08:31:59 <kubi001> #startmeeting Yardstick Work Meeting 08:31:59 <collabot> Meeting started Mon Apr 17 08:31:59 2017 UTC. The chair is kubi001. Information about MeetBot at http://wiki.debian.org/MeetBot. 08:31:59 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic. 08:31:59 <collabot> The meeting name has been set to 'yardstick_work_meeting' 08:32:06 <kubi001> #topic roll call 08:32:08 <kubi001> #info kubi 08:32:13 <rbbratta> #info Ross 08:32:29 <JackChan> #info jack 08:32:38 <Kanglin> #info Kanglin 08:32:44 <JingLu> #info JingLu 08:33:03 <Mingjiang> #info Rex 08:33:08 <kubi001> kalyan: ping 08:33:28 <kubi001> #topic action items follow up 08:33:32 <kubi001> #link http://ircbot.wl.linuxfoundation.org/meetings/opnfv-yardstick/2017/opnfv-yardstick.2017-04-10-08.32.html 08:33:46 <kubi001> #info AP1: Kubi will try to book a time slot of room for this plugfest 08:33:57 <kalyan> Hi all :) 08:34:16 <kalyan> #info kalyan 08:34:44 <kubi001> #info AP1: Done https://wiki.opnfv.org/pages/viewpage.action?pageId=9569132 08:35:38 <kubi001> #info kubi have booked the serveral time slots for yardstick, one for E-planning, two for integration with Ixia 08:35:53 <kubi001> kalyan: will you attend the plugfest 08:36:24 <kalyan> kubi001: no, i am not attending. Yunhong will 08:36:46 <kubi001> #info Yunhong Jiang from KVM4NFV will attend the plugfest 08:36:51 <kubi001> kalyan: thanks 08:37:27 <kubi001> rbbratta: are the time slots suitable for you? 08:37:41 <kubi001> rbbratta: I mean the plugfest yardstick time slot 08:37:49 <rbbratta> yes 08:38:03 <kubi001> rbbratta: great 08:38:23 <kubi001> please let me know if any time slot you want to change 08:38:38 <kubi001> #info AP2: yardstick team to review flavor create in heat context and feedback 08:38:58 <rbbratta> Martin was working on an update, I just updated the Jira with the new spec. 08:39:44 <kubi001> #info AP2: Done Martin was working on an update, Ross just updated the Jira with the new spec. 08:39:54 <kubi001> thanks for your work 08:39:55 <rbbratta> it will support inline flavor definition, and re-using flavors 08:40:21 <kubi001> #topic agenda for todya 08:40:24 <kubi001> #undo 08:40:24 <collabot> Removing item from minutes: <MeetBot.ircmeeting.items.Topic object at 0x248a290> 08:40:32 <kubi001> #topic agenda for today 08:40:34 <Mingjiang> that would be great 08:40:56 <kubi001> #link https://wiki.opnfv.org/display/yardstick/Meetings 08:41:31 <kubi001> #topic PTL election process 08:41:57 <kubi001> #link https://etherpad.opnfv.org/p/yardstick_ptl_election 08:42:30 <kubi001> #info kubi have initialized the etherpad for ptl election process 08:43:51 <kubi001> #action all, please review the ptl election process and give comment, we will finalize the process at plugfest 08:44:10 <kubi001> #topic E-planning 08:44:21 <kubi001> #link https://etherpad.opnfv.org/p/yardstick_release_e 08:45:08 <kubi001> I forwarded the email from kalyan to everyone 08:45:32 <kubi001> they have some requirements for yardstick framework 08:45:55 <kubi001> I guess we can put that into E-release to implenment 08:46:22 <rbbratta> there is also the proposed NFVBench project 08:47:15 <rbbratta> whatever NFVbench does will probably impact what we do for yardstick. 08:48:13 <rbbratta> we also need to discuss scale-out and scale-up tests 08:48:36 <kubi001> rbbratta: yes, good point 08:49:39 <Mingjiang> scale-out and scale-up tests +1 08:50:21 <rbbratta> or anything with dynamic contexts, unless we can use bottlenecks to do scaling 08:50:22 <Mingjiang> for NFVbench, we may need a discuss among yardstick team 08:50:32 <kubi001> rbbratta: what's your idea for NFVBench, From my view, I think we should let the yardstick be reused 08:51:15 <rbbratta> it really depends what the code looks like. If they aren't using Heat, and what they do for scale-out. 08:52:01 <rbbratta> and also what it does that VSPerf doesn't do. We had planned to integrate with VSPerf 08:52:43 <rbbratta> I see yardstick as using Heat to bridge the gap until we have some MANO. I don't see yardstick doing tests that don't use Heat. 08:52:57 <rbbratta> but if Heat is insufficient for NFVbench 08:54:46 <kubi001> I remembered that the evolution of VSperf should be data plane benchmarking, right? 