13:58:31 <mackonstan> #startmeeting FD.io CSIT project meeting 13:58:31 <collab-meetbot> Meeting started Wed Jul 1 13:58:31 2020 UTC and is due to finish in 60 minutes. The chair is mackonstan. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:58:31 <collab-meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:58:31 <collab-meetbot> The meeting name has been set to 'fd_io_csit_project_meeting' 14:02:11 <mackonstan> #chair 14:02:11 <dwallacelf> #info Dave Wallace 14:02:11 <collab-meetbot> Current chairs: mackonstan 14:02:14 <jgelety> #info Jan Gelety 14:02:18 <vrpolak> #info Vratko Polak. 14:02:21 <mackonstan> #topic Agenda bashing 14:02:39 <mackonstan> #info Juraj: question of about adding new jobs 14:03:07 <mackonstan> #info Maciek: adding a new 2-node amd testbed 14:04:09 <mackonstan> #topic Inputs from projects - VPP 14:04:19 <mackonstan> #info Dave - no updates, all on track 14:04:31 <mackonstan> #topic Inputs from projects - VSAP 14:05:40 <mackonstan> #info Dave - meeting happened yesterday, major topics: packaging, csit patches refactor agreement to align with hoststack 14:06:00 <mackonstan> #info Dave - it wasn't a public meeting, so no notes 14:06:18 <mackonstan> #topic Inputs from projects - TSC 14:07:32 <mackonstan> #info Vratko - no actual progress on GPL relicensing, this week TSC meeting not happening due to volume PTOs, proposal to have monthly webinars (1st is end of August about IPsec performance) 14:07:58 <mackonstan> #topic Releases - CSIT-2005 14:09:22 <vrpolak> #link https://wiki.fd.io/view/Fd.io_Webinar_Topics 14:11:07 <mackonstan> #info Maciek - summary of where we are with CSIT-2005 benchmarks: 14:14:16 <mackonstan> #info Summary: 14:14:17 <mackonstan> 1. Moving from 2005.0-3 to 2005.0-4 made the comparison clearer but further work is required (-3 and -4 are CSIT environment versions). 14:14:17 <mackonstan> a. -3 to -4 change is reverted TRex 7to15-workers change, that has introduced different packet burst profile, and potentially also different packet flow sequencing on the wire. 14:14:17 <mackonstan> b. -3 was responsible for a large number of regressions with vhost and memif packet paths. 14:14:17 <mackonstan> 2. Inconclusive bisecting of identified 2005.0-4 vs 2001.0-2 progressions. 14:14:17 <mackonstan> a. Findings to-date are inconclusive from the VPP code point of view, indicating impact of further CSIT changes, in addition to TRex 7to15-workers change that has been already reverted. 14:14:18 <mackonstan> 3. In order to address point 2., we agreed to do following 14:14:18 <mackonstan> a. Separate out any CSIT changes by re-applying comparison methodology developed in CSIT-2001 cycle, and test vpp v20.01.0 in CSIT-2005-4 environment. 14:14:18 <mackonstan> b. Focus on bisecting identified 2005.0-4 vs 2001.0-2 regressions only. 14:14:19 <mackonstan> c. This approach is consistent with what has been done in CSIT-2001 to address skx and clx ucode induced CSIT environment change. 14:16:08 <mackonstan> #info Tibor to run 2001.0-4 set of tests (vpp v20.01.0 tested in CSIT-2005-4): 14:16:08 <mackonstan> ◦ 4 runs on 2n-clx. 14:16:08 <mackonstan> ◦ 1 run on 2n-skx, 3n-skx, 3n-hsw, 3n-tsh. 14:22:43 <vrpolak> #link https://gerrit.fd.io/r/c/csit/+/27396/7#message-460ca8013124dfd04431a3f61732d208fa479128 <-- Source for the "clx testbed faster with hsw variant" claim. 14:23:54 <mackonstan> #topic Releases CSIT-2001, CSIT-1908.2 14:23:58 <dwallacelf> #info Once the definition of "CSIT environment" is complete, the release schedule should include appropriate freeze dates on env changes so previous release testing can begin early in the release cycle. 14:24:04 <mackonstan> #info no updates 14:24:39 <mackonstan> #topic Physical Infrastructure 14:29:02 <mackonstan> #info Arm build servers - Juraj, Peter working directly with Vexxhost - latest status in Vexxhost ticket (2 servers still dead). Separate tickets for each. Need to coordinate with on-site presence in DC. 14:30:42 <mackonstan> #action Juraj, Peter, Maciek - propose a consolidated way-forward proposal to address all Arm servers down tickets (3 servers in total, 2 build, 1 thunderX). 14:31:25 <mackonstan> #topic Virtual Infrastructure, CI/CD 14:32:42 <dwallacelf> #info Dave running ping monitor scripts and updating Vexxhost ticket daily. Will send out instructions for running scripts while I'm on PTO. 14:33:39 <mackonstan> #info Juraj - question of about adding new jobs, want to add Centos job, question to DaveW 14:34:57 <mackonstan> #info Juraj - all details in email from xxx, Juraj to confirm 14:35:17 <mackonstan> #topic Environment, Framework 14:35:27 <mackonstan> #topic Performance Tests 14:37:37 <mackonstan> #info Vratko - tests using DPDK plugin are failing, per VPP build all tests using DPDK plugin are either failing or passing 14:38:05 <mackonstan> #info Peter - saw some discussion on this topic, will investigate it 14:38:23 <jlinkes> Jieqiang wang 14:38:51 <mackonstan> #info re above `xxx` note, it is `Jieqiang wang` 14:40:07 <mackonstan> #info Maciek, Jan, Maros - work continues with NAT44-EI and -ED, using TRex ASTF and TRex STL with symmetric burst profiles 14:41:36 <mackonstan> #endmeeting