14:00:50 #startmeeting sahara 14:00:51 Meeting started Thu Nov 29 14:00:50 2018 UTC and is due to finish in 60 minutes. The chair is tellesnobrega. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:52 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:54 The meeting name has been set to 'sahara' 14:01:24 o/ 14:01:53 o/ 14:02:00 o/ 14:02:06 hi jeremyfreudberg, good to see you around 14:02:59 I'm giving a couple more minutes to see if anyone jumps in 14:03:54 #topic News/Updates 14:04:26 Split plugins got out of my plate and jumped into tosky's 14:04:40 for now, until I find bugs :) 14:04:55 I wrote the payload response patch on apiv2 14:04:58 it is up for review 14:05:11 yes, i will review it soon 14:05:32 and that is all I did 14:05:50 i don't have much to report either 14:06:50 I deployed devstack with sahara/split plugin (tellesnobrega, remember to fix the URL of the repositories in the devstack plugin) 14:07:19 will do, by the way, I worked the devstack plugin to deploy the split plugin 14:07:26 I tested two images so far, working (exception: EDP, thanks to a broken radosgw on my system, which I'm trying to fix) 14:07:41 I'm planning to test other images too 14:08:11 I also sent an email to rdo-dev@ asking for advices for the packaging, and when I have a plan there I will send the more complete email to openstack-discuss@ 14:08:36 I also need to complete the patch which adds support for S3 testing to sahara-tests 14:08:38 and that's it 14:09:16 moving on... 14:09:21 #topic APIv2 14:09:35 jeremyfreudberg, since you are here, we have some questions on APIv2 14:09:44 what is the current status, considering the payload patch is fine 14:09:53 what is left to do? how can we make this happen 14:10:16 M2 is around the corner, January 10th, we need to get it asap 14:11:42 do I remember correctly that we just need to deploy with unversioned endpoints? Do we need to do anything else to enable it (say) in devstack? And with other deployment tools? 14:11:50 according to my notes: 14:13:28 on the server side 1)tenant_id->project_id , 2)proper microversioning, 3)fix 500 on malformed query string 14:13:46 4) get as much of boot from volume enhancements and health repair done now, so we don't have to change the api again 14:14:43 on the client side, there is 1) OSC 2) adding new features like decomission of a specific node and keypair replacement to where they are missing 14:14:50 and of course, proper testing and API REF 14:15:10 we have a patch for OSC 14:15:25 #link https://review.openstack.org/#/c/584110/ 14:15:37 oh, right 14:15:56 please review it :) 14:16:10 along side #link https://review.openstack.org/#/c/620307/ 14:16:17 yes, it needs a re-review by me, certainly 14:17:12 about testing, tempest tests requires a new version of the client, but what about scenario tests? As they use saharaclient, should they be able to use v2 if we enable it? 14:17:18 so, if those two patches go in soon, the next big priorities are the other server-side patches 2,3,4 14:17:20 I think you tried to hack them to work with v2 14:18:18 can you explain more what the 500 on malformed query string is? 14:18:42 is it all around, or specific call? 14:18:49 tosky: to answer your earlier question about deployment: with unversioned endpoints, yes, and a new-enough python-saharaclient (although it should still be able to discover v2 from the old endpoint) 14:19:13 tosky: the scenario tests require some tweaks, because some of the names have changed in v2 14:20:35 tosky: https://review.openstack.org/#/c/536157 was a very haphazard/rushed/sloppy attempt 14:20:40 shouldn't saharaclient mostly abstract from the changes? 14:20:42 * tosky checks 14:21:08 tosky: it abstracts from the changes except for stuff like job/job template/job execution where we had to rename in order to have some amount of sense 14:21:15 tellesnobrega: for any call 14:21:29 oh, right 14:21:53 jeremyfreudberg, ok 14:23:11 tellesnobrega: any unexpected value in the query string besides the ones used for pagination eventually throw an error in sqlalchemy about not being able to find a table with the name of whatever value was in the string 14:23:39 do we have all the items (1-4 for server, 1-2 for client, + fixing the API, scenario tests and tempest tests) documented on storyboard? 14:24:10 tellesnobrega: it should be simple enough to fix, just have a whitelist of allowed things in query string which can be checked in flask.before_request (I'll take care of the patch when I have time) 14:24:12 tosky: uh, maybe 14:24:40 jeremyfreudberg, thanks, do you think you will have time anytime soon? 14:25:36 so: does anyone take the time to check if we have everything tracked on storyboard? :) 14:26:13 tosky, I will do that 14:26:19 oki, thanks 14:26:27 and update accordingly 14:26:43 there definitely is something somewhere (some individual stories?) 