14:11:11 <mackonstan> #startmeeting FD.io CSIT Project Meeting
14:11:11 <collab-meetbot> Meeting started Wed Sep 15 14:11:11 2021 UTC and is due to finish in 60 minutes.  The chair is mackonstan. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:11:11 <collab-meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:11:11 <collab-meetbot> The meeting name has been set to 'fd_io_csit_project_meeting'
14:11:25 <mackonstan> #info mackonstan
14:11:28 <fan_zhang> #info Fan Zhang
14:11:32 <mackonstan> #topic Agenda bashing
14:11:41 <mackonstan> #info Andrew Yourtchenko
14:11:42 <mackonstan> #chair
14:11:43 <collab-meetbot> Current chairs: mackonstan
14:11:47 <mackonstan> #topic Agenda bashing
14:12:05 <mackonstan> #info only few of us here, so short meeting, focusing on incremental updates
14:12:14 <mackonstan> #topic VPP update
14:12:33 <mackonstan> #info vpp rc1 and branch pull in 1 week
14:12:41 <mackonstan> #topic TSC update
14:13:54 <mackonstan> #info Maciek: CY 2022 FD.io budget planning, Trishan is collecting feedback, review on TSC meeting tomorrow
14:14:10 <mackonstan> #topic New tests, refactored tests
14:15:11 <mackonstan> #info Fan, performance tests require more careful design, unlikely to happen for this release
14:17:18 <mackonstan> #info Fan, vpp_device tests are not working on 1n-skx, gettiner error: the "vpp_sw_index" of E810CQ port is None.
14:18:15 <mackonstan> #info Maciek, may be related to issue with one of the NICs, per Peter's vexxhost ticket (replace E810 with only 1 port 200GE showing up):
14:18:17 <mackonstan> #link https://secure.vexxhost.com/billing/viewticket.php?tid=BGY-197011&c=3b1qMRWQ
14:23:34 <mackonstan> #info Fan, re performance tests for async crypto, crypto worker scheduler is working fine, suspecting I/O worker (the one being faster than crypto workers) being too fast, pushing too many packets to crypto workers, and as the buffer pool is full, dropping them to the floor. But no hard proof yet.
14:26:07 <mackonstan> #info Fan, need a counter similar to rx_no_buf on interfaces.
14:29:05 <mackonstan> #info Fan, problem apparent for 1518B and 9000B. Trying to address the problem, not sure how to do it on VPP side, but for now focusing on getting it working in CSIT by extending the size of the buffer pool size and increasing number of crypto-workers.
14:30:13 <mackonstan> #info Maciek, TRex b2b tests are completed, used for TRex performance monitoring, calibration and will be added shortly to trending pages.
14:30:55 <mackonstan> #info Maciek, scale tests are now fixed and working, replaced broken VAT with CLI executor for now, will be moving to VAT2 next
14:31:16 <mackonstan> #topic CSIT-2110 rls
14:31:28 <mackonstan> #info Maciek wiki link updated
14:32:28 <mackonstan> #link https://wiki.fd.io/view/CSIT/csit2110_plan
14:33:34 <mackonstan> #info Fan, will review with the team and update
14:35:50 <mackonstan> #topic Virtual Infrastructure, CI/CD
14:36:06 <mackonstan> #info infra stable after move to the new YUL1 DC location
14:37:00 <mackonstan> #info Andrew, after rc1 branch pull, will be stress testing the infra with dummy changes ahead of the busy pre-release period
14:37:34 <mackonstan> #info Maciek (on behalf of Vratko): oper branches optimisation in CSIT, see discussion in Vratko's patch
14:38:01 <mackonstan> #link 33690: Docs: Discussion on oper branches | https://gerrit.fd.io/r/c/csit/+/33690
14:38:12 <mackonstan> #topic Environment
14:38:55 <mackonstan> #info Maciek (on behalf of Peter) started to upgrade FVL NIC FWs on TRex machines to bring it to recommended version for DPDK 21.02, and test verifying any anomalies
14:39:16 <mackonstan> #topic Trending
14:39:33 <mackonstan> #info Maciek two tests failing in daily trending:
14:41:37 <mackonstan> #info 1) nat44ed 16M sessions test, sessions not getting established in allotted time for some tests
14:42:46 <mackonstan> #info 2) IPSec async scheduler test 9000B, 1518B - no traffic forwarded
14:42:49 <mackonstan> #endmeeting