21:01:16 #startmeeting Server Working Group Weekly Meeting (2016-12-06) 21:01:16 Meeting started Tue Dec 6 21:01:16 2016 UTC. The chair is sgallagh. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:01:16 Useful Commands: #action #agreed #halp #info #idea #link #topic. 21:01:16 The meeting name has been set to 'server_working_group_weekly_meeting_(2016-12-06)' 21:01:16 #chair nirik sgallagh mhayden dperpeet smooge jds2001 vvaldez adamw mjwolf 21:01:16 Current chairs: adamw dperpeet jds2001 mhayden mjwolf nirik sgallagh smooge vvaldez 21:01:16 #topic roll call 21:01:16 .hello sgallagh 21:01:17 sgallagh: sgallagh 'Stephen Gallagher' 21:01:23 .hello vvaldez 21:01:26 .hello dperpeet 21:01:26 .hello adamwill 21:01:27 vvaldez: vvaldez 'Vinny Valdez' 21:01:30 dperpeet: dperpeet 'None' 21:01:33 adamw: adamwill 'Adam Williamson' 21:01:42 .hello mhayden 21:01:43 mhayden: mhayden 'Major Hayden' 21:02:59 .hello jstanley 21:03:00 jds2001: jstanley 'Jon Stanley' 21:03:02 .hello smooge 21:03:04 smooge: smooge 'Stephen J Smoogen' 21:03:14 .hello langdon 21:03:15 langdon: langdon 'Langdon White' 21:03:25 Wow, full house today. 21:03:38 #topic Agenda 21:03:55 I didn't remember to put together a real agenda this week, but I have a couple topics anyway. 21:04:18 #info Agenda Item: Usenix LISA Booth 21:04:18 #info Agenda Item: NFS Server Role Discussion 21:04:23 Anyone have other topics? 21:05:55 nothing 21:05:56 Did I fall off the network again? 21:05:59 isn't lisa now 21:05:59 oh ok 21:06:11 Excellent segue! ;-) 21:06:11 sorry i got called into a wok meeting 21:06:20 #topic Usenix LISA Booth 21:07:00 who is running the lisa booth this week? 21:07:03 OK, so langdon and I will be operating the Fedora Server booth at LISA tomorrow. 21:07:08 Smooge: you making me dinner? 21:07:09 hah 21:07:28 no i am in north carolina 21:07:29 "wok" meeting. Ha. 21:07:46 i am wokkin on a meeting 21:08:07 :( 21:08:10 I am not, I’m in my basement this week 21:08:22 Fedora Server is in kind of a transitional place this year, so it's a little less obvious than in the past what we should be talking about. 21:08:37 * jds2001 in his apartment, which is on the third floor if that counts for anything :D 21:09:01 I'm probably going to yammer on about our new ansible-based role plans as well as preaching the gospel of Cockpit. 21:09:11 sgallagh: i'd talk about the future, about the roles. 21:09:14 sgallagh: +1 21:09:15 But I'd like to hear from all of you what you think will be the most attractive stuff. 21:09:31 Crazy server! 21:09:36 i wish i could be at LISA! :P 21:09:41 langdon: For the last time, we're not calling it that. 21:09:46 servers? who needs those anymore? :D 21:09:59 * langdon still working on the acronym 21:10:00 everyone has containers! 21:10:02 jds2001: haha, been watching a lot of re:invent coverage :P 21:10:51 Oh, actually I have a third topic to stick into the agenda later... today there was a mini-summit on Ansible Container that I attended. Some interesting stuff that could be relevant for the roles. 21:10:59 /me puts a pin in that for now 21:11:04 seriously, I think talking a bit about the container stuff could be valuable as well 21:11:08 and where we fit in there. 21:11:14 /me ndos 21:11:15 *nods 21:12:18 I'm probably going to avoid talking too much about the Domain Controller and DNS stuff that's been going on, mostly because the Identity Management folks from Red Hat have their own booth. 21:13:08 go ahead, steal their thunder :D 21:13:10 I'll probably throw together a couple VMs for people to poke at Cockpit 21:13:24 jds2001: Problem with stealing thunder is the lightning that comes with it 21:13:26 I think they are right next to each other though 21:14:30 OK, let's not spend too much time on this topic; anyone have a particular piece they *really* want us to talk up? 21:15:39 *crickets chirp* 21:15:55 sgallagh, you mentioned what I would say 21:15:55 sgallagh: It would be cool i fyou could speak to people about: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/DTXEZ3UACKA6VIVMVNCPLTB336BXIFJI/ 21:16:00 And get some direct feedback 21:16:16 geppetto: Ah, great idea! I will do that. 21:16:25 I imagine mattdm will do the same (he will also be around) 21:16:30 * geppetto nods 21:16:40 #info Talk about future of Server Roles powered by Ansible 21:16:52 #info Discuss future container plans 21:17:10 #info Show off Cockpit 21:17:35 #info Poll people on Fedora release frequency 21:18:09 #topic Ansible Container Mini-Summit 21:18:11 sgallagh, ok.. not to be too snarky... but 21:18:16 ... 21:18:18 why not cockpit in a container? 