14:58:46 #startmeeting Server Working Group Weekly Meeting (2014-05-06) 14:58:46 Meeting started Tue May 6 14:58:46 2014 UTC. The chair is sgallagh. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:58:46 Useful Commands: #action #agreed #halp #info #idea #link #topic. 14:58:49 #chair sgallagh mizmo nirik davidstrauss Evolution adamw simo tuanta mitr 14:58:49 Current chairs: Evolution adamw davidstrauss mitr mizmo nirik sgallagh simo tuanta 14:58:55 ahoyhoy 14:58:56 #undo 14:59:01 #chair sgallagh mizmo nirik davidstrauss adamw simo tuanta mitr 14:59:01 Current chairs: Evolution adamw davidstrauss mitr mizmo nirik sgallagh simo tuanta 14:59:09 #topic roll call 14:59:23 Who is around today? 14:59:42 I am 14:59:44 here 14:59:59 * mizmo_afk here 15:00:02 * tuanta is here 15:00:03 simo: here 15:00:13 nirik is on a well-deserved vacation, IIRC 15:00:40 Hello 15:01:12 * masta lurks 15:01:27 Ok, so let's proceed 15:01:32 #topic agenda 15:01:39 #info Agenda Item: New WG Members 15:01:42 #info Agenda Item: Server Role API Discussion 15:01:53 Any other agenda items I missed? 15:02:37 not here 15:02:54 * sgallagh will take that as a "no" 15:02:58 #topic New WG Members 15:03:25 So we need to fill a missing chair on the WG and we have three self-nominated candidates. 15:03:41 adamw: hmm? 15:03:58 #info Dan Mossor: https://lists.fedoraproject.org/pipermail/server/2014-April/001094.html 15:04:05 #info Stef Walter: https://lists.fedoraproject.org/pipermail/server/2014-April/001091.html 15:04:11 #info Thomas Woerner: https://lists.fedoraproject.org/pipermail/server/2014-May/001102.html 15:05:06 Having not faced a situation previously where we had more candidates than seats, we should decide how best to proceed. 15:05:42 sgallagh: hard to decide :) 15:05:50 Our charter requires majority consensus to fill a seat 15:05:55 Presumably, we vote here and forward a choice to FESCo? 15:05:57 these are three very strong candidates 15:06:05 davidstrauss: FESCo isn't involved here 15:06:14 So hard :) 15:06:34 ok as a first observation I can say that stef and twoerner kind of overlap in experience and background 15:06:53 so may be we can simplify the choice by using binary exclusion 15:07:01 ie we put people in this tree 15:07:01 stefw, twoerner, danofsatx-work: I would also like to reiterate that, regardless of a voting seat, you are all encouraged to continue participating in our SIG and meetings 15:07:10 certainly 15:07:15 sgallagh: sure 15:07:20 top - ((Dan), (Stef, Thomas)) 15:07:21 i'd be inclined to pick danofsatx, on the basis he's a solid non-RH candidate: there seems to be a degree of discontent about fedora.next being overly RH-driven. and it'd be nice to have another person representing the target user base of the product rather than another person likely to be involved in building it. 15:07:56 My pick is Stef 15:07:57 we are very RH-employee heavy right now 15:08:02 adamw: that was one of my considerations too 15:08:19 adamw: I was writing something similar; Jim Perrin was originally chosen to represent the needs of downstream. It might be sensible to pick someone else who represents a consumer rather than a producer. 15:08:24 Yeah, pragmatically (cynically?) people who are going to participate in any case don't so much need the mild encouragement of a regularly scheduled meeting 15:08:27 but I was more atracted by the non-pure-engineer aspect 15:08:29 +1 mizmo 15:08:36 mitr: point 15:08:46 because too many of us are already sw engineers involved in server software 15:09:24 tbh I would welcome them all 15:09:28 That said, about everyone has experience with UNIX-style servers, and having a very visible voice from Stef, involved in an explicitly not-old-UNIX GUI, would also be valuable 15:09:42 simo: Agreed. I wouldn't say "no" to any of these candidates. 15:09:45 but the wg seat is only about steering not necessarily doing 15:10:01 hopefully the engineers will have their voice anyway by *doing* 15:10:26 so I think in the end I would agree with adamw for a variety of reason mostly to do with diversity of the group 15:11:21 Given the choice, I think I'd rather have steering weighted towards user concerns rather than engineering ones. 15:12:18 Dan's perspective on user concerns is very different from mine. 15:12:29 He's the "pet" perspective. 15:12:37 davidstrauss, what do you mean by pet perspective 15:12:43 To my "livestock." 15:12:52 ahhhh 15:13:34 the pets&cattle meme 15:14:25 is anybody not comfortable voting on a choice right now? 15:14:38 Tools like Roles are actually more for pet keepers than livestock keepers. 