fedora_coreos_meeting
LOGS
16:29:11 <dustymabe> #startmeeting fedora_coreos_meeting
16:29:11 <zodbot> Meeting started Wed Jan  9 16:29:11 2019 UTC.
16:29:11 <zodbot> This meeting is logged and archived in a public location.
16:29:11 <zodbot> The chair is dustymabe. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:29:11 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
16:29:11 <zodbot> The meeting name has been set to 'fedora_coreos_meeting'
16:29:18 <dustymabe> #topic roll call
16:29:23 <dustymabe> .hello2
16:29:24 <zodbot> dustymabe: dustymabe 'Dusty Mabe' <dusty@dustymabe.com>
16:29:37 <jbrooks> .fas jasonbrooks
16:29:38 <zodbot> jbrooks: jasonbrooks 'Jason Brooks' <jbrooks@redhat.com>
16:29:47 <rfairley> .hello rfairleyredhat
16:29:48 <zodbot> rfairley: rfairleyredhat 'Robert Fairley' <rfairley@redhat.com>
16:29:57 <ajeddeloh> .hello2
16:29:58 <zodbot> ajeddeloh: ajeddeloh 'Andrew Jeddeloh' <andrew.jeddeloh@redhat.com>
16:30:16 <slowrie> .hello2
16:30:19 <jlebon> .hello2
16:30:19 <zodbot> slowrie: slowrie 'Stephen Lowrie' <slowrie@redhat.com>
16:30:22 <zodbot> jlebon: jlebon 'None' <jonathan@jlebon.com>
16:30:38 <ksinny> .hello sinnykumari
16:30:38 <zodbot> ksinny: sinnykumari 'Sinny Kumari' <ksinny@gmail.com>
16:30:41 <kaeso> .hello lucab
16:30:41 <mnguyen_> .hello mnguyen
16:30:42 <zodbot> kaeso: lucab 'Luca Bruno' <lucab@redhat.com>
16:30:45 <zodbot> mnguyen_: mnguyen 'Michael Nguyen' <mnguyen@redhat.com>
16:31:17 <miabbott> .hello2
16:31:19 <zodbot> miabbott: miabbott 'Micah Abbott' <miabbott@redhat.com>
16:32:20 <dustymabe> #chair slowrie jlebon ajeddeloh jbrooks rfairley mnguyen_ kaeso ksinny miabbott
16:32:20 <zodbot> Current chairs: ajeddeloh dustymabe jbrooks jlebon kaeso ksinny miabbott mnguyen_ rfairley slowrie
16:32:29 <sayan> .hello sayanchowdhury
16:32:30 <dustymabe> welcome :)
16:32:30 <zodbot> sayan: sayanchowdhury 'Sayan Chowdhury' <sayan.chowdhury2012@gmail.com>
16:32:37 <dustymabe> #chair sayan
16:32:37 <zodbot> Current chairs: ajeddeloh dustymabe jbrooks jlebon kaeso ksinny miabbott mnguyen_ rfairley sayan slowrie
16:32:54 <dustymabe> #topic Action items from last meeting
16:33:07 <dustymabe> * kaeso to check/document why CL images are garbage-collected on gcloud
16:33:09 <dustymabe> * jlebon to ask rpm-software-management org on github about name move of
16:33:11 <dustymabe> rpm-ostree repo
16:33:13 <dustymabe> * bgilbert[1] dustymabe to follow up with releng on GC policy in #99
16:33:15 <dustymabe> * bgilbert[1] to file a ticket for live PXE support
16:33:42 <bgilbert> .hello2
16:33:43 <zodbot> bgilbert: bgilbert 'Benjamin Gilbert' <bgilbert@backtick.net>
16:34:16 <bgilbert> #link https://github.com/coreos/fedora-coreos-tracker/issues/99#issuecomment-451035255
16:34:16 <dustymabe> #chair bgilbert
16:34:16 <zodbot> Current chairs: ajeddeloh bgilbert dustymabe jbrooks jlebon kaeso ksinny miabbott mnguyen_ rfairley sayan slowrie
16:34:32 <bgilbert> #link https://github.com/coreos/fedora-coreos-tracker/issues/105
16:34:54 <dustymabe> #info bgilbert filed ticket for live PXE support #105
16:35:37 <kaeso> I did some git history spelunking, but didn't find anything interesting. There was a placeholder for image GC in the first gcloud PR, and later bgilbert just implemented it.
