08:01:28 <gongysh> #startmeeting tacker
08:01:29 <openstack> Meeting started Tue Jul 17 08:01:28 2018 UTC and is due to finish in 60 minutes.  The chair is gongysh. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:01:30 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
08:01:32 <openstack> The meeting name has been set to 'tacker'
08:01:41 <gongysh> #topic roll call
08:02:04 <dkushwaha_> o/
08:02:20 <tpatil> Hi
08:02:21 <YanXing_an> o/
08:02:31 <joxyuki> hi team
08:02:37 <gongysh> hi
08:03:02 <Li_Jiale> o/
08:06:25 <gongysh> #topic bp
08:06:27 <gongysh> YanXing_an, Li_Jiale hi
08:06:46 <gongysh> how is the barbican task doing?
08:07:31 <Li_Jiale> conding finished. now testing.
08:07:46 <gongysh> Li_Jiale, great.
08:07:54 <gongysh> thanks
08:08:14 <gongysh> when will we be ready for merge?
08:09:20 <Li_Jiale> Before Sunday, i think.
08:09:51 <Li_Jiale> it's ok ? @gongysh
08:10:03 <gongysh> Li_Jiale, fine
08:10:10 <gongysh> look forward to it.
08:10:24 <dkushwaha_> Li_Jiale, could you please update patch, i would like to test it in parallel
08:10:46 <gongysh> dkushwaha_ is reaching out for help.
08:10:53 <phuoc_> me too :)
08:11:43 <gongysh> Li_Jiale, hi
08:11:50 <Li_Jiale> ok , update it in tomorrow .
08:12:11 <dkushwaha_> Li_Jiale, Thanks
08:12:24 <gongysh> phuoc_, I have +2 for the ns with vnffg patches.
08:12:48 <phuoc_> gongysh, thanks
08:13:04 <gongysh> let dkushwaha_  +2 and have a test.
08:13:28 <dkushwaha_> gongysh, phuoc_ sure, i will test and give my +2+A
08:13:52 <gongysh> phuoc_, but https://review.openstack.org/#/c/558166/45/tacker/common/eventlet_utils.py
08:14:11 <gongysh> I still don't know why you modify the eventlet monkey patch here?
08:14:29 <gongysh> is there any problem if we don't change this code?
08:15:27 <dkushwaha_> gongysh, phuoc_ meanwhile, I will finish functional test patch by tomorrow for ns-vnffg.
08:15:48 <gongysh> dkushwaha_, it is kind of you. thanks
08:15:49 <phuoc_> actually, I changed it to the past
08:16:24 <phuoc_> That doesn't harm anything
08:18:08 <gongysh> phuoc_, if it doesn't harm anything, lets keep it as it is.
08:18:34 <gongysh> I am afraid it will lead to old problem which is fixed by that patch.
08:19:34 <phuoc_> gongysh, I mean I changed it to match with https://github.com/openstack/tacker/blob/stable/pike/tacker/common/eventlet_utils.py#L23
08:19:40 <phuoc_> before Queens version
08:20:20 <gongysh> it is changed by https://review.openstack.org/#/c/529116/
08:20:59 <nguyenhai93> https://review.openstack.org/#/c/558166/ should be rebased
08:21:00 <gongysh> I am afraid change back into https://github.com/openstack/tacker/blob/stable/pike/tacker/common/eventlet_utils.py#L23 will lead to the problem which is fixed by https://review.openstack.org/#/c/529116/
08:21:38 <phuoc_> but I recognized that my change make Tacker runs slowly
08:22:48 <gongysh> phuoc_, can we leave it alone in your vnffg ns patch, and file a bug and a new independent patch to change it?
08:23:17 <gongysh> I mean  not to change that event monkey in you vnffg ns patch.
08:23:32 <gongysh> if you need to change it, lets do it in an independent patch.
