16:00:41 #startmeeting Ansible Network Working Group 16:00:41 Meeting started Wed Oct 10 16:00:41 2018 UTC. 16:00:41 This meeting is logged and archived in a public location. 16:00:41 The chair is Qalthos. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:41 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:00:41 The meeting name has been set to 'ansible_network_working_group' 16:01:15 #chair dagrawal ganeshrn gundalow ikhan justjais nilashishc pabelanger privateip rcarrillocruz trishnag 16:01:15 Current chairs: Qalthos dagrawal ganeshrn gundalow ikhan justjais nilashishc pabelanger privateip rcarrillocruz trishnag 16:01:30 #topic Core Updates 16:01:49 #info Ansible 2.7 was released last week! 16:02:18 o/ 16:02:25 who was at fest? 16:02:27 any feedback 16:02:29 You can find it in all the usual places, which for most cases means pypi 16:02:36 o/ 16:02:45 \o/ 16:03:02 ansiblefest was great, I thought it had a lot of energy this time around 16:03:09 not to say others did't have energy 16:03:12 o/ 16:03:21 but 1400 people talking about ansible was awesome 16:03:28 #info AnsibleFest was last week as well 16:03:31 I wasn't there, but I heard it was nice 16:04:26 Not a lot of updates from the team this week due to fest 16:05:15 (but since many are here, they can talk about that if they want) 16:06:02 * privateip waves 16:06:07 Beyond that, the agenda can, as always, be found at 16:06:09 #link https://github.com/ansible/community/labels/network 16:08:28 Right, then, let's get on with it 16:08:45 #topic A brief note on HTTPAPI 16:09:38 I tried to cover the concerns raised in the issue, and I would have loved to sit down with you and talk about it if I'd been there 16:10:30 The gitst is that httpapi has had some development over the 2.7 cycle to make it friendlier to uses other than the one it was designed for 16:11:15 If you want to use it or can't figure out how to do something, let me know and we can talk about how to make it better 16:12:23 That out of the way, we have some proposals on the agenda 16:12:35 #topic Interface manager role 16:12:46 #link ansible-network/network-proposals#1 16:13:09 privateip: do you want to take this? 16:13:12 sure 16:13:25 i think most (hopefully) of the details are in the proposal 16:13:43 the proposal covers creating a new platform agnostic role for network interfaces 16:13:52 this would ultimately replace net_interface in core 16:14:05 looking for comments / critiques / suggestions / opinions 16:15:27 * privateip waits for Jeopardy! theme song to stop playing in his head to continue 16:15:30 i've commented all of my comments on it i think 16:15:36 are you good with it? 16:15:43 yeah i'm good with it 16:16:01 does the extensions make sense to you? 16:16:07 totes 16:16:15 excellent 16:16:29 i was just interested in agreeing that "this is what we'll call vendor extensions 16:16:52 as far as "put them in a dictionary and it passes through" that's a-ok with me 16:17:13 yep, we will carry that same pattern through to every role 16:17:21 vendor should also be responsible for documenting in their role those extensions 16:17:27 yes correct 16:17:30 k 16:17:36 that should be done in their lower order provider role 16:17:43 correct 16:18:10 ok i think we are good to move to the next topic 16:18:29 #topic VRF definitions role 16:18:34 #link ansible-network/network-proposals#3 16:18:53 same conversation just different role 16:19:06 the role was intentionally kept brief 16:19:23 i know there is quite a bit more we can (and will) add to the model 16:19:27 but wanted to start somewhere 16:19:51 since we release every other week, i took the short path to get the role work started 16:19:56 anyone disagree with this approach? 16:20:15 i don't know how my platform would conform to this, but that's a work item for me. the gist of the proposal i'm ok with 16:21:11 are there model changes that would help? 16:21:20 we can defer a week if you want some more time to absorb it 16:21:32 i got it published a little late for this weeks meeting 16:21:33 no, its more of a problem of i believe my product supports this...i just dont know where 16:21:39 ok 16:22:45 as you absorb it, please lets discuss if we need to make changes ... with the roles construct and release cycle we have infinitely more ability to make changes on the fly 16:22:53 k 16:23:51 i guess we are on to vlans 16:23:53 #topic VLANs role 16:23:57 #link ansible-network/network-proposals#2 16:24:05 that was some impressively fast typing :) 16:24:19 had a question about vlans proposal actually 16:24:19 You got to it while I was typing (: 16:24:26 how does it differ from this role? https://github.com/ansible-network/configure-vlan 16:24:48 we will be folding configure-vlan into this proposed role 16:25:00 can we annotate that on the proposal? 16:25:07 yes absolutely! 16:25:15 i can take that as an AI 16:25:53 #action privateip Update description that configure-vlan will be folded into this role 16:27:28 general question.... is there enough detail in the proposals? 16:28:27 we could probably use a section that outlines what the generic params will be? 16:28:39 or are vlan, name, status, and state all of them 16:28:45 vlan_id* 16:29:37 I'm leaning towards no, but there's not a lot of functionality for these proposals given the early-and-often approach 16:29:41 for this initial release those are it unless you have others you want to propose 16:29:44 So I'm not actually sure what more to add 16:29:59 is vlan_id another way of saying "vlan tag"? 16:30:07 i think we need a generalized "role architecture" write up 16:30:10 yes 16:30:13 k 16:30:23 that we can link to in the proposals 16:30:43 to address things like caphrim007_ commentary around extensions 16:31:11 you can tag me for that AI as well 16:31:15 privateip: So the job of these roles would be to generate vlan/vrf/interface config and then utilize the provider roles' config_manager function to push it to the device? 