08:55:30 <kubi001> https://wiki.opnfv.org/display/testing/TestPerf?preview=/https%3A%2F%2Fwiki.opnfv.org%2Fdownload%2Fattachments%2F8688867%2FOPNFVTestEcosystem%2528Danube-v02%2529.png%3Fversion%3D1%26modificationDate%3D1488487829000%26api%3Dv2 08:55:39 <rbbratta> yes, without Openstack, NFVbench is supposed to be with Openstack. That is good. 08:56:13 <rbbratta> Openstack usually uses double-briding, br-int and br-tun, which I don't think VSPerf usually does. 08:57:47 <rbbratta> I think probably NFVbench doesn't use Heat, but does manual config of VMs and Neutron to ensure correctness and reproducability. 08:59:15 <rbbratta> anyway, we will see 09:00:19 <Mingjiang> yes, we miss the details, now just with the proposal 09:01:01 <Mingjiang> do nfvbench go to this thursday tech-discuss meeting? 09:01:52 <rbbratta> april 20th 09:02:05 <kubi001> yes, one idea from me is that we should try to clarify with them and try to reuse the existing projects 09:02:46 <rbbratta> I'm sure both VSperf and yardstick would like to re-use. 09:03:37 <kubi001> yes 09:03:38 <Mingjiang> +1 09:04:10 <kubi001> kalyan: would you like to discuss your requirements now or with email? 09:05:51 <kalyan> kubi001: will discuss now, atleast what we need..then after an internal kvm discussion will share full details in an email 09:06:35 <kubi001> kalyan: OK, go ahead 09:08:02 <kubi001> 1. post-execute script 09:08:03 <kalyan> yes, so as mentioned https://etherpad.opnfv.org/p/yardstick_release_e first requirement is ..we need an option in yardstick to be implemented for executing some post execution scripts 09:09:01 <kubi001> kalyan: so the key point is that you need the post-execute script run after context undeploy, right? 09:09:35 <kalyan> this is required because, as a part of kernel debugging we need trace to be disabled soon after the test execution..so that we will avoid unwanted logs for effective debugging 09:09:43 <kalyan> kubi001: yes, exactly 09:11:17 <kubi001> rbbratta: what's your opinion? 09:11:32 <rbbratta> the script runs on the nodes? 09:12:05 <kalyan> rbbratta: yes, runs on the nodes 09:12:53 <rbbratta> I think we just added a teardown script section on node.py 09:13:09 <kalyan> so basically, we need post-execute script option..if required we wil pass the script as an argument to the yaml file .. 09:13:12 <rbbratta> which I update for ansible scripts. 09:14:34 <rbbratta> Jing Lu added for https://jira.opnfv.org/browse/YARDSTICK-573 09:15:04 <rbbratta> env: setup: - node1: script: .... 09:15:15 <rbbratta> that should already work. 09:15:34 <JingLu> the node context now supoort run config on nodes 09:15:46 <rbbratta> I'd prefer ansible playbooks over shell scripts, but it depends on the shell script. 09:16:16 <rbbratta> JingLu: do we have a better example that uses 'teardown'? 09:17:31 <JingLu> Here is a context example for cpu pinning test 09:17:35 <JingLu> contexts: 09:17:35 <JingLu> - 09:17:36 <JingLu> type: Node 09:17:36 <JingLu> name: env-prepare 09:17:38 <JingLu> file: {{file}} 09:17:38 <JingLu> env: 09:17:40 <JingLu> type: ansible 09:17:40 <JingLu> setup: cpu_pin_setup.yaml 09:17:42 <JingLu> teardown: cpu_pin_teardown.yaml 09:19:18 <JingLu> you can specify setup and teardown ansible playbook yaml in the env section 09:20:28 <JingLu> The post-execute script needs to be implemented using ansible 09:22:56 <JackChan> contexts: 09:22:56 <JackChan> - 09:22:56 <JackChan> type: Node 09:22:58 <JackChan> name: LF 09:23:00 <JackChan> file: {{ file }} 09:23:02 <JackChan> env: 09:23:04 <JackChan> type: ansible 09:23:06 <JackChan> setup: migrate_setup.yaml 