15:14:50 I'd ideally like to seek consensus on a choice here. 15:15:10 So I'd prefer to start with "Does anyone want to recommend *against* a particular candidate?" 15:15:19 So, I'll motion to appoint danofsatx. 15:15:21 Oh 15:15:24 nvm 15:15:34 davidstrauss: We could do that too 15:15:39 i think they are all great candidates, so for me it's more, which way do we want to go 15:15:47 I vote against the "pet&cattle" meme, I really hate that ... 15:15:52 it sounds like we may be leaning towards having a more full complement of user input 15:15:54 As I said, my ideal state is that we have consensus. 15:16:03 * adamw not strongly against anyone 15:16:20 does everybody have a choice in mind at this point or is anybody still struggling to choose 15:16:32 sgallagh: I find it hard to get consesus about picking one in my own brain 15:16:39 I think the views of the other two (not Dan) will get represented regardless of our decision. 15:16:43 hopefully will be easier for others 15:16:44 simo: Fair enough 15:16:55 also evolution left right? why did he leave? lack of time? 15:17:01 mizmo: Lack of time, yes 15:17:07 how is the time availability for the current candidates 15:17:54 I think they nominated themselves 15:17:56 danofsatx-work, stefw, twoerner - do you guys think you'll have sufficient time to commit to this? 15:18:13 eg following the threads on the list and this meeting weekly 15:18:26 in my case i'm working on cockpit, and trying to deliver it for fedora 21, and i've checked with my manager about spending time on the WG 15:18:29 so that's a yes. 15:18:29 Good point. The candidates should be aware that the commitment isn't just the hour a week for the meeting. It may include research and other involvement 15:19:34 * stefw agrees with the points about having sysadmins properly represented in the WG 15:19:37 I will also have time for the server WG 15:19:44 I also spoke with my manager about this 15:20:16 danofsatx-work: Can you confirm as well? 15:20:43 yeah time is fundamental, if there is no commitment I that should be an exclusion cirteria 15:21:07 proposal: if there is no explicit time commitment the candidate will be excluded 15:21:28 well and to be fair we should make clear the time commitment required, which can be sometimes quite high depending on the volume of list threads 15:21:56 I'm not sure we need to enshrine this in policy. 15:22:11 I trust that people who discover they don't have time will do as Jim did and voluntarily vacate the position 15:22:32 danofsatx-work: Still here? 15:22:32 mizmo: well I would say no less than 2 hours a week but be available occasionally for more, depending on task assignment 15:22:47 1 hour is the meeting 15:22:59 1 hour to be up to date with mailing list, meeting notes etc 15:23:13 more as needed but not required necessarily every week 15:24:35 I think we may need to assume that danofsatx-work does not meet the time requirements if he is unable to participate in his own nomination :-/ 15:24:45 * mizmo now you see how tricky mizmo is :) 15:24:47 sgallagh: Agreed 15:25:10 unfortunate, but yeah, it seems to be a concern. 15:25:19 there isn't anybody working on cockpit currently in the wg now right 15:25:31 My second choice is stefw, mostly because I think Cockpit is more germane to our current work than firewalld. 15:25:51 mizmo: Not directly, no 15:26:11 until sgallagh starts contributing to cockpit :P 15:26:23 oh man 15:26:24 stefw: I have three patches contributed! 15:26:25 i'd be OK with that 15:26:34 it choosing Dan would have made things easier 15:26:36 Granted they're all trivial... 15:26:36 sgallagh, i stand corrected :D 15:26:50 now we need to choose between the rock and the hard place :-D 15:27:10 * davidstrauss nominates the rock. 15:27:23 For the sake of translation, which one is the rock? :) 15:27:34 :) 15:28:00 i think twoerner and stefw are both great :( 15:28:08 #info danofsatx-work has been disqualified from nomination due to inability to meet the time requirements. 15:28:57 eenie meenie minie mo 15:29:04 I do as well. Personal bias leads me to recommend stefw slightly over twoerner. This is entirely due to my having worked with stefw in the past. 15:29:49 ok can we have a deathmatch or some gladiator style selection process ? 15:29:49 I hope that both will be deeply involved in any case, but I do want to have the GUI / interactive behavior concerns with an extra strong voice. So I'd prefer Stef. 15:30:07 * adamw votes for a gladiatorial contest 15:30:12 cage fight 15:30:18 Tux Racer? 