16:31:49 I mean the provider specific load functions. 16:32:10 sorta ... the high order role (vlans in this case) provide data validation and any munging (if necessary) ... the lower order roles (provider roles) are wholly responsible for the implementation 16:32:23 config_manager is not involved 16:32:24 #action privateip create role architecture writeup that can be linked in to role proposals to answer common questions regarding roles 16:32:25 Oh got it. 16:32:55 privateip: the top level 'state' option is provided for intent check right? 16:33:03 at least not required (some roles may choose to re-use config_manager tasks) 16:33:09 ganeshrn: correct 16:33:33 in that case we will might need the delay option as well 16:33:39 the overall idea is to give complete control to the provider for implementation 16:33:51 privateip: on that note, more detail on the purpose of each option could be added to the proposals 16:33:58 ganeshrn: two functions... configure and validate 16:34:13 playbook writer is responsible for any delay if both are used back to back 16:34:23 Qalthos: +1 16:35:01 #action privateip describe each role option on proposals 16:35:47 any final comments? 16:35:52 privateip: does that mean there should be a validate.yaml function as well? 16:35:59 yes 16:36:01 or is that in configure? 16:36:03 and there is 16:36:21 network_interfaces support config and validate 16:36:25 vlans just config (for first release) 16:36:34 vrf_definitions just config (for first release) 16:36:44 we will add validate to both vlans and vrf in subsequent releases 16:36:56 i anticipate a third function as well 16:36:58 facts 16:37:31 so over time a higher order role will support functions: configure, validate, facts 16:37:40 but we can phase this in over release cycles 16:37:49 +1 16:38:18 the whole intention of bi-weekly release cycles is to avoid huge feature addtions 16:38:25 privateip can we also make a configurable 'function' in these roles 16:38:39 dagrawal: can you explain your idea? 16:38:48 provider roles can implement it the way they want 16:39:26 e.g. get_existint_vlan function for vlan role 16:39:53 higher order role does not know there is a function like it 16:40:17 and it can be passed as kv pair 16:40:26 ah yes 16:40:35 definitely 16:41:03 i will try hard to get the draft of the role architecture document done by next week and there we can define abstractly how that would function 16:42:46 Anyone have any other comments on either the VLAN role or role direction in general? 16:43:30 (that second half may have been way more broad in scope than I meant) 16:44:48 We have one more last minute entry on the agenda 16:45:02 #topic Lenovo cleanup PR 16:45:53 Anil_Lenovo: There is a bit much here to go through right now, but I'll look it over and try to get something back to you today 16:46:17 #link https://github.com/ansible/ansible/pull/46623 16:46:26 Qalthos: Thats Ok take your time 16:46:32 #topic Open Floor 16:46:49 I can give a little update on zuul / infrastructure things 16:47:15 #topic zuul / infra 16:47:49 pabelanger: go for it 16:48:07 While I was at ansiblfest last week, I ran into logan- who is a public cloud provider. Was talking about our future infrastructure needs, and he's been able to via us some credit for POC using nodepool / zuul on limestone. The super cool things about it, is it took only a few hours to bring online 2 more regions for our testing 16:48:29 so today, we have 2 public cloud providers (openstack), giving 4 regions for testing 16:48:36 making our jobs much more redundant 16:48:57 neat 16:49:20 I also had a meeting with packet.net this week, again to help bootstrap some infrastructure things. And again, they've give us some credits so we can do a POC atop of their baremetal servers 16:49:39 which, once working could be another region for clouds or maybe k8s / openshift 16:50:38 So far, both regions vexxhost / limestone have been working great, and we've had no outages there 16:52:31 I have one topic to discuss pertaining to cumulus network modules. Can I ? 16:52:59 sure, that is all I had 16:53:00 Anil_Lenovo: Sure 16:53:21 One of my customer has both Lenovo and Cumulus switches. 16:54:05 And he want to use image transfer and vlan configurations using our respective modules 16:54:28 but for cumulus the modules were there, they are all deprecated now 16:55:06 and they have a single module only now viz. CCLU 16:55:09 NCLU 16:55:21 #topic cumulus modules 16:55:44 any one from cumulus here to help me out ? 16:56:29 what are their plans on deprecated module like image download etc 16:57:34 it comes here lib/ansible/modules/network/cumulus 16:58:55 Next query I have is that they dont have example on vlan configurations ? Will it work with nclu.py ? 17:00:31 I'm not sure myself, I think we might need to take an action item here and loop back next week 17:01:59 Yeah. Some contact point is also fine 17:02:21 I can then chat with them privately 17:02:38 my understanding from talking with them is that nclu is the only module they are going to write 17:02:55 whether or not it supports vlans... i dont know 17:03:07 ok 17:03:21 that said, we dont have to wait for cumulus 17:03:46 we can develop modules in this community 17:03:51 if someone is willing to take on the task 17:05:40 then we need to know the cli syntax and may need a switch too 17:10:02 Well it doesn't sound like we'll be getting anywhere with that here and now 17:10:52 VM? 17:10:59 CumulusVX? 17:12:35 No issues. Will see next week, they may review the meeting minutes and will respond. 17:14:20 A better course might be to open an issue against the NCLU module, though I can't say if they'll see that either 17:15:08 I will try that 17:15:36 We're a bit over time, so thanks everyone for coming 17:15:46 #endmeeting