08:23:55 <phuoc_> It may run fail some functional test, because of bad performance
08:24:50 <gongysh> this way, we will  know it has nothing to do with the vnffg patch.
08:25:27 <gongysh> phuoc_ we can merge the eventlet monkey patch before the vnffg patches.
08:26:03 <phuoc_> gongysh, okie
08:26:10 <phuoc_> I will make a patch for that
08:26:13 <gongysh> fine
08:26:41 <tpatil> gongysh: I want to discuss about bp: Support affinity/anti-affinity policies for VDUs
08:27:01 <gongysh> tpatil, hi
08:27:07 <gongysh> your turn
08:27:15 <tpatil> #link : https://review.openstack.org/#/c/570845/
08:27:46 <tpatil> I have added negative test as suggested by dkushwaha which will run in single cell environment (nova)
08:27:59 <tpatil> name of the test is test_vnf_with_placement_policy_anti_affinity_on_single_comp_node
08:28:21 <tpatil> Adding multinode support in under development : https://review.openstack.org/#/c/578170/
08:28:50 <tpatil> Once this patch is merged, positive test written for anti-affinity will work in single cell environment
08:29:31 <tpatil> Request you all to please consider this feature for Rocky release as we are having hard time to make multinodes job work
08:30:06 <gongysh> tpatil, the multinode job is still failing.
08:31:04 <gongysh> tpatil, do you have idea how to fix the multinode job problem?
08:31:19 <tpatil> Yes, we are having hard time to make this multimodes job work. I have tested anti-affinity in single cell environment and it is working as expected
08:32:21 <tpatil> gongysh: No, I don't have the knowhow of CI jobs. but I have been successful in adding single cell support
08:32:47 <tpatil> #link : https://review.openstack.org/#/c/570845/13/.zuul.yaml
08:33:08 <tpatil> Need to set CELLSV2_SETUP: singleconductor in devstack localrc
08:33:21 <tpatil> and that is the reason why the negative test is passing
08:34:22 <tpatil> As explained in the patch, multi cells environment doesn't support anti-affinity. We plan to use single cell environment until this issue is fixed in multi-cells environment
08:35:52 <gongysh> dkushwaha_, what is your opinion for https://review.openstack.org/#/c/570845/?
08:36:39 <gongysh> http://logs.openstack.org/45/570845/13/check/tacker-functional-devstack/c9be107/testr_results.html.gz
08:37:12 <tpatil> You can see "test_vnf_with_placement_policy_anti_affinity_on_single_comp_node" functional test is passing
08:37:38 <dkushwaha_> gongysh, tpatil earlier, I had tested it on single node environment, but it was not as expected. I could not see updated patch, i needs to check that
08:38:33 <tpatil> dkushwaha: Did you setup CELLSV2_SETUP: singleconductor in devstack localrc before installing OpenStack?
08:39:09 <tpatil> devstack by default installs nova in multi-cells environment
08:39:18 <dkushwaha_> tpatil, i just used default settings
08:39:48 <tpatil> <dkushwaha_: No that's won't work becoz it installs multi-cell environment
08:40:53 <tpatil> dkushwaha_: as pointed out in https://review.openstack.org/#/c/570845/13/.zuul.yaml, it installs nova in single cells environment. Please try it and let me know if you face any issues.
08:41:08 <dkushwaha_> tpatil, yes, in multi-cell env, there is nova bug, and that is the issue
08:41:58 <tpatil> We plan to use this feature in single cell environment until this issue is fixed in multi-cell env. So request you all to please consider this feature for Rocky release
08:42:19 <gongysh> tpatil, your skip test is seen as failed tests, please make it as skipped.
08:42:22 <tpatil> I will try to make multinode job work but I cannot commit it will be done before the feature freeze
08:42:46 <dkushwaha_> tpatil, with single cell it may work, but with multi-celll we have issue and it is a known bug
08:43:30 <tpatil> gongysh: Will do in next PS
08:43:33 <gongysh> tpatil, don't use a inner decorator, you can use direct call of unittest.skipif function.
