10:49:42 <aday> #startmeeting Workstation WG (2020-01-14)
10:49:42 <zodbot> Meeting started Wed Jan 15 10:49:42 2020 UTC.
10:49:42 <zodbot> This meeting is logged and archived in a public location.
10:49:42 <zodbot> The chair is aday. Information about MeetBot at http://wiki.debian.org/MeetBot.
10:49:42 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
10:49:42 <zodbot> The meeting name has been set to 'workstation_wg_(2020-01-14)'
10:49:42 <aday> #meetingname workstation
10:49:42 <aday> #chair cmurf aday
10:49:42 <aday> #info These minutes could be subject to subsequent revision.
10:49:42 <aday> #topic rollcall
10:49:42 <zodbot> The meeting name has been set to 'workstation'
10:49:42 <zodbot> Current chairs: aday cmurf
10:49:43 <aday> #info present: cmurf, tpopela, aday, benzea, ngompa, mclasen, langdon, jens, otaylor, kalev
10:49:45 <aday> #info regrets: mcatanzaro
10:49:47 <aday> #info missing: no one!
10:49:49 <aday> #topic Approve previous meeting minutes
10:49:51 <aday> #info https://lists.fedoraproject.org/archives/list/desktop@lists.fedoraproject.org/thread/NPHPMQCAMAQY7MUR4NMRN2MO7B3DX3CZ/
10:49:54 <aday> #agreed minutes approved
10:49:56 <aday> #topic Announcements
10:49:58 <aday> #info Chris has decided to use Pagure and Taiga together. The latter will be used for tracking work items, mostly by Chris and Allan. Discussion will mostly happen on Pagure.
10:50:01 <aday> Allan: WG members will need to use Taiga if they have tasks assigned to them. Not all WG members are present in the project at the moment; Langdon has offered to help with this.
10:50:04 <aday> #info DevConf is happening 24-26 January in Brno.
10:50:06 <aday> Langdon won't be available on 21st and 28th. Tomas will be busy around DevConf too.
10:50:08 <aday> #agreed We will go ahead with the meetings on the 21st and 28th, unless we get too many regrets.
10:50:10 <aday> #topic Workstation WG vacancy - https://pagure.io/fedora-workstation/issue/122
10:50:14 <aday> Vote: add tomas to the working group
10:50:16 <aday> +1 Allan, Chris, Jens, Kalev, Langdon, Matthias, Michael (voted ahead of time in the ticket), Neil, Owen - passed unanimously
10:50:19 <aday> #agreed Add Tomas as a member of the working group
10:50:21 <aday> #action Allan to update the wiki
10:50:23 <aday> #topic earlyoom by default in F32 - https://pagure.io/fedora-workstation/issue/119
10:50:25 <aday> Chris: concern that if we add earlyoom it will stick around for those who upgrade in the future. Neal: the obsoletes package can be used to remove packages. Tomas: this could be an issue if someone has manually added/configured earlyoom themselves. Neal: yes, but if we go with a different oom solution, someone really shouldn't have more than one running.
10:50:30 <aday> Allan: note Bastien's recent comments on the ticket - fairly strongly against earlyoom. Owen: doesn't have a good sense of Bastien's approach to some of the problem cases in the ticket. Matthias is also unsure about this.
10:50:34 <aday> Neal: moving to systemd assumes that all the processes are going to be properly scoped. At the moment this doesn't happen: examples of terminal processes, Electron apps. systemd is not going to be a complete solution. Still thinks that earlyoom has a role - it needs to be something that works universally. It has consistent behaviour. Enabling earlyoom could bring us to the same level as Windows and MacOS - give applications a chance to
10:50:39 <aday> gracefully exit by sending SIGTERM before SIGKILL.
10:50:43 <aday> Owen: would like to hear more from Bastien what he thinks the long-term plan could look like.
10:50:45 <aday> Matthias: any solution that avoids catastrophic failure is worthwhile. (General agreement to this.)
10:50:48 <aday> Ben: The systemd move has no effect beyond lmm triggering the kernel OOM killer. However, that is just changing the OOMPolicy setting which is trivial to do. (systemd scopes are only applicable to oomd currently as it always kills a whole cgroup).
10:50:52 <aday> Chris: does the WG prefer to defer the question about earlyoom to FESCO?
