Hi gang,
Kubernetes 1.12 was released yesterday, today I've packaged up k8s
1.12 in devel:kubic
Please feel free to test it, especially for testing the following:
- kubeadm init (aka creating a cluster)
- kubeadm upgrade' from a running Kubernetes 1.11 cluster to 1.12
- anything else you think is worthwhile
Note, cri-o and cri-tools don't have 1.12 versions yet so they aren't
in devel:kubic.
Therefore, until they're available and until we've done sufficient
testing, please make sure to NOT forward the kubernetes package from
devel:kubic to Factory
Thanks!
--
To unsubscribe, e-mail: opensuse-kubic+unsubscribe(a)opensuse.org
To contact the owner, e-mail: opensuse-kubic+owner(a)opensuse.org
Hello,
Another week, another minutes.
```
Meeting: Weekly Kubic Project sync meeting
Date: 2018-09-25 15:00:00 UTC
Location: IRC (chat.freenode.net, #kubic channel)
Attendees (alphabetically sorted):
* drpaneas
* ereslibre
* sysrich
Only attendees that interacted are recorded. This is an open meeting, so the list of attendees could potentially be bigger.
Minutes:
* Default CRI runtime, upstream work for 1.13 (sysrich)
2018-09-25 17:00:40 sysrich well I've been missing the last few meetings so to be verbose and recap some of the stuff I missed sharing
2018-09-25 17:00:48 sysrich * I think CRI-O by default has been smooth - no negative feedback (yet?). Some informally reported issues for users sticking with Docker, but no bug reports hindering any solution there.
2018-09-25 17:01:04 sysrich * Politically speaking SUSE received some direct negative feedback from Docker, Inc regarding Kubic's embrace to CRI-O. SUSE's CTO clarified to Docker, Inc that Kubic is an openSUSE community project and SUSE does not control what openSUSE does or says. I believe the discussions are ongoing, as SUSE seem to be keen to use this opportunity to discuss potential improvements in the relationship between SUSE & Docker.
2018-09-25 17:01:17 sysrich * Config structuring & general kubeadm stuff has slowed down the last week or two as I've been doing that with upstream and they're all focused on 1.12 atm.
2018-09-25 17:01:23 sysrich * Autodetection of CRI runtime is likely to be a kubeadm 1.13 feature
2018-09-25 17:01:29 sysrich * Our general mindset regarding how to layer multiple configurations (CLI switches and config files in /etc and /usr) is helping shape the discussions for 1.13 and later
2018-09-25 17:01:50 sysrich * I'm working on adding tests for kubeadm to openQA, as well as some other Kubic-related features like transactional-server in SLE, Leap, and TW.
2018-09-25 17:02:30 sysrich But, I'm unlikely to have upgrade testing available before k8s 1.12 is released, so for the whole team I have a request. Please lets make sure we do NOT submit k8s 1.12 to Factory until we've done some manual testing of upgrading a Kubic 1.11 cluster to 1.12 using kubeadm upgrade
2018-09-25 17:02:59 sysrich devel:kubic will be fine for k8s 1.12, but I want to keep it out of Factory until we at least think upgrades will work (and we can document it ;))
* `etcd-operator` (ereslibre)
2018-09-25 17:03:25 ereslibre from my side: I have been pretty away from kubic development this last week. I am still working on the `etcd-operator` deployment and checking how we can work with it
* `kubeadm` CRI testing (drpaneas)
2018-09-25 17:05:56 drpaneas I am going to test the new kubeadm CRI stuff
2018-09-25 17:06:01 drpaneas that's my goal for next week
```
Cheers,
Rafa.
Hello,
Another week, another minutes.
```
Meeting: Weekly Kubic Project sync meeting
Date: 2018-09-18 15:00:00 UTC
Location: IRC (chat.freenode.net, #kubic channel)
Attendees (alphabetically sorted):
* ereslibre
Only attendees that interacted are recorded. This is an open meeting, so the list of attendees could potentially be bigger.
Minutes:
* Kubic init (ereslibre)
2018-09-18 17:03:19 ereslibre as for my report: we finally made the `etcd-operator` to work and identified an issue with CoreDNS being assigned an IP outside of the servicesCIDR set by kubic-init. This was fixed, and `etcd-operator` can at least start now. I'll continue digging on the `etcd-operator` and having a look at the backup/restore procedure and how we can pivot from the etcd started by the static manifest to the one created by `etcd-operator`
```
Given that these meetings have had less and less people in the three times we have
held it, maybe it makes sense to move this meeting bi-weekly?
