15:00:14 #startmeeting openstack-helm 15:00:15 Meeting started Tue Aug 27 15:00:14 2019 UTC and is due to finish in 60 minutes. The chair is portdirect. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:16 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:19 The meeting name has been set to 'openstack_helm' 15:00:19 o/ 15:00:29 lets give it a few mins for people to arrive 15:00:43 agenda is here: https://etherpad.openstack.org/p/openstack-helm-meeting-2019-08-27 15:00:48 #topic rollcall 15:01:13 o/ 15:01:17 \o 15:01:41 o/ 15:01:48 \o/ 15:01:51 o/ 15:01:59 o7 15:02:31 \o/ 15:02:31 | 15:02:31 \ 15:02:35 buu 15:02:41 o/ 15:02:44 didnt work, I forgot to escape it 15:03:02 itxaka: that looks painful, i hope you are ok 15:03:22 o/ 15:03:59 ok - lets go 15:04:11 #topic - time 15:04:41 hey, just thought i should communicate this - partly as im so excited about it 15:04:48 o/ 15:05:08 ive recently changed roles within my employer 15:05:20 oh? 15:05:24 :o 15:05:27 that's for open discussion? 15:05:31 lmao 15:05:44 and so should now be able to devote an approprite amount of time moving forward 15:06:04 portdirect unleashed and unshackled!! 15:06:13 ok, so topi is worng, I read it as "minus time" 15:06:16 i really need to thank both mattmceuen and srwilkers so much for helping cover my absence over the last few weeks 15:06:17 mattmceuen: haha 15:06:20 like less time 15:06:25 weren't you already full time portdirect? 15:06:44 yes - but the flux capactitor was broken 15:06:51 no problem :) 15:07:03 he's now able to hit 88 miles per hour 15:07:09 great scott! 15:07:14 wear a helmet 15:07:15 srwilkers: itxaka hahaha 15:07:35 portdirect: I see 15:07:42 anyway - lets move on 15:07:50 good news, congrats I guess? :) 15:07:56 #topic value overrides for gates 15:08:10 georgk: i think this is you (and also a great topic) 15:08:19 thanks 15:08:48 primarily just a questions as i ran into this as part of a review on my dpdk patches 15:09:16 when checking the gate scripts, they don't use the values_overrides functionality yet 15:09:28 I didn't find a good example at least 15:09:46 so, shall we extract the gate-specific overrides? 15:10:03 and if so, how to model this 15:10:25 its hard, wile we are transitioning for sure 15:10:47 i think neutron and keystone between them offer some good examples here 15:11:44 yeah keystone uses a feature_gate for ldap I think 15:11:51 in particular the keystone ldap job, where we make use of the 'feature gate' functionaility: https://github.com/openstack/openstack-helm/blob/master/zuul.d/jobs-openstack-helm.yaml#L55-L74 15:11:55 yes, I saw that 15:12:05 and neutron, where we have release specific over-rides 15:12:08 but I wasn't feeling good about modeling a gate configuration as a feature 15:12:15 and I sent one patch today for testing the resource which also uses that: https://review.opendev.org/#/c/678578/ 15:12:19 like, running non-ha in a gate 15:12:26 is that a feature gate? 15:12:35 i think it should be 15:12:45 either a feature or a bug, your choice :p 15:12:51 :-) 15:13:11 yeah - 'feature' doesn't always mean good :) 15:13:48 so, looking at this https://review.opendev.org/#/c/643284/33/neutron/values_overrides/dpdk-rocky-ubuntu_bionic.yaml 15:14:04 I could split this into a gate.yaml and a dpdk.yaml as corresponding overrides 15:14:06 i think this is a good ref for where the term comes from: https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/ 15:14:16 georgk: that would be great! 15:15:03 ok, if there is agreement, fine with me 15:15:22 +1 15:15:31 I can give it a try, but I'll go on vacation soon, so rihabb will probably take over (just FYI) 15:15:56 georgk: would you mind if I take a poke at the ps sometime this week? 15:16:47 please poke it 15:16:57 carefully... 15:16:59 ;-) 15:17:05 will do :D 15:17:37 * portdirect looks at: 15:17:37 * portdirect https://www.amazon.com/M-Z-Livestock-Electric-Cattle-Batteries/dp/B07GRKK6GP/ref=asc_df_B07GRKK6GP/?tag=hyprod-20&linkCode=df0&hvadid=343257889873&hvpos=1o1&hvnetw=g&hvrand=8404574012282221268&hvpone=&hvptwo=&hvqmt=&hvdev=c&hvdvcmdl=&hvlocint=&hvlocphy=9022859&hvtargid=aud-801381245258:pla-750808049640&psc=1&tag=&ref=&adgrpid=68270154919&hvpone=&hvptwo=&hvadid=343257889873&hvpos=1o1&hvnetw=g&hvr 15:17:37 * portdirect and=8404574012282221268&hvqmt=&hvdev=c&hvdvcmdl=&hvlocint=&hvlocphy=9022859&hvtargid=aud-801381245258:pla-750808049640 15:17:55 oh geez thanks 15:18:05 100% SATISFACTION GUARANTEED 15:18:06 I totally needed that in my amazon history 15:18:16 oh, i think i can get that into the office 15:18:22 EASY TO CARRY -> 31" 15:18:28 gives me a way to get portdirect's attention when i need to 15:18:40 * portdirect regrets link 15:18:43 srwilkers: I see 15:18:57 srwilkers: I thought it was more: https://www.amazon.com/Agrilabs-VetGun-Delivery-System/dp/B00L2G5RSG/ref=pd_sbs_328_1/144-3730140-0510016?