08:44:03 <dkushwaha_> tpatil, and another point is, we should have complete test coverage in multi-node job
08:44:12 <tpatil> gongysh: Sure. Will check and update next PS soon
08:44:33 <gongysh> tpatil, we can make it ready for merge after your fix.
08:44:34 <gongysh> thanks
08:44:50 <tpatil> gongysh: Thanks
08:46:02 <gongysh> tpatil, by the way, is there a way to do affinity with some existed vdu(s)?
08:46:14 <tpatil> dkushwaha_: Do you see any functional tests are missing? If yes please mention it on the patch. I will add them.
08:46:51 <tpatil> gongysh: you mean, already deployed VDUs?
08:47:10 <gongysh> yes, but it is a next story.
08:47:33 <gongysh> we can write it into your doc.
08:47:39 <dkushwaha_> tpatil, i have to see the latest patch, and will comment there if i have some comments
08:48:16 <tpatil> dkushwaha_: Sure, thank you
08:48:17 <dkushwaha_> tpatil, my just main point to have tested it on git
08:48:36 <dkushwaha_> :%s/git/gate
08:49:00 <gongysh> tpatil, one more question if I have just have one vdu, and write a affinity policy, what will happen?
08:50:00 <tpatil> gongysh: No error in that case. Heat actually create a server group internally, so it will place that VDU on one compute node. Nova will not complain
08:50:12 <gongysh> tpatil, ok
08:50:35 <gongysh> dkushwaha_, it is great if we can make the multinode ci job running
08:50:49 <dkushwaha_> gongysh, in case of single vdu, affinity things will not come not pic till you have previous vdu in same server group
08:51:21 <gongysh> and then merge the policy patch.
08:51:35 <dkushwaha_> gongysh, yeah, so there are 2 things. 1. complete test coverage as functional tests 2. multinode job setup.
08:52:05 <gongysh> however, if  we cannot make the multi job ci run, we have to test it manually.
08:52:21 <dkushwaha_> gongysh, ohk, after gate setup we can release it but by mentioning that this will not work for multi-cell because of  known nova bug
08:52:34 <tpatil> gongysh: That will be great.
08:52:42 <gongysh> we don't need to do multi cell
08:53:03 <gongysh> multi node desenot mean multi cell
08:53:04 <dkushwaha_> gongysh, IMO manually testing is not enough to release as it depends on nova bugs etc.
08:53:05 <tpatil> dkushwaha_: I agree. Should I update release notes accordingly?
08:53:30 <dkushwaha_> yeah, for multicell i agree that we can mention that limitation in reno.
08:54:05 <dkushwaha_> tpatil, yes, that is one way to release it so that user know it is not for multicell yet
08:54:20 <gongysh> tpatil, ok lets fix the multi node job and then merge the policy patch.
08:54:34 <tpatil> dkushwaha: Sure, I will add this info in the reno
08:55:00 <dkushwaha_> thanks.
08:55:23 <dkushwaha_> I will review the latest patch set.
08:55:39 <tpatil> gongysh: we will try our best
08:56:40 <gongysh> tpatil, lets try to fix it together.
08:56:40 <gongysh> thanks
08:56:45 <gongysh> #topic open discussion
08:56:51 <gongysh> anything else?
08:57:05 <gongysh> I have found we have registered one topic session for berlin summit.
08:57:09 <gongysh> thanks phuoc_  and dkushwaha_
08:57:17 <tpatil> gongysh: Sure, I will definitely need some help from the team. Thanks
08:57:38 <gongysh> tpatil, stay online, lets find a way.
08:58:05 <tpatil> gongysh: Ok
08:58:08 <gongysh> to get to the top of the mountain. :)
08:59:23 <gongysh> ok, thanks everyone
08:59:28 <gongysh> #endmeeting