16:35:41 <dustymabe> #info bgilbert added a follow up on the GC discussion in #99
16:36:37 <jlebon> #info jlebon emailed who he thinks is the appropriate person to ask about the rpm-software-management org, waiting to hear back
16:36:38 <bgilbert> kaeso: I think that may have been at crawford's request, but I'm not certain
16:38:01 <dustymab1> hello
16:39:42 * dustymab1 here
16:39:56 <dustymab1> can anyone see my messages ?
16:39:59 <ajeddeloh> yes
16:40:08 <ajeddeloh> why are you dustymab1 ?
16:40:22 <slowrie> probably dc'd
16:40:24 <mnguyen_> yes
16:40:37 * ksinny got disconnected couple of minutes back
16:40:41 <dustymabe> ok
16:40:43 <dustymabe> better now
16:41:22 <dustymabe> where were we?
16:41:32 <dustymabe> did we handle all items from last meeting?
16:41:55 <jlebon> seems like it
16:41:59 <ajeddeloh> yeah
16:42:04 <dustymabe> i think there was one for kaeso and jlebon that maybe got addressed when I was away
16:42:06 <dustymabe> ok good
16:42:08 <dustymabe> moving on
16:42:35 <dustymabe> #topic Produce live PXE images
16:42:43 <dustymabe> #link https://github.com/coreos/fedora-coreos-tracker/issues/105
16:42:53 <dustymabe> welcome geoff-
16:43:05 <dustymabe> bgilbert: i believe you added this one
16:43:09 <bgilbert> I don't have a particular discussion in mind for this one
16:43:23 <bgilbert> just wanted to bring it to people's attention and ask about use cases.
16:43:29 <dustymabe> bgilbert: +1
16:43:50 <dustymabe> i think this has some overlap with our discussion about bare metal installer and what artifacts we should deliver
16:43:52 <bgilbert> the ticket describes Container Linux's design for PXE-to-live support, and we'll want similar capability for FCOS
16:43:56 <bgilbert> dustymabe: yes
16:44:00 <dustymabe> which led to https://github.com/coreos/fedora-coreos-tracker/issues/91
16:44:24 <dustymabe> i'll try to consider #105 as I investigate #91
16:44:37 <bgilbert> as a starting point, I'm assuming FCOS will essentially follow the CL model for PXE-to-live
16:44:37 <jlebon> is this something we'd want from the start?
16:44:43 <dustymabe> and generate discussion when questions come up
16:44:52 <bgilbert> but I'm interested in how people would use it and/or think it should work
16:44:59 <bgilbert> jlebon: yes, for migrating CL users
16:45:28 <dustymabe> jlebon: in the very least we should make sure whatever plan we come up with doesn't prevent this use case from being successful
16:45:47 <bgilbert> I don't _think_ it's a ton of work
16:45:50 <dustymabe> i.e. we make a design decision for something else that makes this not possible
16:46:37 <jlebon> right, that makes sense
16:46:53 <dustymabe> bgilbert: i've actually got some questions for you on this topic but I'll grab you in channel
16:47:00 <lorbus[m]> .hello lorbus
16:47:01 <zodbot> lorbus[m]: lorbus 'Christian Glombek' <cglombek@redhat.com>
16:47:04 <dustymabe> welcome lorbus[m]
16:47:06 <bgilbert> dustymabe: +1
16:47:07 <jlebon> we might want to add something to the roadmap too
16:47:15 <dustymabe> #chair lorbus[m] geoff-
16:47:15 <zodbot> Current chairs: ajeddeloh bgilbert dustymabe geoff- jbrooks jlebon kaeso ksinny lorbus[m] miabbott mnguyen_ rfairley sayan slowrie
16:47:19 <lorbus[m]> sorry all, had some problems connecting
16:47:22 <dustymabe> jlebon: yes, will do
16:47:33 <dustymabe> lorbus[m]: i got kicked like 10 minutes ago
16:47:38 <dustymabe> so it seems you aren't the only one
16:48:03 <dustymabe> anything else for this topic before we move on?