14:26:57 tellesnobrega: maybe not, until the end of the semester (3 weeks) 14:27:04 you can give that patch a try if you want 14:27:23 we need to organize some stuff 14:27:41 so maybe microversioning should take more time and would be more important to work on now 14:28:55 microversioning is definitely the trickiest (to have it really work and not only imitated), i have ideas in my head which are putting me on the right path to solving it, i think i should be able to accomplish it after the semester is over 14:29:19 i think it would be most helpful if you took the query string stuff off my plate so that i don't have to worry about it when i actually have free time 14:29:40 jeremyfreudberg, ok, I will leave microversioning to you and work ont he query stuff 14:29:51 email me if you have questions 14:29:55 will do 14:30:37 what about unversioned deployment? what do we need to do on that? devstack deployment? what else? 14:32:00 I had few patches but they were mostly not working 14:32:13 for puppet, packstack, kolla and tripleo 14:33:36 https://review.openstack.org/#/q/topic:bug/1782147 right? 14:33:42 or are there more with a different topic? 14:34:00 that is it 14:35:01 yep, probably the topic is not correct, but let's not debate it 14:35:54 ok, so we have a plan on apiv2 14:36:06 jeremyfreudberg, tosky please review the patches we have asap 14:36:16 I will start working on 500 issue 14:36:23 and update storyboard 14:36:55 lets try to make this happen, I don't want to postpone one more cycle on apiv2 14:37:41 agree 14:38:06 I have a question on the unversioned endpoint 14:38:14 yes 14:38:27 how it works, what do I need to do to have it working on devstack? 14:40:42 having the latest client is essentially enough for basic operation 14:41:05 assuming that you tell the client that's the version you want 14:41:21 same for osc and dashboard, you have to tell it, i think (I can't remember what we did for dashboard, actually) 14:41:33 i also realized that we are still doing unversioned endpoints in devstack: https://github.com/openstack/sahara/blob/master/devstack/plugin.sh#L38 14:41:56 probably that should be changed (the client can still discover v2 from the old kind of endpoint) 14:42:36 about dashboard-- 14:42:36 Enable it by setting the "data-processing" API version to "2" in the `OPENSTACK_API_VERSIONS` dictionary found in Horizon's local_settings.py. 14:42:48 we need to change that to "$SAHARA_SERVICE_PROTOCOL://$SAHARA_SERVICE_HOST:$SAHARA_SERVICE_PORT/\$(tenant_id)s" \ 14:42:51 ??? 14:42:57 at this point we can assume the latest client (that's what we do in master) 14:43:19 just "$SAHARA_SERVICE_PROTOCOL://$SAHARA_SERVICE_HOST:$SAHARA_SERVICE_PORT" 14:43:32 ok 14:43:44 i think we didn't do it yet because grenade had been breaking for other reasons, and we wanted to write a grenade test 14:43:48 but i may be misremembering 14:44:09 uh, you wrote "we are still doing unversioned endpoint in devstack", did you mean "we are still doing versioned endpoint in devstack"? 14:44:48 grenade should work, because the new version should still work when you have the old endpoint only 14:44:50 I think he meant versioned 14:44:56 yes, yes, the endpoints are still versioned right now, and need to be made UNversioned 14:45:03 an earlier typo 14:45:04 unless grenade requires us to write the code to also add the new endpoints 14:45:28 yes, i don't expect the changes to break grenade 14:45:39 one question about endpoints (maybe for keystone): can you have multiple endpoints for the same type (public, etc) and service type (data-processing)? 14:45:59 or should upgraded deployment change the existing endpoints instead of adding new ones? 14:46:29 it should probably change them 14:47:40 good to hear, as we need to tell people who works on deployment tools that support upgrades (like tripleo) 14:47:50 not sure whether openstack-ansible supports upgrades 14:48:04 tosky: out of curiosity, i just did this: http://paste.openstack.org/show/736419/ 14:48:12 not sure what the resulting behavior would be 14:50:33 those are the same endpoint for different types, not sure if you can do it with the same type 14:50:53 tellesnobrega: there are two "admin" 14:50:54 they are for the same type (admin) 14:51:13 oh, I see 14:51:21 yes, maybe we need to test the output of that 14:51:50 or ask on #openstack-keystone 14:53:42 I guess we have a plan, I will send a patch on devstack plugin soon 14:54:06 yep 14:54:11 thanks 14:54:30 6 minutes left 14:54:38 any other topics to discuss? 14:55:35 not from me 14:55:46 not from me 14:56:02 ok 14:56:12 thanks you both 14:56:18 thank you! 14:56:21 this was a really good meeting 14:56:24 see you all next week 14:57:35 jeremyfreudberg, wait on some emails on the 500 stuff, I will definitely need help 14:57:42 sure 14:57:44 thanks all 14:58:12 #endmeeting