========================================= #fdio-meeting: FD.io CSIT Project Meeting ========================================= Meeting started by mackonstan at 14:59:50 UTC. The full logs are available at http://ircbot.wl.linuxfoundation.org/meetings/fdio-meeting/2023/fd_io_csit_project_meeting/fdio-meeting-fd_io_csit_project_meeting.2023-01-18-14.59.log.html . Meeting summary --------------- * Juraj Linkeš (jlinkes, 15:03:13) * mackonstan (mackonstan, 15:04:17) * Vratko Polak. (vrpolak, 15:04:32) * Agenda bashing (mackonstan, 15:05:35) * VPP project updates (mackonstan, 15:05:42) * Andrew rc1 in progress today (mackonstan, 15:05:58) * Andrew so far according to plan (mackonstan, 15:06:19) * FD.io TSC update (mackonstan, 15:06:32) * Vratko - some discussion in the last meeting about a new way to log into LFN project meetings using LF id - see meeting notes (mackonstan, 15:08:46) * E810 flow "offload" tests (mackonstan, 15:09:20) * Kai - discussed with Yulong, still seeing discrepancies between perf results in CSIT labs and Intel labs. (mackonstan, 15:11:02) * Kai - suspecting HW differences between the two environments. (mackonstan, 15:12:16) * Kai - will send email to csit-dev with the current status and proposed next steps, goal is to have a full clarity of where this is before we ship 23.03 (mackonstan, 15:13:08) * correction: 23.02 not 23.03 (mackonstan, 15:13:29) * Scale tests API optimization (mackonstan, 15:13:48) * Vratko - latest changes merged, code is faster than before, and execution time is stable. (mackonstan, 15:14:55) * Vratko - next step is to watch both trending results and rc1 results to see if any issues. (mackonstan, 15:15:26) * VPP ip reassembly perf tests (mackonstan, 15:15:59) * Vratko - not ready for rc1 (mackonstan, 15:16:08) * Vratko - but some required improvements, namely jumbo tests, may be ready - will send summary to csit-dev (mackonstan, 15:17:25) * wireguard 10k tunnels (mackonstan, 15:18:13) * Kai - patch inside vpp waiting for review - yulong - fan already reviewed the patch. once reviewed and merged, and then wireguard 10k tunnels tests are unblocked. (mackonstan, 15:19:09) * Andrew - as a bug fixes it should be addressed asap before rc2 milestone (mackonstan, 15:19:27) * Release CSIT-2302 (mackonstan, 15:21:03) * Tibor - will be pulling the branch tomorrow once vpp rc1 artifacts are published (mackonstan, 15:21:29) * Tibor, Maciek - inspect current job specs and verify that any updates are applied before doing rc1 tests incl. dpdk mlx, hoststack set, gtpu, wireguard. (mackonstan, 15:23:47) * Tibor, Maciek - send update to csit-dev with what's tested in rc1. (mackonstan, 15:25:49) * Kai - to take care of job specs - gtpu, wireguard. (mackonstan, 15:26:38) * Tibor, Maciek - to take care of job specs - dpdk mlx, all hoststack. (mackonstan, 15:27:54) * csit.fd.io telemetry - Tibor - working on reading telemetry data - issues with big memory consumption, addressing those, still on target to ship in 23.02 timeframe, "confidence level 80%" :^) (mackonstan, 15:29:32) * Tibor - not aware of any changes in CSIT environment, so no need to retest vpp v22.10 in CSIT-2302 for current vs previous comparion, unless anybody comes up with a valid reason why this would need to be retested now. (mackonstan, 15:31:33) * Juraj - asking if (deprecated from kernel) isolcpu got addressed, Peter confirmed it has not been addressed, need to do it in the next rls cycle. (mackonstan, 15:34:30) * Physical Infrastructure (mackonstan, 15:34:57) * mackonstan - mlx 8x cx6 and 24x cx7 nic cards arrived to FD.io labs, will be going into mainly SPR servers, some cx6 will be going into ICX servers. (mackonstan, 15:36:21) * question to Juraj - confirm which mlx cards should go into Ampere servers (mackonstan, 15:37:15) * Virtual Infrastructure, CI/CD (mackonstan, 15:37:46) * Trending and Test Failures (mackonstan, 15:38:34) * mackonstan - re Hoststack tests - live meeting tomorrow - vpp-dev Dave, Florin, Filip Tehlar; csit-dev: Peter, Tibor, Viliam, Maciek (mackonstan, 15:39:21) * Viliam - CSIT Test Failures Summary and Status (mackonstan, 15:40:23) * Viliam - dpdk tests fixed on all testbeds except 2n-zn2 because it was down (HDD dead), but should be back now (mackonstan, 15:41:19) * Viliam - correction dpdk tests on 2n-zn2 should be working now, will know after the next weekly trending testing cycle this weekend (mackonstan, 15:42:42) Meeting ended at 15:43:23 UTC. People present (lines said) --------------------------- * mackonstan (42) * collab-meetbot` (9) * jlinkes (1) * vrpolak (1) * tifrank (0) * viliamluc (0) Generated by `MeetBot`_ 0.1.4