16:49:00 <dustymabe> #topic roadmap
16:49:10 <dustymabe> #link https://github.com/coreos/fedora-coreos-tracker/blob/master/ROADMAP.md
16:49:30 <dustymabe> as mentioned last week we'll review the roadmap weekly and try to make sure we are making progress
16:50:27 <dustymabe> disclaimer: most things in the roadmap are purely an effort to make sure we discuss high priority topics often, dates aren't strict requirements set by anyone
16:50:50 <dustymabe> so for this week we have:
16:51:01 <dustymabe> 2019-01-07
16:51:03 <dustymabe> M - complete bare metal installer: POC #91
16:51:05 <dustymabe> Proof of concept complete
16:51:08 <dustymabe> H - finalize strategy ostree mirroring for better UX #54
16:51:09 <dustymabe> H - investigate no cloud agents #95
16:51:11 <dustymabe> openstack #68, open new tickets for work items
16:51:28 <dustymabe> I'm poking around on #91
16:51:38 <dustymabe> sinny is working on #54
16:51:59 <dustymabe> for #68 it looks like from the discussion in the ticket we are good
16:52:13 <dustymabe> i.e. "for openstack we think we have everything covered and don't need any cloud agents"
16:52:33 <dustymabe> does anyone disagree with that? is there any follow up work needed for openstack that we should document?
16:53:20 <dustymabe> acks/nacks ?
16:53:23 <ajeddeloh> ack
16:53:25 <jlebon> ack
16:53:39 <dustymabe> slowrie: bgilbert ?
16:53:45 <slowrie> ack, I don't think there's anything but I haven't looked into it personallly
16:53:50 <jlebon> already using FCOS in openstack today successfully :)
16:53:55 <dustymabe> +1
16:54:09 <bgilbert> +1
16:54:17 <dustymabe> #action dustymabe to open design doc PR to cloud out openstack cloud agent issue #68
16:54:22 <dustymabe> sigh
16:54:24 <dustymabe> #undo
16:54:24 <zodbot> Removing item from minutes: ACTION by dustymabe at 16:54:17 : dustymabe to open design doc PR to cloud out openstack cloud agent issue #68
16:54:30 <dustymabe> #action dustymabe to open design doc PR to close out openstack cloud agent issue #68
16:54:35 <dustymabe> "cloud out" is a new thing
16:54:53 <dustymabe> :)
16:55:00 <ajeddeloh> We should add Ignition spec 3.0.0 to the roadmap. Maybe have a beta by end of feb?
16:55:13 <ajeddeloh> since we want to ship fcos with Ignition spec 3.0.0
16:55:27 <dustymabe> ajeddeloh: how much actual work is left there ?
16:55:36 <dustymabe> IMHO the sooner we get a beta the better :)
16:56:06 <ajeddeloh> yeah, thre's the implementation bit that will take a bit, we have a pretty good idea what the spec itself will look like at this point
16:56:21 <ajeddeloh> plus the work to add the ign-mount/umount in the initramfs
16:56:36 <jlebon> yeah, i suspect that'll be the biggest part
16:57:01 <bgilbert> #link https://github.com/coreos/ignition/milestone/5
16:57:16 <dustymabe> cool cool
16:57:21 <jlebon> because we'd want to make sure it works with separate /var at the same time
16:57:29 <ajeddeloh> yeahh
16:57:39 <dustymabe> ajeddeloh: I think there is already a ticket for ignition 3.0.0 spec in the fcos tracker ?
16:58:02 <dustymabe> do you want to open a PR to the roadmap to get it updated?
16:58:25 <ajeddeloh> I'll PR the roadmap
16:58:38 <dustymabe> #action ajeddeloh to add igntion 3.0.0 spec to roadmap
16:59:19 <dustymabe> #info ksinny has been investigating the ostree mirroring/repo optimizations. some investigation going on over in https://github.com/ostreedev/ostree/issues/1541#issuecomment-452721757
16:59:48 <dustymabe> ksinny: are we getting promising results? do you have any idea which way we might lean for our implementation for FCOS?
17:00:12 <ksinny> would like to know what other folks think about it
17:00:51 <ksinny> using CDN cloudfront for entire repo/ is definitely a +1 from me
17:00:52 * dustymabe needs to read the remaining comments on that ticket I just linked - will do that after the meeting
17:00:59 <jlebon> ksinny: thanks for the awesome work there. it's great to see this getting attention.