_encoding=UTF8&pd_rd_i=B00L2G5RSG&pd_rd_r=dcdad721-2509-453a-858b-7223d52539ca&pd_rd_w=5ou42&pd_rd_wg=5Ed1M&pf_rd_p=43281256-7633-49c8-b909-7ffd7d8cb21e&pf_rd_r=P5SMGN01WZG0D51RG82N&psc=1&refRID=P5SMGN01WZG0D51RG82N 15:19:14 that's also acceptable 15:19:15 Only 7 left in stock, get yours soon! 15:19:45 ok - sorry georgk 15:20:20 untill we got distracted with office equipment, was this helpful? 15:20:32 :-) 15:20:44 yes, I got all the infos I was looking for 15:20:49 maybe too much actually... 15:21:11 awesome, thanks dude 15:21:24 #topic mariadb helm chart: Be closer to mariadb upstream helm chart? 15:21:30 evrardjp: i think this is you? 15:22:07 oh my, why did you link those... 15:22:16 now Im on a spiral of reading the comments in there 15:22:34 and yeah, that has to be evrardjp! 15:23:11 "does it work for cats and dogs?" oh my 15:23:21 haha 15:23:30 yeah it's me 15:24:08 Today a coworker just noticed that our maria helm chart could deserve some love to be image agnostic (due to presence of hardcoded paths) 15:24:39 As an action item, we'll make that more flexible with the current mechanism we have but... 15:25:07 The said image is very close to the upstream image on dockerhub, with the same kind of entrypoint etc. 15:25:17 so my question is the following: 15:25:35 Should we try again to be closer to upstream mariadb, so that we don't have to maintain it? 15:25:56 That could be either using the mariadb official helm chart (under beta!) or the mariadb operator 15:26:01 (under beta!) 15:26:20 so the image is actually based on the upstream image, just with the python k8s client added: https://github.com/openstack/openstack-helm-images/blob/master/mariadb/Dockerfile.ubuntu_xenial#L12-L15 15:26:23 is there someone evaluating those now that they have gained maturity? 15:26:30 we'd need to review again 15:26:38 ok 15:26:59 but last time we looked the chart in 'helm/charts' had a couple of differences 15:27:28 its active/passive ratehr than active/active 15:27:39 here I say there are two sources to look for, but indeed the one in helm/charts is the most visible one 15:27:46 which would make it more useful in some circumstances, and less in others 15:27:47 it's still set up that way portdirect 15:27:51 i just took a look 15:28:18 it also had some resulanccy issues in that recovering from a 'hard' outage was very tricky 15:28:33 *resiliency 15:28:43 but this may well have changed? 15:28:47 cant we send our changes to the upstream one? or are they reluctant to be accepted? 15:28:56 not sure, but was wondering if it was evaluated 15:29:08 upstream = official chart 15:29:10 itxaka: I like this approach 15:29:21 I prefer not having to maintain stuff. :D 15:29:33 exactly :D 15:29:33 its cretainly worth a shot - shelling out tech debt is one of the goals we have 15:30:03 i think it's important to note though 15:30:40 i also remember some security issues, and we'd need to evaluate the alignment of the values layouts 15:30:53 s/values/config 15:30:57 that while we have a mariadb chart in openstack-helm-infra, i dont think it's the only option for an openstack-helm deployment 15:31:26 it's just the option we've provided for those looking for a solution that aligns closely with the chart design patterns we've followed 15:31:35 this is very true 15:33:04 its also woth noting the chart in helm/charts relies on logic built into the bitnami images 15:33:10 so while im not opposed to exploring the deltas, i think at this point it's important to keep in mind the tradeoffs from differing points of view 15:33:22 srwilkers: what's the alternative? Because I think the design patterns are not broad, so it might require some hard work. But I agree on the fact it's free for the user, which is what I like :) 15:33:46 srwilkers: it would be nice to list and document those 15:34:04 I mean things are done for a reason, yet it's not explicit, because it's not documented 15:34:28 then it is in the memory of a few, which might or might not be active in the project anymore 15:35:12 anyway, I have my answer, thanks for the info! 15:38:20 ok - i think thats it for the agenda this week 15:38:40 any other topics we should be aware of before moving into reviews? 15:40:06 #topic reviews 15:40:30 one of the services we have been sorely lacking is designate 15:40:39 so its great to see a chart come in for it: 15:40:46 https://review.opendev.org/469981 (Designate chart) 15:41:10 one of the chalenges with this service has been gettign a backend that works well in k8s 15:41:37 in a past life, i had to do this (not fun), and ended up with the same conclusion as the author has here 15:42:00 in that powerdns seems to make the most sense as an initial backend to leverage: https://review.opendev.org/676142 (PowerDNS chart) 15:42:08 (in turn using mysql) 15:42:29 the layers run deepeth 15:42:33 would be great to see these merged, and then neutron updated to be able to take advantage of them 15:44:27 also georgk's and cheng1 's ovs dpdk work is in the final straight - lets get it home 15:45:11 sounds good to me 15:45:12 +1 :) 15:47:36 ok - if thats it lets give everyone 10 mins back 15:47:48 thanks everyone, see you in #openstack-helm 15:47:53 #endmeeting