18:15:24 #startmeeting cperf 18:15:24 Meeting started Thu Aug 3 18:15:24 2017 UTC. The chair is dfarrell07. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:15:24 Useful Commands: #action #agreed #help #info #idea #link #topic. 18:15:24 The meeting name has been set to 'cperf' 18:15:53 #chair mattw4 jamoluhrsen LuisGomez mgkwill sai 18:15:53 Current chairs: LuisGomez dfarrell07 jamoluhrsen mattw4 mgkwill sai 18:16:10 joining 18:16:29 joining 18:19:23 #info mattw4 is up for vote as PTL of CPerf, has Jamo/Luis/Marcus vote, missing Tim, dfarrell07 will ping 18:19:49 #info mattw4 is up to 210 nodes on scale tests, now that he's actually placing the nodes deliberately 18:20:08 #info mattw4 has some container patch merged, next up is ansible deployment stuff 18:25:06 #info mattw4 is doing some cleanup of ansible, dfarrell07 and mattw4 talk about where to put it once ready, seems like merge into ansible-odl 18:25:16 #link https://git.opendaylight.org/gerrit/gitweb?p=integration/packaging/ansible-opendaylight.git;a=tree ansible-opendaylight role 18:26:00 #info jamoluhrsen is using CPerf's docker image downstream again, so will have feedback maybe 18:27:55 #info sai talks about next round of carbon testing RH will do, planning what focus will be 18:28:29 #info will be using 50+ compute, control plane tests, longevity tests, will try to check longevity issues cable labs has been reporting 18:28:54 #info sai expects first iteration of testing will start in early september, if you have issues/ideas please report them ASAP 18:29:15 #info mattw4 asks sai about how they are spreading out openstack services 18:30:56 #info sai talks about first model, HA with 3 controllers all openstack services on the box, 2nd with ODL on bare metal nodes and ODL isolated, will try to figure out at what scale ODL needs to be on dedicated hardware nodes 18:31:18 #info mattw4 is basically planning the same thing, currently running ODL on shared box w/ openstack, looking to try ODL on isolated box 18:31:42 #info big focus in these next tests will be to check for RAM issue, which we saw last time, make sure we get RAM dumps to analyze 18:32:15 #info mattw4 talks about standard model being compute nodes in containers, one instance per compute node 18:33:16 #info mattw4 clarifies that it's a real instance VM via qemu 18:33:36 #info jamoluhrsen bets anyone a case of beer that there's a RAM issue, they are seeing in CI with fairly simple tests 18:33:48 #info jamoluhrsen also expects there will be big perf hit when running in a cluster 18:34:14 #info sai says they will only be deploying with a clustered ODL, will only use single node once stuff breaks 18:34:44 #info jamoluhrsen says it would be helpful to see comparison of cluster and non-cluster for many tests, 18:35:16 #endmeeting