15:05:38 <mackonstan> #startmeeting FD.io CSIT Project Meeting 15:05:38 <collab-meetbot> Meeting started Wed Nov 15 15:05:38 2023 UTC and is due to finish in 60 minutes. The chair is mackonstan. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:05:38 <collab-meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:05:38 <collab-meetbot> The meeting name has been set to 'fd_io_csit_project_meeting' 15:05:46 <mackonstan> #info Maciek Konstantynowicz 15:05:54 <vrpolak> #info Vratko Polak. 15:05:55 <mackonstan> #topic Agenda bashing 15:06:20 <tifrank> #info Tibor Frank 15:07:44 <mackonstan> #topic Inputs from projects - VPP 15:09:21 <mackonstan> #info Vratko - yday vpp-dev call, Lijian/Arm team are working on crypto acceleration, expecting 20-30% improvement for openssl library(?), will be pushing a patch to gerrit shortly 15:10:09 <mackonstan> #topic Inputs from projects - TSC 15:12:00 <mackonstan> #info Maciek - DaveW submitted lab extension budget request for cy2024 - see list of equipment noted in one of the previous meetings 15:12:51 <mackonstan> #topic Intel HW accelerator status 15:17:21 <mackonstan> #info SPR QAT2.0 - Kai Peter updating - Peter it is enabled on 2n-spr and can test initialization and uni-dir tests. 3na-spr and 3nb-spr require work due to the use of containers to run VPP instances per NUMA. 15:19:02 <dwallacelf> #info Dave Wallace 15:20:48 <mackonstan> #info SNR QAT2.0 - Peter - CPU replacement units arrived, Ali/Vexxhost replaced them, and working on getting them to boot successfully. 15:22:43 <mackonstan> #info aside - ICX-D installation - Peter Maciek Dave - Ali/Vexxhost had to use cables from Vexxhost stock, we should replenish "used 2X 4 ft DAC SFP28cables brand 10GTEK", $15 on amazon. 15:24:33 <mackonstan> #info agreement - Dave Maciek Peter - ask Ali/Vexxhost to invoice LFN for those cables 15:26:20 <mackonstan> #info agreement2 - Peter Dave Maciek - agreed to involve Vexxhost team in reviewing the new equipment shipping list and highlight any gaps (e.g. rail kits, cables) before the shipment is executed 15:27:37 <mackonstan> #info SPR DSA - Peter - BIOS v1.4 - updated on all devices 2n-, 3n-spr. 1n-spr to be updated. DSA enabled in BIOS. 15:29:18 <mackonstan> #topic 1n-spr work 15:30:11 <mackonstan> #info Peter - interfaces vanishing issue fixed, currently only IP timeouts happening sporadically 15:31:37 <mackonstan> #info back to hw accelerator - E810 GTPU offload - Kai - intel team are looking at the results 15:31:56 <mackonstan> #topic CSIT-2310 15:33:37 <mackonstan> #info testing - Tibor - all testing done incl. coverage, last testbed was 3nb-spr, finished earlier today. okay for LFN IT Vanessa to go ahead with maintenance of: Production Jenkins, Sandbox Jenkins, Gerrit, Jira. 15:34:38 <mackonstan> #info Tibor - 23.10 rls was very smooth, few hiccups fixed by Peter. no unexpected bugs or issues from Tibor's perspective. 15:35:25 <mackonstan> #info Tibor - latest coverage results will be published in csit.fd.io by tomorrow, as it takes up to 24hrs to process all results. 15:36:50 <mackonstan> #info Tibor - during rls testing implemented some improvements based on Vratko's feedback: i) more info for off-canvas, ii) more links, few more (total of 5 small tweaks). 15:37:36 <mackonstan> #info Tibor - added GSO tests to iterative tests - see rls report data - box graphs, comparison. 15:39:10 <mackonstan> #info Vratko - in documentation rls notes - new issues got added - total of 8 issues 15:39:22 <mackonstan> #link https://csit.fd.io/cdocs/release_notes/current/vpp_performance/#new 15:41:36 <mackonstan> #info CSIT-1935 rls 2310: Zero traffic reported in udpquic tests 15:41:44 <mackonstan> #link https://jira.fd.io/browse/CSIT-1935 15:42:43 <mackonstan> #info CSIT-1936 TRex occasionally sees link down in L2 perf tests 15:42:53 <mackonstan> #link https://jira.fd.io/browse/CSIT-1936 15:45:27 <mackonstan> #info CSIT-1937 Small but frequent loss in ASTF UDP on cx7 mlx5 15:45:34 <mackonstan> #link https://jira.fd.io/browse/CSIT-1937 15:45:40 <mackonstan> #info CSIT-1938 3n-alt: high scale ipsec policy tests may crash VPP 15:45:45 <mackonstan> #link https://jira.fd.io/browse/CSIT-1938 15:45:52 <mackonstan> #info CSIT-1939 3na-spr: VPP fails to start in first two test cases 15:45:58 <mackonstan> #link https://jira.fd.io/browse/CSIT-1939 15:46:05 <mackonstan> #info CSIT-1940 hardware acceleration does not work yet 15:46:12 <mackonstan> #link https://jira.fd.io/browse/CSIT-1940 15:46:17 <mackonstan> #info CSIT-1941 TRex may wrongly detect link bandwidth 15:46:23 <mackonstan> #link https://jira.fd.io/browse/CSIT-1941 15:46:30 <mackonstan> #info CSIT-1942 3nb-spr hoststack: interface not up after first test 15:46:37 <mackonstan> 15:48:00 <mackonstan> #info RCAs - Vratko - 2 RCAs 15:48:02 <mackonstan> #link https://csit.fd.io/cdocs/release_notes/current/vpp_performance/#root-cause-analysis-for-regressions 15:49:10 <mackonstan> #topic Virtual Infrastructure, CI/CD 15:49:17 <mackonstan> #info maintenance tomorrow - see note above 15:49:38 <mackonstan> #topic Environment, Framework 15:50:28 <vrpolak> #link 39549: interface dpdk avf: introducing setting RSS hash key feature | https://gerrit.fd.io/r/c/vpp/+/39549 15:50:40 <mackonstan> #info Vratko - someone added an VPP API call to set the RSS key 15:50:42 <mackonstan> #link 39549: interface dpdk avf: introducing setting RSS hash key feature | https://gerrit.fd.io/r/c/vpp/+/39549 15:51:20 <mackonstan> #info Vratko - running gerrit triggered perf tests that are still to complete 15:52:57 <mackonstan> #info re CLI without API - Vratko - some virtio vhost tests failing due to typo - fix is easy, see jira ticket 15:53:37 <mackonstan> #link https://jira.fd.io/browse/VPP-2088 virtio: bad CLI argument parsing introduced with tx-queue-size 15:54:11 <mackonstan> #topic Trending and Test Failures 15:54:37 <mackonstan> #info Tibor - we are back to normal trending cadence 15:55:43 <mackonstan> #info Vratko - still working through the coverage results, will move to trending next 15:56:23 <mackonstan> #info Vratko - for the first time no difference observed between release and trending testing. 15:56:43 <mackonstan> #endmeeting