15:35:40 <sergmelikyan> #startmeeting fuel@opnfv
15:35:40 <collabot`> Meeting started Tue Feb 28 15:35:40 2017 UTC.  The chair is sergmelikyan. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:35:40 <collabot`> Useful Commands: #action #agreed #help #info #idea #link #topic.
15:35:40 <collabot`> The meeting name has been set to 'fuel_opnfv'
15:35:47 <sergmelikyan> #topic Rollcall
15:35:51 <sergmelikyan> #info Serg Melikyan
15:35:53 <sergmelikyan> o/
15:37:26 <billyo_> #info Billy O'Mahony
15:38:10 <sergmelikyan> billyo_: Hi Billy, I should get first DPDK results by Wed :)
15:39:00 <billyo_> sergmelikyan: great thanks looking fwd to it
15:40:12 <davidchou1> #info David Chou
15:42:06 <sergmelikyan> #topic CI Status
15:42:13 <sergmelikyan> folks, we have number of bugs in CI
15:42:24 <sergmelikyan> and I need your help with these bugs
15:42:41 <sergmelikyan> Specifically this one https://jira.opnfv.org/browse/KVMFORNFV-75
15:42:59 <sergmelikyan> looks like pma_ got stuck with this bug, any chance you guys can help?
15:43:09 <sergmelikyan> davidchou1, billyo_ ?
15:43:22 <billyo_> just looking at it now
15:44:12 <sergmelikyan> we need to make all gates green ASAP, if we want to release Danube 1
15:48:49 <billyo_> I presume that the nofeature scenario is good ?
15:49:39 <sergmelikyan> there are other bugs, specifically with Doctor
15:49:50 <sergmelikyan> it leads to failure to nofeature scenario
15:50:08 <Greg_E_> hi serg
15:50:15 <sergmelikyan> Greg_E_: o/
15:50:22 <Greg_E_> how are we doing for the docs milestone
15:50:47 <Greg_E_> David was pinging folks yesterday
15:50:56 <sergmelikyan> our docs didn
15:51:04 <sergmelikyan> 't change much from last release
15:51:07 <sergmelikyan> even screens are same
15:51:26 <sergmelikyan> question is how to publish them to the right places, I will solve that on this week
15:52:25 <pma_> about doctor disable, it has been delayed for 1 week: https://gerrit.opnfv.org/gerrit/29483
15:53:29 <billyo_> sergmelikyan: there is a new structure for the doc in Danube. It's actually very straightforward to convert to the new docs structure (content changes are a different story of course)
15:53:41 <billyo_> #link https://docs.google.com/presentation/d/1_PszNevu1J7KK_46p8K9L4y5AjbnDWnXfsxQN-SmETI/edit#slide=id.g20d9418851_0_0
15:54:01 <sergmelikyan> billyo_: thank you!
15:54:04 <billyo_> that links show the procedure - only takes about 15mins. Just some git mv's
15:56:16 <sergmelikyan> billyo_: so what about kvm bug?
15:56:33 <sergmelikyan> any thoughts? or any names of people who I should ping for help?
15:56:56 <Greg_E_> Billy have to tried to generate packets inside a VM for performance?
15:57:29 <Greg_E_> what is the max number of packets that you were able to achieve?
16:02:28 <billyo_> sorry. back now.
16:04:26 <billyo_> sergmelikyan: at least the VM is getting an interface. Looks like issue would be with path to the dhcp agent on the controller. If this issue is also present on nofeature scenario then is not dpdk specific so we should debug the simpler non-dpdk scenario.
16:05:08 <billyo_> CI guys sometimes will remove jobs from CI pod to allow a manual run and then time to go in and have a look around.
16:05:38 <billyo_> Very hard to debug thise things on live CI system - really need a dedicated pod.
16:06:41 <billyo_> Greg_E_: with dpdk 16.11 and ovs 2.6.1 iperf3 (default tcp test)  will/should get about 4 to 5 Gbps VM to VM.
16:07:39 <Greg_E_> what about using a DPDK based packet generator
16:07:53 <Greg_E_> we need to get 25 mpps through OVS for a customer test
16:07:59 <pma_> <billyo_> it works fine even on pure ovs & with pure kvm scenario, but ovs+kvm doesn't
16:08:51 <pma_> <billyo_> kvm plugin installs kernel 4.4.6-rt4nfv on computes
16:09:11 <billyo_> I haven't actually used a dpdk s/w pkt generator yet. I know their are some around - moongen & t-rex. But it's on my short term agenda.
16:10:42 <Greg_E_> can you ping folks at Intel to see if anybody else has
16:10:43 <billyo_> pma_: ah ok. I'm meeting with the scenario owner this evening so I'll suggest to him that he really needs to get this setup on a pod somewhere/how so then it can be debugged.
16:11:08 <Greg_E_> we are running into bottlenecks around 5.5mpps in the VM
16:11:13 <Greg_E_> and we do not understand why
16:11:25 <Greg_E_> host can do 13-14mpps
16:11:30 <Greg_E_> but VM cannot
16:12:03 <pma_> <billyo_> thanx
16:12:17 <billyo_> pma_: "ovs & with pure kvm scenario" is there actually such a kvm4nfv only scenario? I thought the scenario was just kvm+dpdk
16:13:33 <davidchou1> Regarding the https://jira.opnfv.org/browse/KVMFORNFV-75 issue, I will ask kvmfornfv group take a look.
16:15:14 <billyo_> Greg_E_: what dpdk based pkt gen are you using in the VM? I know there is a public one called dpdk-iperf but I have heard bad reports about it. t-rex seems to be the most widely used within Intel.
16:15:34 <Greg_E_> moongen
16:15:39 <Greg_E_> pktgen
16:16:03 <sergmelikyan> davidchou1: thx!
16:17:10 <billyo_> Greg_E_: what is the ovs and dpdk versions you are using?
16:18:16 <pma_> billyo_> kvm4nfv only scenario: https://git.opnfv.org/fuel/tree/deploy/scenario/ha_nfv-kvm_heat_ceilometer_scenario.yaml
16:18:33 <billyo_> pma_: I stand corrected ;)
16:23:41 <OPNFV-Gerrit-Bot> Merged fuel: patch-packages: Disable mechanism once again  https://gerrit.opnfv.org/gerrit/29459
16:29:29 <Greg_E_> this is for MOS 9.2
16:29:36 <Greg_E_> so 2.6.1
16:29:42 <Greg_E_> DPDK 16.07
16:57:30 <sergmelikyan> #endmeeting