16:04:17 #startmeeting fdio-vpp 16:04:17 Meeting started Tue Nov 10 16:04:17 2020 UTC and is due to finish in 60 minutes. The chair is DaveBarach. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:04:17 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:04:17 The meeting name has been set to 'fdio_vpp' 16:04:34 https://docs.releng.linuxfoundation.org/en/latest/meetbot.html 16:04:47 #topic CSIT (Maciek) 16:05:10 #info CSIT-2009 report published on 15-Oct, see announce email from me 16:06:25 #info All but AMD and nat44ed test data are there 16:06:34 #link https://docs.fd.io/csit/rls2009/report/ 16:07:06 #info Ongoing work on: nat44ed cps tunings, gso tests, geneve tests 16:08:02 #info CSIT-2009 RCA update by Vratko, no big enough regressions (< -5%), spent time on some smaller regressions (l2patch, see ticket) 16:08:55 #info Trending update by Vratko, one regression that's not real (introduced by nat44ed tests), side-effect of not trusting TRex when it comes to test time measurement ( <1% change) 16:09:35 #info Vratko, some progressions noticed but not investigated 16:10:52 #topic Host Stack (Florin) 16:11:16 #info various cleanups, to avoid bit-rot 16:11:57 #info to improve compile time, add {tcp,udp}_packet.h vs. protocol implementation headers 16:12:45 #info per-worker cut-through sessions. Gives much improved cut-through session setup performance 16:13:28 #info host stack no longer supports legacy binary API (shared memory) connections. Security measure. 16:14:04 #info Question: why does dpdk map 128gb of memory? 16:14:23 #info Damjan talked to Anatoly (dpdk guy) yesterday about that 16:14:57 #info dpdk maps a huge chunk, then finds an appropriate boundary 16:16:04 #info host stack can trip over the dpdk etc. 16:16:55 #info vcl improved handling of out-of-order events. Small cleanups / fixes in addition. 16:17:03 #topic Docs (Ray) 16:17:36 #info Ray points to Ole... ;-)... 16:18:34 #topic Release Management (Andrew, DaveW) 16:19:31 #info 21.01 roughly 1 month to [sic] API freeze pre-throttle pull cool-down 16:19:55 #info Actual release in January, most of a month between RC1 and RC2 16:20:54 #info 20.09 one outstanding question: maintenance release 20.09.1 16:21:12 #info Any requests for the dot release? Timing? 16:22:08 #info DaveW suggests that waiting until February would be quite a while since the formal release 16:22:45 #info 19.08 Job removal in January? 16:23:10 #info Not planning to do another 19.08 point release. Revisit in December. 16:24:36 #info Cadence after 21.01: September is tricky due to summer vacation. 16:25:17 #info Possibly 21.06 and 21.10. 16:25:41 #info 22.01, etc. 16:26:01 #topic Infra Status (Dave W) 16:26:17 #info 4 [sic] emergency maintenance ops in the last month 16:26:40 #info All of the LF instances have increased in capacity 16:27:01 #info Some positive results from the exercise 16:27:23 #info Finally root-caused the multiple cherry-pick issues. 16:27:48 #info Need to reduce the Nomad cluster cleanup interval. 16:28:18 #info Docker image creation scripts merged into ci-mgmt 16:30:26 #info Centos-7 job will go away in the next release; removing from master builds later today 16:31:04 #info Debian 9 is old. Would like to switch to Debian 10 in the 21.0 time frame 16:32:47 #info DaveB reports > 20 static analysis complaints 16:33:39 #info Ray reports Intel internal discussion about moving away from Centos-7 16:33:59 #info Centos-8 patch validation online 16:38:04 #topic Kubecon (Jerome) 16:38:28 #info Presenting VPP / Calico integration at Kubecon North America 16:38:42 #info joint preso between Aloys and the Calico guys 16:39:13 #info three-way comparison between Calico/VPP, Calico/Linux and Calico/eBPF 16:39:27 #info expect recording to be available 16:40:12 #info $50 to join live 16:43:42 #endmeeting