17:01:06 <dustymabe> yes, thanks ksinny
17:01:21 <jlebon> "slow OSTree upgrades" is like... in the top 3 things we constantly hear about
17:01:36 <ksinny> Maybe we can explore a bit around static delta, but that will need some poking around https://github.com/ostreedev/ostree/issues/470
17:01:38 <dustymabe> obviously this isn't the only thing we have to do as there is work to "control update rollouts" we still need to do
17:01:56 <dustymabe> but either way making the sources faster to get will make everything better
17:02:26 <dustymabe> jlebon: yeah. also I just found out yesterday the configs for silverblue installs don't even point at the CDN at all
17:02:29 <dustymabe> *facepalm*
17:02:49 <ajeddeloh> jlebon: on the other hand this will be happening in the background rather than because a user told it to update directly
17:03:08 <jlebon> ouch
17:03:15 <dustymabe> I have a plan to update the repo URLs to not include 'atomic' when we do that we'll make sure the SB configs are right for F30
17:03:43 <dustymabe> #info dusty has a plan to remove 'atomic' from our repo URLs
17:03:46 <dustymabe> #link https://pagure.io/fedora-infrastructure/issue/7487
17:04:06 <dustymabe> ok moving on to next topic
17:04:27 <dustymabe> #topic Roadmap for next week
17:04:39 <dustymabe> so we just covered the items that are being worked on this week
17:04:46 <dustymabe> obviously there is a lot that is going on in parallel
17:05:05 <dustymabe> here is what we have scheduled to look at/try to resolve for next week
17:05:28 <dustymabe> H - collaborate fedora releng integration #44
17:05:30 <dustymabe> H - finalize strategy,collaborate Network Management #24
17:05:32 <dustymabe> gaps identified feature work requested
17:05:34 <dustymabe> H - finalize strategy Kubernetes/OKD strategy #93
17:05:37 <dustymabe> M - finalize strategy Collect metrics from Fedora CoreOS machines design #86
17:05:38 <dustymabe> #link https://github.com/coreos/fedora-coreos-tracker/blob/master/ROADMAP.md
17:06:12 <dustymabe> I'm trying to set up a time to meeting with fedora releng - they had a lot of PTO built up they need to use before February so many of them are still away
17:06:41 <dustymabe> for #24 - kaeso can we make sure we've 'gaps identified feature work requested' for #24 ?
17:06:48 <dustymabe> in the next week?
17:08:23 <kaeso> dustymabe: yes, although I need to re-ping the NM team as they never followed with their workitems
17:08:35 <kaeso> *followed back
17:08:55 <dustymabe> kaeso: yes. persistence is key :(
17:09:06 <dustymabe> please do follow up with them if you would
17:09:19 <dustymabe> I know jbrooks is looking at #93
17:09:40 <dustymabe> and bgilbert - i think the next step on #86 was to make a fedmag post ?
17:10:32 <bgilbert> I believe so
17:10:53 <dustymabe> considering the DNF uuid proposal we could possibly do a joint post
17:11:10 <dustymabe> could you work with mattdm to strategize there ?
17:11:21 <bgilbert> sure
17:11:40 <bgilbert> #action bgilbert to coordinate with mattdm about FCOS metrics
17:11:46 <dustymabe> also is it reasonable to have that item for next week or should I move it
17:12:06 <bgilbert> dustymabe: I don't think it's a major priority, honestly
17:12:21 <bgilbert> dustymabe: it's one of the few tasks we can defer until after the initial release
17:12:36 <bgilbert> we should certainly publicize it and decide on a direction
17:12:42 <dustymabe> ok that's good to know
17:12:45 <bgilbert> but we won't need much code for F30
17:13:07 <dustymabe> yes that sounds great - so publicie and decide on direction - that sounds like a relatively small amount of work
17:13:40 <dustymabe> I'll keep it at next week for now - if no progress in this next week then we'll push to something like end of feb
17:14:29 <dustymabe> jbrooks: did you want to make any comments about #93 before we move to open floor ?