15:30:20 bzflag deathmath? 15:30:22 I will use a coin 15:30:24 simo: No loosing arms, we hope to receive patches from both 15:30:24 er deathmatch 15:30:27 lol 15:30:29 altho deathmath sounds good too 15:30:35 hehehe 15:30:38 More seriously, I motion for stefw to fill the vacant seat. 15:30:50 davidstrauss: +1 15:30:56 +1 15:31:02 neither is a bad choice, +1 15:31:03 +1 15:31:14 ok I'll chip in, I'll go with Stef as he was proactive and nominated himself even before we officially looked out, he demonstrated extreme attention to the WG 15:31:25 +1 15:31:25 +1 15:32:02 stefw: contgratulations on your new burden :) 15:32:03 I count +7 (all attending members) 15:32:09 * mizmo updates wiki 15:32:13 mitr, oh my. thanks 15:32:26 #agreed stefw (Stef Walter) is selected to serve on the Fedora Server Working Group 15:32:33 stefw: grats 15:32:37 thanks 15:32:40 Welcome aboard, stefw! 15:32:46 #chair stefw 15:32:46 Current chairs: Evolution adamw davidstrauss mitr mizmo nirik sgallagh simo stefw tuanta 15:32:48 hi stefw! 15:32:53 #unchair Evolution 15:32:53 Current chairs: adamw davidstrauss mitr mizmo nirik sgallagh simo stefw tuanta 15:32:58 congrats stefw 15:33:21 Congrats stefw. 15:33:28 #topic Server Role API Discussion 15:33:59 I'd like to emphasize our continued invitation for twoerner to continue participating. 15:34:02 twoerner and I continued discussing this a bit on Friday last week 15:34:14 if another solid contributor does show up to help balance the redhatness of the WG, i'd be happy to swap out 15:34:23 davidstrauss: And danofsatx-work as well, whenever he is able 15:34:29 Yes 15:34:53 #undo 15:34:53 Removing item from minutes: 15:35:11 #info The Server WG thanks all candidates and hopes they will continue to participate 15:35:15 #topic Server Role API Discussion 15:35:35 sorry got called into a meeting here :) 15:35:38 twoerner: So where are we at this point? 15:35:38 er, :( 15:37:00 twoerner: Specifically, what open questions do we have left? 15:37:19 sgallagh: I still have some open questions.. :-) 15:37:28 * stefw notes that twoerner and i discussed a bit about how cockpit might (initially) use the API 15:37:45 #info twoerner and stefw have discussed a bit about how cockpit might (initially) use the API 15:37:57 sgallagh: but at first I think it would be good to collect all things the API and a role shoud/could do 15:38:05 danofsatx-work: sorry, we had too choose, and it was a very hard choice to make, we would welcome any contribution you can keep making though 15:38:17 i understand 15:39:43 So what can and should we solve about this API here? 15:39:48 I started a piratepad for requiresments at https://www.piratepad.ca/p/ServerRoleRequirements 15:39:59 sgallagh: and aditionally a schedule for the parts and the whole thing 15:40:42 cockpit will likely handle installation of a server role in a non-generic manner 15:40:43 (As an aside, IMHO it would be generally beneficial to hold most of the conversation, or at least to record results of personal conversations, on the mailing list; IRC is great for clarifications of misunderstandings but random IRC chats over the week are difficult for people with incompatible schedules to follow) 15:40:47 How generically is Cockpit wrapping DBus APIs? 15:41:02 davidstrauss, very generically, we can (now) talk to any system service 15:41:10 even ones that are pretty ugly like NetworkManager 15:41:22 cockpit will likely handle status of a server role in a generic manner 15:41:46 stefw: So the client connects to the API as a proxy to send, say, native systemd DBus requests? 15:41:47 sgallagh, should i add the sort of status things cockpit would like to inquire about installed roles ... to the piratpad? 15:41:52 sgallagh: Most of those "must be able to indicate" should also include "... and act on it", e.g. the "install a working role" call. 15:41:56 stefw: please do 15:41:59 davidstrauss, yes, via cockpit-agent 15:42:44 mitr: +1 Feel free to edit the pad 15:42:50 #url https://www.piratepad.ca/p/ServerRoleRequirements 15:42:57 davidstrauss, https://raw.githubusercontent.com/cockpit-project/cockpit/master/doc/cockpit-transport.png 15:44:16 * davidstrauss added a proposed API for config file listing 15:45:32 Sidebar: does 'rpm -ql' have a mechanism for listing %config entries only? 15:45:46 sgallagh: -qlc 15:45:54 mitr: Good to know, thanks 15:47:55 Back when I worked on bcfg2, we used that mechanism to audit our way to 100% classification of config files into (1) defaults, (2) desired changes to defaults, and (3) extra, desired files 15:48:03 in general i would say that any configuration and or installation apis should be specific to the role. 15:48:08 For the sake of not scaring twoerner away, can we agree that all NTH items are post-F21 work? 15:48:20 sgallagh: Sure 15:48:32 stefw: for config, do you mean "specific to the role but using that D-Bus service", or "a completely separate role-specific implementation not even necessarily using D-Bus"? 