21:18:38 * langdon was wandering so couldn't type too fast 21:18:39 langdon: ? 21:18:48 you said VMs to play around with cockpit.. 21:19:01 we can discuss in person tomorrow.. may not be relevant for everyone 21:19:01 at this point I'm not aware that we can have all of cockpit in a container 21:19:13 langdon: Because I don't want them mucking with my actual laptop :) 21:19:40 also what dperpeet said 21:19:42 dperpeet, oh.. didnt know that 21:19:58 I concur with sgallagh's plan to use VMs 21:20:02 dperpeet: Atomic's version of Cockpit is a subset of functionality, right? 21:20:33 we're veering a bit off-topic here, but briefly put the web service part can run in a container 21:20:50 but the base parts need to be installed 21:20:57 which is what Atomic ships with in the ostree 21:21:50 So, Ansible Container: basically this was a meeting of a number of interested parties to discuss future use and requirements. 21:21:52 I'm happy to explain that in more detail, but probably better not here and right now 21:22:38 Quick overview: basically Ansible Container is a declarative replacement for dockerfiles with pluggable output to support multiple container formats (including Docker) 21:24:01 It looks quite appealing as a potential implementation mechanism for the server roles (as microservices). 21:24:32 One of the major benefits to it is that it builds OpenShift-compatible containers by default, which means an easier migration path between Fedora Server and Fedora Atomic. 21:24:34 sgallagh: i heard a little about it at ansiblefest 21:25:03 threw it on the list of "this looks cool, but I really don't see the usefulness in practical terms" 21:25:29 that would be nice, ideally it could output json needed for runc system containers I’d imagine 21:25:37 jds2001: That was my thought then as well. After today, I think there's direct value to what we're working on. 21:26:27 vvaldez: The output plugins look fairly straightforward, so I expect that wouldn't be terribly difficult to accomplish 21:27:01 nice 21:27:03 Anyway, I don't have a *lot* to report on this at the moment. Today's discussion opened the possibility that this could be a good fit for us, so I'm putting it on our radar. 21:27:32 (Surprises are good for birthdays, not for software development projects :-D ) 21:28:10 just fyi.. modularity team has a person specifically working on modularity + ansible/ansible-container 21:28:20 so we should have help for you there 21:28:21 sgallagh: my birthday is 2/6 :D 21:28:31 is that june or feb? ;) 21:28:39 langdon: Who is that person? 21:28:44 ttomeck 21:28:52 oops.. i think i dropped an e 21:28:53 Ah, ok. Yes, he was on the call. 21:28:58 ttomecek 21:29:28 OK, shall we move on to the NFS discussion? 21:29:34 Also if anyone wants to have a look easier, there is a copr: http://copr-fe.cloud.fedoraproject.org/coprs/james/ansible-container/ 21:31:25 geppetto++ 21:31:26 sgallagh: Karma for james changed to 1 (for the f25 release cycle): https://badges.fedoraproject.org/tags/cookie/any 21:31:52 #topic NFS Server Role Discussion 21:32:02 We have the google doc here: https://docs.google.com/document/d/1jLyKsECdHdlKltmHGgf_-iOKj-hj4Qjbh5Zgm7a-eMc/edit and andreasn updated the Cockpit feature page https://github.com/cockpit-project/cockpit/wiki/Feature:-NFS-Server 21:32:06 dperpeet, jds2001: Mind giving a quick... 21:32:10 ok, you beat me to it 21:32:12 :) 21:32:40 at this point we need to make sure that the user stories are good 21:32:45 dperpeet: I read the user story that Andreas put together. 21:33:10 I think we need to call out a few others (that may have some or much overlap) but are very important. 21:33:34 yeah, I iterated a bit also 21:34:07 I emphasized different permission settings 21:34:11 One extremely common example is NFS-mounted home directories. 21:34:16 right 21:34:52 sgallagh: do we want to include autofs configuration? 21:34:53 I think we also need to focus on getting Kerberos-protected shares done right. 21:35:10 jds2001: That's a client-side configuration; we can't easily do that on the server-side except via FreeIPA. 21:35:26 However, I *do* think we want to start looking at how to do that optimally with FreeIPA 21:35:37 I suggest adding feedback to the wiki page 21:35:40 Sorry, probably need some background info there. 21:35:43 so andreas can reference it there 21:35:57 or comment on trello https://trello.com/c/00TuMHlI/414-design-nfs-server-configuration 21:36:09 jds2001: I would +1 an autofs config, as it is there’s a default /net by just installing/enabling the package we could go from there 21:36:11 dperpeet: Sure, will do. 21:36:30 sgallagh: does kerberos have any impact on the UI? 21:36:32 vvaldez: /net is a special-case 21:36:56 we can add important stories that interact with other technologies if they are important 21:36:58 jds2001: Possibly. 