#opnfv-cperf: cperf

Meeting summary

1. 1. dfarrell07 has sent some docs patches for the release, is generally keeping track of what we need to do for release (dfarrell07, 18:20:51)
   2. jamoluhrsen has some cperf job fixes pushed, waiting on code review from Tim who is busy with release (dfarrell07, 18:21:10)
   3. mat is working with clusters in intel labs (dfarrell07, 18:21:53)
   4. starting to see some results around limits, what we eventually hit when scaling (dfarrell07, 18:22:16)
   5. mattw4 is still fighting with devstack (dfarrell07, 18:22:28)
   6. lots of work going into finding cofig that is equilv, odl vs neutron deploys, lots of digging under the hood (dfarrell07, 18:23:49)
   7. trozet ninjas in and merges jamoluhrsen patch to maybe-fix cbench job (dfarrell07, 18:24:39)
   8. jamoluhrsen and mattw4 talk about devstack config (dfarrell07, 18:25:18)
   9. mgkwill has kolla patch that is looking good, seems close to merge (dfarrell07, 18:28:42)
   10. dfarrell07 warns mattw4 that there is new dockerfile linting in int/pack that will impact his under-dev patches (dfarrell07, 18:29:01)
   11. mgkwill says kolla patches are passing gates, working in lab, will be helpful for OOO (dfarrell07, 18:29:44)
   12. mgkwill is also working on patch to ansible to allow some overrides that are needed (dfarrell07, 18:30:11)
   13. mgkwill has been running upstream tests in intel test env, has some results (dfarrell07, 18:31:25)
   14. mgkwill some tests seem to be hardcoded to run on snapshots, bgp and netconf tests seem to break on release distros, others are running (dfarrell07, 18:31:51)
   15. jamoluhrsen confirms that^^ issue is true, vratko talked about it on int/* call in different context recently (dfarrell07, 18:32:23)
   16. LuisGomez says that someone from his team is going to start working on netconf (dfarrell07, 18:33:06)
   17. mgkwill shares results from tests they have run in lab so far (dfarrell07, 18:34:23)
   18. mgkwill reports he is seeing 10-20% regression in Boron SR2 over Be SR1 for OF tests, LuisGomez reminds that we totally changed OF plugin between those releases (dfarrell07, 18:36:10)
   19. LuisGomez reports that flow pushing is slower in new OF plugin, but stats collection is tons faster (dfarrell07, 18:36:32)
   20. mgkwill clarifies that both sides of this test are using the new plugin (dfarrell07, 18:37:23)
   21. LuisGomez reports that OF team has been cherry-picking lots from master to Boron, and they have't been looking at perf much yet, so maybe explains (dfarrell07, 18:37:53)
   22. mgkwill reports seeing 5x better NB API REST results with flow blaster script, vs ROBOT implementation of same test, needs looking into (dfarrell07, 18:39:15)
   23. mgkwill talks about RPC perf reduction, LuisGomez explains that RPC is now sync, gets confirm of flow install, should not be used for perf (dfarrell07, 18:40:33)
   24. mgkwill is seeing about a 50% drop in OF bulk flow programming tests in 3-node cluster vs single node, LuisGomez says this is about what he expects based on upstream results (dfarrell07, 18:42:22)
   25. acm is at IETF today (dfarrell07, 18:49:48)
   26. acm gave them feedback about doing pkt in tests in 0-loss env like we with with latte and Nikos, that is now in RFC draft (dfarrell07, 18:50:39)
   27. acm asks mgkwill about running some tests with latte to see 0 loss rate (dfarrell07, 18:51:10)
   28. LuisGomez's past issue about results reporting has been fixed, now has output that can be used in tests (dfarrell07, 18:51:48)
   29. cperf is canceled next week due to ONS (dfarrell07, 18:53:05)

Action items

1. (none)

People present (lines said)

1. dfarrell07 (38)
2. collabot` (5)
3. trozet (4)
4. jamoluhrsen (3)