15:48:37 sgallagh: this is the next question on my list.. :-) 15:49:00 mitr, as far as cockpit is concerned it's either one 15:49:17 but i understand that option a) would be easier for both a command line tool *and* cockpit to make use of 15:49:22 at least that's the impression i got 15:49:37 stefw: Fair. From the command-line/automation POV having a single "kickstart-like" format (with role-specific items) would be desirable 15:49:37 stefw: I was just typing exactly that 15:49:43 My additions aren't really server-role centric 15:50:24 But they'd be useful for maintaining installed roles 15:50:31 anyway, my point is merely that the installation and configuration apis for the freeipa role and the database role are very distinct 15:50:53 whereas the ongoing status, backup, monitoring, upgrade are served well by a generic api 15:51:04 stefw: Yes, absolutely 15:51:09 this generic api should be able to handle roles in containers 15:51:18 but that's not a big leap ... just stating the obvious 15:52:03 stefw: +1 15:54:22 * davidstrauss has to head off. 15:54:46 Can we move the openlmi question in the piratepad back here? 15:54:57 o/ 15:55:10 What, specifically, is the concern? 15:55:10 so openlmi providers exit when not in use (after a timeout) 15:55:16 * tuanta got FAmSCo meeting in next five minutes 15:55:26 Yes, unless they have active indications registered 15:55:31 (Same as my comment about D-BUS signals) 15:55:35 nod 15:55:41 ok, that is ok 15:56:02 The master Pegasus process does not yet also terminate, but that's in progress 15:56:07 (Currently under review upstream, IIRC) 15:56:10 I think that exiting when not in use is nice-to-have but not at all a "must" and IMHO not even a priority. 15:56:21 twoerner, storaged has a solid race-free implementation of dbus services exiting when not in use. 15:56:27 the realmd one is a bit more complex. 15:56:46 if you want i'd be happy to document how it works. 15:57:03 mitr, i disagree 15:57:05 these are servers 15:57:06 stefw: sure, that would be good 15:57:13 they're not here to be configured, but are here to do the work tehy've been configured to do. 15:57:19 It's also easier to do this right at the start than change it later. 15:57:25 stefw: what's the impact? A few megabytes in swap space? 15:57:30 "Shut down when not in use" is a fundamental design decision 15:57:34 mitr, adoption 15:57:37 adoption of fedora server 15:58:04 mitr: For reference, when I gave the Summit talk on OpenLMI, the first question we received was "Does this require a daemon running all the time?" 15:58:05 many people are already having a hard enough time seeing all the daemons we're running 15:58:11 Real users don't like that. 15:58:22 OK; I don't think it's that important (people who really care won't be using the generalized roles) but I can see the case. 15:58:23 especially since this daemon essentially does configuration that a tool could do. 15:58:33 obviously the dbus api is a nice structured way of accomplishing it. 15:58:40 guys i gotta go, i have a 12pm meeting 15:58:50 I have to leave in 2 mins also 15:59:07 Ok, we can't really continue without twoerner here, so shall we adjourn for today? 15:59:15 #topic Open Floor 15:59:23 Anything quick for Open Floor? 15:59:39 * stefw is making some effort to make polkit and dbus services usable by non-console users 15:59:40 I'd just like to repeat my plea to have the API discussion more on-list. 15:59:44 I would like to apologize for getting pulled away. It was unavoidable at that point in time. 15:59:56 * danofsatx-work congratulates stefw 16:00:27 mitr: Agreed. 16:00:36 thanks danofsatx-work, hope to see you around, and looking forward to your insight/angle on fedora server 16:00:49 I'll be here ;) 16:00:51 nothing from me 16:00:53 twoerner: Please ask further questions by email and we'll use the piratepad for capturing the results 16:01:32 sgallagh: ok 16:01:34 will do 16:01:46 Thank you all for coming 16:01:48 #endmeeting