09:23:08 <JackChan> teardown: migrate_teardown.yaml 09:23:10 <JackChan> - 09:23:12 <JackChan> name: migrate 09:23:14 <JackChan> image: yardstick-image 09:23:16 <JackChan> flavor: {{ flavor }} 09:23:18 <JackChan> user: ubuntu 09:23:20 <JackChan> placement_groups: 09:23:22 <JackChan> pgrp1: 09:23:24 <JackChan> policy: "availability" 09:23:26 <JackChan> servers: 09:23:28 <JackChan> jack: 09:23:30 <JackChan> floating_ip: true 09:23:32 <JackChan> placement: "pgrp1" 09:23:34 <JackChan> rose: 09:23:36 <JackChan> placement: "pgrp1" 09:23:38 <JackChan> floating_ip: true 09:23:40 <JackChan> networks: 09:23:42 <JackChan> test: 09:23:44 <JackChan> cidr: '10.0.1.0/24' 09:23:48 <JackChan> This is an example. 09:24:07 <JingLu> Yes I think we can orchestate a "node" type context after the "heat" type context to do the post-execute action 09:24:09 <JackChan> it will use ansible playbook to execute migrate_teardown.yaml after undeploy the stack. 09:28:26 <JackChan> The contexts is a stack. So when setup it will first deploy context1 then context2, when teardown it will first undeploy context2 then context1. 09:30:02 <Mingjiang> since the context is undeployed in a reverse way, so you need to add Node context with teardown in at the top? 09:30:21 <JackChan> Mingjiang: yes 09:30:53 <kubi001> kalyan: Is that make sense for you? 09:31:52 <Mingjiang> I think this solution requires that the post-execute script should be run on the node. 09:32:58 <kalyan> kubi001: yes, I am actually new to these terminologies, what i understood is under node context 'file:' option will be provided to execute the script..right? 09:34:35 <JackChan> kalyan: the file is nodes config. actually it will execute the migrate_teardown.yaml using ansible. 09:36:04 <kalyan> Jackchan: thanks for clarfying that, just to mention the post execute script we will be executing is a shell script.. 09:37:39 <JackChan> kalyan: it also support execute script by set the type keyword to 'script' in env. 09:38:05 <Mingjiang> does this shell scripts run on some specific nodes? 09:38:18 <kalyan> yardstick -d task start cylictest-node-context.yaml with this command we need an option to execute some scripts which we pass.... 09:38:59 <kalyan> yes, those will be run on nodes depending on the type of job verify/daily 09:40:31 <JackChan> kalyan: maybe I can send you an email to clarify it. What's your email address? 09:41:09 <kalyan> JackChan: sure, thanks reddyx.gundarapu@intel.com 09:41:36 <JackChan> kalyan: ok 09:41:55 <kubi001> #action JackChan will clarify the technical detail for kalyan's requiremetn with email 09:43:33 <kubi001> kalyan: As the time is quite over, About 2nd requirement, I think we can discuss in etherpad page, 09:43:58 <kalyan> sorry if any confusion caused, will discuss again on this point after getting some clarification from Jackchan mail. 09:44:26 <kalyan> https://gerrit.opnfv.org/gerrit/#/c/33407/ 09:44:44 <kubi001> kalyan: sure, will do 09:44:49 <kalyan> kubi001: this will do the second point i think 09:45:18 <kalyan> but i will wait for some inputs from yunhong, to check if he is ok or if something else to be added as well 09:47:01 <kalyan> kubi001: sure, will discuss in etherpad page if needed.. 09:47:56 <kubi001> #action please discuss the detail with the etherpad page https://etherpad.opnfv.org/p/yardstick_release_e 09:48:45 <kubi001> #topic Aob 09:48:51 <kubi001> anything else? 09:49:23 <rbbratta> I updated the ansible code, using some existing code from other project. 09:49:48 <kubi001> rbbratta: thanks 09:49:58 <rbbratta> https://gerrit.opnfv.org/gerrit/#/c/33411/ 09:50:29 <rbbratta> still trying to get providers networks working. 09:50:55 <rbbratta> Deepak I think is still working on Ixia code, but that may wait until plugfest when we meet with the Ixia people. 09:51:13 <rbbratta> that's it from my side. 09:52:16 <kubi001> thanks for for today' s meeting 09:52:22 <kubi001> #endmeeting