#opnfv-cperf: cperf

Meeting summary

1. 1. jamoluhrsen reports that he is working with Tim on getting us deployed on POD 2, making good progress (dfarrell07, 18:20:05)
   2. Al reports that intel pods will be shut down friday, Jamo will still try to see it work (dfarrell07, 18:20:22)
   3. Intel will have some power outage this weekend, will be down until Tuesday morning (dfarrell07, 18:20:53)
   4. Nikos has been working on Latte for code quality, better stats (dfarrell07, 18:21:28)
   5. Nikos merged PR from Al, good PR, thanks Al (dfarrell07, 18:21:38)
   6. Next steps are to add matching logic for other traffic scenarios, like CBench, or other request/responses like stats from controller (dfarrell07, 18:22:13)
   7. dfarrell07 jamoluhrsen and Nikos have been talking about NSTAT Dockerfile (dfarrell07, 18:22:28)
   8. Current NSTAT logic is Ubuntu-based, should we port to CentOS and install in current Dockerfile or have more than one container (dfarrell07, 18:22:51)
   9. dfarrell07 thinks we should use container-per-tool, to best leverage containers (dfarrell07, 18:23:06)
   10. Discussion about options for NSTAT Dockerfile (dfarrell07, 18:28:04)
   11. Discussion about what boxes in pods have internet (dfarrell07, 18:28:15)
   12. Jumphost will have internet, containers can live there (dfarrell07, 18:28:43)
   13. will go with route of using nstat upstream dockerhub image, pulling into our test, not hosting copy Dockerfile or building into OPNFV account (dfarrell07, 18:30:16)
   14. Al and Nikos talk about stats in Latte, getting more useful ones (dfarrell07, 18:30:56)
   15. Luis took a quick look at Latte this weekend, but not so many cycles, will look at againsoon (dfarrell07, 18:32:31)
   16. Luis reports that MT-CBench is already installed in ODL CSIT, will start using that going forward, vs CBench (dfarrell07, 18:32:58)
   17. Luis will try to use minimal set of building blocks, tests from NSTAT/Latte/MT-CBench/Multinet (huge thanks to Nikos and Intercom Telco for these open source tools!) (dfarrell07, 18:33:44)
   18. Luis expects to be able to get Latte running in ODL CSIT easily (dfarrell07, 18:34:00)
   19. Latte currently produces a histogram, which we can't really plot, but Al is patching Latte to report stat numbers that we can plot in ODL CSIT (dfarrell07, 18:34:35)
   20. Matt has been working with mgkwill on NetVirt scale testing, looking at Kola, stuff with containers (dfarrell07, 18:35:33)
   21. Matt is optimistic about getting this scale stuff done for ODL Whitepaper (dfarrell07, 18:36:20)
   22. Comparing ODL to native Neutron networking (dfarrell07, 18:36:43)
   23. Luis reports that two of the current blocking bugs in OpenFlow Plugin would impact scale/perf, and clustering (dfarrell07, 18:37:12)
   24. ODL upstream has 15 blocking bugs currently for Boron, many in projects relevant to us (OFP), they are being fixed but they do impact our testing (dfarrell07, 18:37:54)
   25. Luis for sure can't do any clustering stuff currently (dfarrell07, 18:38:15)
   26. Luis can't scale switches either, another blocker (dfarrell07, 18:38:35)
   27. Luis can do the flow pushing test he showed last week, as well as MT-CBench and Latte w/o hitting blockers (dfarrell07, 18:38:57)
   28. Discussion about NetConf, BGP, that they are broken in CSIT Luis reports (dfarrell07, 18:39:42)
   29. dfarrell07 is chair for OPNFV TestPerf meeting next week, fun times (dfarrell07, 18:43:20)
   30. dfarrell07 is working on NetVirt Scale in CSIT test, running and debugging Robot logic, racking up that sandbox job count ;) (dfarrell07, 18:43:54)

Action items

1. (none)

People present (lines said)

1. dfarrell07 (32)
2. collabot (3)