21:37:00 vvaldez: im used to working in enterprise environments where we obliterate that :) 21:37:09 vvaldez: so i honestly completely forgot about it. 21:37:20 it doesn't mean that it has to be implemented right away, but certainly we should point that story out in the design phase 21:37:28 sure, same here, we use our own custom ones but still, very nice to have on new servers that get brought up 21:37:42 autofs itself, not just /net. ack sgallagh 21:39:12 Autofs setup would be nice, but I don't think it can be implemented without a way to "push" config to the clients. 21:39:51 I suppose we could offer ansible snippets that admins could download and add into their playbooks to configure the NFS clients, though. 21:40:02 But now I'm getting into implementation, so I'll stop 21:40:12 (though that might impact UI...) 21:40:16 sgallagh: really, it's a separate role 21:40:41 "file server client 21:40:47 " or something like that 21:40:49 jds2001: Well, not necessarily if we do the integration work necessary to update a FreeIPA config as well 21:41:04 that just updates maps 21:41:05 Because any SSSD client of FreeIPA can just pick up the autofs config 21:41:23 it wont make sure that autofs is configured correctly on the clients to look at it :) 21:41:29 hmm 21:41:36 I suggest adding this as a potential interaction point with other technology, but not design it fully at this point 21:41:54 dperpeet: Fair point. 21:42:06 jds2001: Would you mind adding these use-cases to the doc? 21:42:17 sure 21:42:49 Much obliged. 21:43:28 As for whether the kerberized NFS will affect the UI... it *might* if we want to offer a choice of secure vs. insecure sharing. 21:44:01 (We can't *only* offer secure sharing unless we want to require that the machines be part of a domain to use this feature... which is a choice we could make, of course) 21:44:32 interesting, but feels a bit too forceful 21:44:50 dperpeet: Which part? 21:45:02 only offering secure sharing 21:45:06 please don't .. unless the domain part is part of the setup.. 21:45:33 langdon: Well, that's where I was going: we could offer sharing if you're a domain member or prompt you to join one first before setting up sharing. 21:45:35 too hard for us non-sysadmins 21:45:51 yeah.. my problem is i don't know how to set up a domain ;) 21:45:53 Joining a domain is quite easy these days. 21:45:59 except on windows 21:46:04 what's a domain? why do i need one just to share files? 21:46:04 langdon: *joining*, not *creating* 21:46:10 * jds2001 puts on inexperienced admin hat 21:46:43 * langdon is/was a developer but played a sysadmin on tv for like 6m .. (by tv, on a contract for some friends) 21:47:11 jds2001: That sounds like a job for... the designer! ;-) 21:47:17 sorry.. i was trying to say.. i don't know how to setup a domain.. so i don't have one to join... or be a member of 21:47:26 let's not limit ourselves like this from the start (regarding domain requirement) 21:47:48 dperpeet: Well, we have a domain requirement for secure sharing no matter what. 21:48:07 So even if we're not *limiting* ourselves to it, we either still have to offer it or else only enable insecure sharing. 21:48:09 right, but we'll want to support quick&dirty sharing as well 21:48:11 I don't like the latter at all 21:48:31 dns domain versus ldap domain versus kerberos domain versus nfs versus samba... goes back to being quiet 21:48:33 So we still have to solve the "why do I need a domain?" question either way. 21:48:43 ok.. don't hit me.. but i can do ssh setup.. so nfs over ssh tunnels? or is that "not recommended" for some reason? 21:48:54 OH GOD NO 21:49:02 PLEASE NO NO NO 21:49:04 ha.. figured that was probably the response 21:49:11 smooge: I tried very hard to convince FreeIPA folks to choose the term "hegemony" instead of "domain", but I was shot down. 21:49:29 well its our cockput ui.. it is a hegemony to me 21:49:35 "set of friendly computers" is probably too wordy 21:49:59 langdon, several reasons.. NFS can use udp.. ssh doesn't do that 21:50:07 langdon: Well, there are two effects to the secure NFS. One is that it's actually encrypted, the other is that the Kerberos layer eliminates the ancient problem about ensuring that UIDs are the same on the clients as the server. 21:50:38 long term we'll want to default to secure unless the user doesn't want that 21:50:56 dperpeet: I don't know that we can call that "long-term". 