17:15:06 <jbrooks> dustymabe, I've been looking at systemd portable services some -- it's no slam dunk
17:15:19 <jbrooks> I'll make some additional comments on the thread
17:15:24 <dustymabe> jbrooks: thanks
17:15:32 <dustymabe> #topic open floor
17:16:43 * ksinny has one
17:16:51 <dustymabe> go for it
17:17:00 <ksinny> Do we have plan to do Test Day for FCOS release during F30? Sumantro (from Fedora QA) asked me, I think I will better ask here :)
17:17:45 <dustymabe> ksinny: we don't have any plans right now
17:17:55 <dustymabe> i think for F30 we'll sit out the test day for FCOS
17:18:05 <dustymabe> but let's try to do one for fedora cloud
17:18:14 <ksinny> +1, I was also thinking same about FCSO
17:18:17 <ksinny> FCOS*
17:18:21 <dustymabe> #link https://pagure.io/cloud-sig/issue/296
17:18:33 <dustymabe> I asked jdoss about ^^ haven't heard back yet
17:18:49 * jdoss is alive
17:19:05 <ksinny> yeah, in the past we did cloud test with FAH but it will be different this time
17:19:50 <jdoss> .hello2
17:19:51 <zodbot> jdoss: jdoss 'Joe Doss' <joe@solidadmin.com>
17:20:06 * dustymabe waves at jdoss
17:20:20 <dustymabe> ok i have one other topic from a future roadmap item
17:20:48 <jdoss> I will respond to the issue but I am not sure of anything specific to test other than what I normally do.
17:20:48 <dustymabe> we have cloud providers/platforms gce/packet/vmware etc that we need to investigate
17:21:20 <dustymabe> jdoss: right.. but we usually organize a test day and invite people from the community to test - do you mind helping organize that test day with sumantro ?
17:21:39 <jdoss> Sure thing.
17:21:50 <ksinny> jdoss++
17:21:50 <zodbot> ksinny: Karma for jdoss changed to 1 (for the current release cycle):  https://badges.fedoraproject.org/tags/cookie/any
17:22:02 <dustymabe> anyone interested in picking up the torch on gce/packet/vmware to investigate and help us identify issues/gaps so we can close out those cards
17:22:30 <dustymabe> essentially we are trying to check off every check box in https://github.com/coreos/fedora-coreos-tracker/issues/95
17:22:51 * ajeddeloh can gce since I'm already managing that stuff for CL
17:22:52 <dustymabe> and also create new checkboxes for new work we've found
17:22:56 <bgilbert> I can take Packet
17:23:07 <dustymabe> perfect
17:24:02 <bgilbert> oh, fine, and VMware :-)
17:24:06 <dustymabe> the only other item i had for open floor is https://github.com/coreos/fedora-coreos-tracker/issues/104
17:24:32 <dustymabe> basically does anyone know of a good place to put udev rules for our cloud platforms ?
17:24:46 <dustymabe> should we just put them in the base udev package ?
17:24:59 <dustymabe> but that would require us to get them upstream
17:25:29 <ajeddeloh> what does FAH do?
17:25:31 <dustymabe> we could put them in a generic fedora package .. or we could create a coreos-udev-rules package
17:25:36 <bgilbert> note that they'll need to be in the initramfs as well
17:25:39 <ajeddeloh> or regular fedora
17:25:52 <dustymabe> ajeddeloh: so for nvme I don't think we have anything (aws)
17:26:04 <dustymabe> but the cloud-init rpm provides some udev rules for azure it looks like
17:26:21 <dustymabe> but we won't be using cloud-init so we can't put them there
17:26:57 <dustymabe> we could just make a generic 'cloud-(provider|platform)-udev-rules package
17:27:04 <ajeddeloh> wfm
17:27:08 <bgilbert> sgtm
17:27:12 <slowrie> might be worthwhile to make it more generic sounding
17:27:14 <dustymabe> jlebon: walters ^^ any ideas
17:27:17 <slowrie> so we have the option to package other things as well
17:27:29 <bgilbert> other than udev rules, you mean?  +1
17:27:37 <slowrie> yes
17:27:41 <jlebon> or the fedora-coreos-release pkg?
17:28:03 <dustymabe> jlebon: maybe - but i would think we would want these rules in fedora cloud base too ?
17:28:20 <dustymabe> also what is the true source of these files?
17:28:28 <dustymabe> do they ever get updated upstream anywhere?
17:29:22 <kaeso> dustymabe: at the least aws NVMe were written from scratch for CL
17:29:50 <dustymabe> k
17:30:00 <dustymabe> anywho we can pick up this discussion in the ticket
17:30:15 <kaeso> dustymabe: aws-linux has some custom rules embedded in the image, but they use python helpers
17:30:15 <dustymabe> either way we should just pick a solution and not stall too long on it
17:30:29 <dustymabe> will close out meeting in one minute unless anyone else has anything for open floor
17:32:23 <dustymabe> #endmeeting