10:50:54 <aday> Owen: -1 to that. It's a workstation-specific question.
10:50:56 <aday> Neal: the only reason to do that would be that it's a system-wide change, but it's only affecting the workstation. -1.
10:50:59 <aday> Matthias: -1.
10:51:01 <aday> Langdon: -1 ?
10:51:03 <aday> Kalev: no opinion
10:51:05 <aday> Jens: thinks we should have an opinion. -1
10:51:07 <aday> Allan: abstains - doesn't know the background to the earlyoom proposal, or why it is workstation-specific. Chris: earlyoom only being proposed for workstation because we haven't been confident about the other use cases, and the proposal hasn't attracted interest from the other working groups.
10:51:13 <aday> #agreed The workstation WG should handle this question itself, rather than defering to FESCO.
10:51:15 <aday> Chris: do we want to come to a decision today? Do we want to invite other participants/advice first? Could invite Bastien, Alexy, someone from oomd? Could ask Lennart.
10:51:18 <aday> Chris: the "universal" Facebook oomd is going to land earliest F34. Neal: then maybe F35 to get systemd-oomd. Shipping earlyoom will help by putting pressure on this effort. Owen isn't so sure about this. Matthias: if earlyoom solves the catastrophic failure for a few people, it is still worth it.
10:51:22 <aday> Chris: I am yet to see earlyoom sigkill something in my testing - always sees sigterm.
10:51:24 <aday> Owen: my overriding concern is that we understand what our long-term plan is. Would adopting earlyoom make it more difficult to transition to oomd later? Thinks we should know about this.
10:51:27 <aday> Neal: is in favour of earlyoom, but would still like some guests to inform how we deploy earlyoom.
10:51:30 <aday> #agreed Let's invite some domain experts, to build some consensus of the general direction between the relevant players, and make sure that our decision is an informed one.
10:51:33 <aday> Kalev: doesn't think that oom is in the domain of the workstation WG. Doesn't think that we're qualified and would prefer someone else to make the decision.
10:51:36 <aday> Owen: who else is considering the workstation use cases if we don't? Things like not killing gnome-shell. We have to have an opinion.
10:51:39 <aday> Kalev: in that case, let's invite the experts. Let's not decide this in a vacuum.
10:51:43 <aday> Vote: the WG supports earlyoom by default for Fedora 32
10:51:45 <aday> Allan: -1 - we should invite experts before making a decision
10:51:47 <aday> Jens: +1 - we should test it in rawhide
10:51:49 <aday> Kalev: no opinion
10:51:51 <aday> Langdon: abstains - either answer is good
10:51:53 <aday> Matthias: +1
10:51:55 <aday> Neal: +1
10:51:57 <aday> Tomas: abstains - wants more expert information
10:51:59 <aday> Owen: +1 - we can't delay indefinitely, there's no reasonable alternative for F32. Would still like us to get expert advice.
10:52:02 <aday> #agreed We will move forward with earlyoom by default for F32. This needs testing and the design needs more work. We will consult with the relevant experts in doing this.
10:52:05 <aday> #action Chris to update the FESCO ticket with the earlyoom decision.
10:52:07 <aday> #action Chris and Allan to invite domain experts to a future meeting.
10:52:09 <aday> #topic Video vs IRC for WG meeting - https://pagure.io/fedora-workstation/issue/113
10:52:13 <aday> Chris: what do we want to do?
10:52:15 <aday> Neal: likes the video meetings. We seem to be making decisions better now.
10:52:17 <aday> Owen: has only done one video meeting. IRC was easier to jump on, but video does seem to be more focused.
10:52:20 <aday> Matthias: in favour of video.
10:52:22 <aday> Langdon: finds video harder - prefers writing and being able to multitask. However, doesn't have a strong opinion. We could try a mix of both types.
10:52:25 <aday> Jens: likes typing, but thinks the video meetings are working well.
10:52:27 <aday> Kalev: would prefer IRC. Thinks that they are more inclusive. Fedora leadership are able to jump in.
10:52:30 <aday> Tomas: finds IRC meetings difficult to follow (people replying over each other,  hard to guess the structure sometimes). More in favour of video meetings.
10:52:33 <aday> #info Chris: we will do video again next week and take the temperature again in the future. Defer to guest speakers' preferences, when applicable.
10:52:36 <aday> #endmeeting