21:51:18 ahh.. yes... i see.. but there is a plan to make domain setup easy too, right? so is it just a chicken and egg thing? like do "both" for now with a warning about insecure and then when the "easy domain part" happens maybe a prompt to go set one up first and a stronger warning on insecure 21:51:18 I wonder if "NFSv3-compatible" would be a cop-out option... 21:51:26 sgallagh, shortly after we can reliably set up a freeipa server in a container 21:51:39 langdon, exactly 21:51:44 dperpeet: So nine months ago? 21:51:53 as part of our roles 21:51:58 properly configured 21:52:03 does that all work? 21:52:11 haven't checked recently 21:52:23 * langdon notes that he thinks it still asks a lot of hard questions 21:52:26 We don't have a role for it yet 21:52:43 But I've been told that the container version of FreeIPA is fairly stable. 21:52:50 I'm fine with taking this into the NFS role, but it shouldn't be a requirement 21:52:54 So it's probably time to start looking into deploying it. 21:53:12 langdon: define "it"? 21:53:24 "freeipa setup" 21:53:36 sgallagh, is there anyone willing to try and make that happen for our upstream tests? I'm happy to spend some time on this as well - it will get tested many times a day 21:53:43 our = Cockpit 21:53:44 langdon: The ipa-install-server script asks hard questions. 21:53:48 like don't you have to have a custom domain (ie. dns domain) 21:53:50 Most of them don't need answering. 21:54:05 (That's why rolekit can install it with nothing more than "domain name") 21:55:17 I think it's probably worth having another look at getting it deployed via Cockpit. 21:55:52 The real problem with the install script is that they require explicit answers to a ton of stuff that could have reasonable defaults instead. 21:56:13 let's fix that upstream before we try to hack something 21:56:42 and for the NFS role we can have a rough concept of defaulting to a secure setup, if available 21:56:45 dperpeet: Not likely to happen. 21:56:51 actually langdon you will be at USENIX with a lot of people who do ths alot.. is there a way you could ask them to try it and ask the questions to htem? 21:57:08 Upstream prefers detailed choices. 21:57:12 smooge, sure.. we can do that.. 21:57:49 since we want to prioritize the NFS role in Cockpit as a proof of concept, I don't want to flesh it out too much in the initial design 21:58:07 otherwise we get delays on all the other roles 21:58:41 sgallagh: You don't think they'll accept pulling defaults to show the user, where pressing return DTRT? 21:58:41 e.g. we can keep actual domain setup out of this for now, out of scope 21:59:07 dperpeet: +1 21:59:13 dperpeet: That's fair 21:59:32 using one seems good 21:59:36 as that affects the UI 22:01:09 geppetto: Maybe, but I doubt it's on their priority list and I don't have the time to hack it up. 22:02:47 ok.. so .. I would like modularity to be involved in this.. so i would like us to build a nfs-server-module and/or container in support of this.. but i am not sure how to make sure they are coordinated with this work 22:03:43 langdon: I think what we'd like to do is finish fleshing out our requirements doc and get the first pass of a UX mock-up from andreas. 22:03:58 Then we can present it to your team to discuss actual implementation under the hood. 22:04:24 dperpeet: Remind me when your current sprint ends (and when you'll have those mock-ups)? 22:04:27 Was it next Friday? 22:04:41 * jds2001 thought it *started* next Friday? 22:04:46 sgallagh, the sprint will end a week from now 22:04:56 ahhh 22:05:03 sgallagh, ok.. i may have people start on the "defining the module" and "container" .. and then we can lay the role stuff on top.. re-work is ok.. i would just prefer to make progress in parallel (i have other requests for nfs-server anyway) 22:05:52 langdon, you can track what we base the design on in the google doc and the cockpit wiki page / trello 22:05:53 langdon: Sounds good to me. 22:06:00 dperpeet, ack 22:06:22 langdon, or ping me :) 22:06:22 dperpeet: OK, so next Tuesday we expect to have mock-ups, or it's happening the following sprint? 22:07:02 sgallagh, I'm cautiously optimistic about getting at least a first iteration of mock-ups before our next meeting 22:07:08 or in time for that meeting 22:07:43 That would be fantastic. Thank you. 22:07:55 thank andreasn :) 22:08:11 dperpeet, i applaud you on your excellent consultant answer 22:08:51 langdon, thanks :) 22:10:04 OK, I don't think I have anything else to add here, and we're over time. 22:10:06 Anyone else? 22:11:51 Thanks for coming, folks 22:11:53 #endmeeting