18:19:20 <dfarrell07> #startmeeting cperf
18:19:20 <collabot> Meeting started Thu Jun 22 18:19:20 2017 UTC.  The chair is dfarrell07. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:19:20 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic.
18:19:20 <collabot> The meeting name has been set to 'cperf'
18:20:40 <dfarrell07> #info mattw4 reports they have 50 cerros256 min instaces on a single compute node
18:20:51 <dfarrell07> #info raghu is looking at rally, getting that working
18:21:06 <dfarrell07> #info mattw4 and raghu are working on test scenarios
18:23:45 <dfarrell07> #info mattw4 had 30 compute nodes + instances running on a server with 64GB RAM
18:24:03 <dfarrell07> #info as the test gets bigger it takes longer for openstack to see the instances, like a min at most
18:24:29 <dfarrell07> #info at 30+ things start to break, will not dhcp, maybe bringing it back up will work but that's not so stable
18:25:05 <dfarrell07> #info with 6 systems in the cperf pod with 64GB ram each, we could get some crazy scale
18:25:25 <dfarrell07> #info because we're using linux bridges for container networking, it should be easy to put odl in a place that has lots of resources and still talk to it
18:27:25 <dfarrell07> #info acm is back from openf summit, welcome
18:27:38 <dfarrell07> #info acm had a very well attended talk on monday about vswitch perf results
18:28:38 <dfarrell07> #info acm talks about traffic gens they are looking at, some RHers have been doing stuff with moongen+??, now in vsperf
18:28:59 <dfarrell07> #info the ones they have played with are a bit hard to use, some strange results, but as it goes
18:29:30 <dfarrell07> #info discussion about scale issue with ODL vswitch scale issue, also on email
18:32:43 <mattw4> We've seen the same issue (messages getting dropped & incomplete rule population) in our early testing
18:33:54 <dfarrell07> #info mattw4 talks about seeing some similar issues
18:36:57 <dfarrell07> #info talking about the 30 nodes, the flow dump from ofp is err small, so also dropping flows
18:38:40 <dfarrell07> #info we should push on ofp to actually fix this issue, we are very able to since we're seeing it in multi tests
18:38:51 <dfarrell07> #info mattw4 gives us a quick demo of the scale setup he has
18:47:35 <dfarrell07> #info mattw4's demo is fucking amazing, everyone swoons over automation, very cool ansible stuff
18:47:44 <mattw4> <blushes>
18:47:45 <dfarrell07> #info this is all very ready to be upstream, will be amazing in ci
18:47:50 <dfarrell07> lol ;)
18:48:11 <dfarrell07> #info mattw4 should send talks to open source summit north america if CFP is still open
18:48:15 <dfarrell07> #endmeeting