * mackonstan (mackonstan, 15:06:48) * Agenda bashing (mackonstan, 15:07:03) * Vratko Polak. (vrpolak, 15:07:16) * Physical labs, testbeds (mackonstan, 15:07:17) * Ed - one machine down from the cluster (total 3 servers) running s5ci. (mackonstan, 15:09:43) * Inputs from projects (mackonstan, 15:09:59) * VPP - Dave - v20.01 stable branch pulling today per plan. (mackonstan, 15:10:36) * VPP - Dave - v19.08.2 maintenance after v20.01 ships, per plan discussed and agreed in the past on VPP and CSIT calls. (mackonstan, 15:11:33) * TSC - Vratko - 2 news: i) VSEP project officially approved; ii) debdpdk project is moving out from fd.io git to somewhere. (mackonstan, 15:12:45) * VSEP -> VSAP (vrpolak, 15:13:28) * TSC - Maciek - we should demand to have a well defined and concrete timetable for FD.io dev community f2f meetups in 2020. (mackonstan, 15:15:26) * CNCF CNF Testbed - no updates (mackonstan, 15:15:53) * Releases - 1908 - no status change - VPP v19.08.2 dependent on CSIT productising weekly stable/1908 performance jobs (mackonstan, 15:16:53) * Vratko - confident no API changes sneaked thru to stable/1908, due to be guarded by api-crc jobs and Andrew Yourtchenko. (mackonstan, 15:18:49) * Releases - CSIT-2001 (mackonstan, 15:19:17) * Today is RC1 milestone - Code complete. Pull first release branch. Only bug fixes accepted in release branch. Date aligned with VPP RC1. Start dry-runs to identify CSIT gaps on less frequently run tests. (mackonstan, 15:20:14) * Jan - pulling CSIT 2001 branch today (rls2001). (mackonstan, 15:22:55) * Few CSIT patches need to go in before the branch is pulled: (mackonstan, 15:23:17) * trex_profile patch by Balaji, Under review by Jan Gelety, to make it python3.x. Expecting later today. (mackonstan, 15:24:28) * Dave - 3 patches for tcp/hoststack tests, merge ETA end of today: https://gerrit.fd.io/r/c/csit/+/24252, https://gerrit.fd.io/r/c/csit/+/24254, https://gerrit.fd.io/r/c/csit/+/24337 (mackonstan, 15:25:27) * that's about it, 4 patches in total to go in, before pulling the rls2001 branch (mackonstan, 15:26:03) * briefly reviewed updated content of CSIT-2001 per wiki (mackonstan, 15:31:48) * LINK: https://wiki.fd.io/view/CSIT/csit2001_plan (mackonstan, 15:31:55) * Operations (mackonstan, 15:32:17) * Govind - re Taishan upgrade from -62 to -72 kernel - did this in Arm internal labs and it failed. Looking to do a fresh install. Will update the community. (mackonstan, 15:33:55) * VPP code performance (mackonstan, 15:34:58) * Maciek - number of tests are reported as failing in daily email summaries on csit-report mailer (mackonstan, 15:35:27) * Peter - majority of the failures are fixed (vxlan, docker, nf_density). Other issues with TRex + dnv environment - fixed. IPsec HW failing due to some strange behaviour/issue on VPP side. MLX tests are failing, waiting for the final VPP patches. Taishan vhost-qemu having issues (testpmd memory allocation failure), Juraj is following. (mackonstan, 15:38:40) * Maciek: Upgrade from from Ubuntu 18.04 LTS kernel 4.15.0-66 to Ubuntu 18.04 LTS kernel 4.15.0-72, introduced Xeon Processors’ ucode update increasing coverage to known (but not all announced) meltdown-spectre vulnerabilities (CSIT code is tracking those by executing spectre-meltdown-checker as part of the test suite). Skylake impact at 2-3%, Cascadelake impact at >30% ! (mackonstan, 15:39:28) * Peter - dnv not touched, need to check with Intel team. (mackonstan, 15:41:34) * Peter - hsw upgraded on 14th or 13th of Feb, so should be visible soon. (mackonstan, 15:42:06) * Govind - re qemu failures on Taishan - asking for logs from Peter (mackonstan, 15:44:45) * Peter - for clx - meltdown-spectre-check diff from -66 to -72 captured in: https://gerrit.fd.io/r/c/csit/+/24349/2/docs/report/introduction/test_environment_sut_meltspec_clx.rst#51 (mackonstan, 15:47:29) * LINK: https://gerrit.fd.io/r/c/csit/+/24349/2/docs/report/introduction/test_environment_sut_meltspec_clx.rst#51 (mackonstan, 15:47:35) * Developments (mackonstan, 15:47:53) * Vratko, Tibor - Deprecated tests: DMM, TLDK - removed from master (mackonstan, 15:48:41) * TRex (mackonstan, 15:49:23) * seeing progressions after TRex upgrade to v2.73 (includes patch to the firmware to remove the limit of 27Mpps for TRex). (mackonstan, 15:50:41) * CSIT-2001 report draft (mackonstan, 15:52:08) * LINK: https://docs.fd.io/csit/master/report/ (mackonstan, 15:52:10) * test data is old or fake (mackonstan, 15:52:32) * Tibor - additions to master already in: new latency hdrhistogram, new comparison tables, full TODO list will be send to csit-dev. (mackonstan, 15:58:10) Meeting ended at 16:00:30 UTC. People present (lines said) --------------------------- * mackonstan (43) * collabot` (4) * vrpolak (2) * dwallacelf (1) * jgelety (1) * tifrank (1) Generated by `MeetBot`_ 0.1.4