Cheers,
Rafa.
```
Meeting: Weekly Kubic Project sync meeting
Date: 2018-09-11 15:00:00 UTC
Location: IRC (chat.freenode.net, #kubic channel)
Attendees (alphabetically sorted):
* ereslibre
* fvogt
* marekcounts
Only attendees that interacted are recorded. This is an open meeting, so the list of attendees could potentially be bigger.
Minutes:
* `kubeadm` (marekcounts)
2018-09-11 17:03:00 marekcounts So I'm pretty new to the project, but this last week I spent some time getting into the upstream kubeadm project
2018-09-11 17:03:25 marekcounts will be updating the PR to the project this evning.
2018-09-11 17:03:49 marekcounts I also looked into the backup and restore of a cluster
* Kubic init (ereslibre)
2018-09-11 17:04:33 ereslibre From my side, we've been advancing with kubic-init. The isues we had with `iptables` vanished the moment we started to use the `hardware` experimental qcow images. It looks like there was an issue with `kernel-default-base` used in the kvm qcow images, others can get more into the details
2018-09-11 17:05:08 ereslibre I've also started to play with the `etcd-operator`, but so far I'm blocked by issues with in-cluster DNS resolution
* kernel-default-base (fvogt)
2018-09-11 17:06:12 fvogt I replaced kernel-default-base with the non-base kernel so that the vm images actually work with crio and docker
2018-09-11 17:06:44 fvogt So kubeadm works well out of the box for me (except a certain upstream bug at kubeadm)
2018-09-11 17:07:07 ereslibre fvogt: do you have the reference of that bug? is it known? maybe we could fix that if it's a bug upstream
2018-09-11 17:07:48 fvogt https://github.com/kubernetes/kubernetes/issues/67431
2018-09-11 17:08:20 fvogt sysrich filed that bug, he just needs to copy it to the correct place to get progress I think
2018-09-11 17:08:43 ereslibre marekcounts: ^ it could be of interest
```
Cheers,
Rafa.
On Fri, 7 Sep 2018 at 15:53, Dominique Leuenberger / DimStar
<dimstar(a)opensuse.org> wrote:
>
> Dear Tumbleweed users and hackers,
>
> This week was not a record when it comes to the number of snapshots,
> but there was a steady flow. There were just a couple days where none
> of the staging projects was in a state to be merged into Tumbleweed –
> which in turn means we also did not produce a new snapshot to throw at
> openQA daily. So, with this in mind, the three snapshots (0831, 0903
> and 0905) released during this week is actually a good number. One
> snapshot (0904) was discarded.
>
> Changes that are in the delivered snapshots:
>
> * dbus 1.12.10
> * ucode-intel 20180807a
> * hexchat2.14.2
> * PHP 7.2.9
> * Virtualbox 5.2.18
>
> There are still a good number of interesting changes upcoming:
>
> * glibc 2.28; the split of libxcrypt causes some build failures
> * OpenSSL 1.1.1
> * Boost 1.68
> * python 3.7
> * ncurses 6.1-20180707: breaks Xen (deprecations added, Xen builds
> with -Werror)
> * Usage of the new System Role selection instead of the legacy
> ‘desktop selection’
>
> Cheers,
> Dominique
>
> PS: for the next two weeks, Max Lin will be taking care of Tumbleweed
For those interested in Kubic the latest snapshot released today is
the first which includes Podman/CRI-O as its default container
runtime.
CRI-O is the increasingly popular replacement the docker engine as the
container runtime for Kubernetes. It was convinently annouced today
that CRI-O is no longer an 'incubator' project but now an official
part of Kubernetes upstream
Podman gives a commandline interface to CRI-O with very similar syntax
to the `docker` cli tool.
I ask people to download the latest snapshot and play about, please
send feedback to the opensuse-kubic(a)opensuse.org mailinglist
Docker remains available in Kubic for those who still wish to use it.
But CRI-O will be the most heavily tested and the first we care about
going forward.
Expect a more detailed blog post on kubic.opensuse.org when I'm back
from vacation
Cheers
Richard
--
To unsubscribe, e-mail: opensuse-kubic+unsubscribe(a)opensuse.org
To contact the owner, e-mail: opensuse-kubic+owner(a)opensuse.org
Hello,
Here you can find the meeting minutes of the Kubic Project weekly sync meeting.
```
Meeting: Weekly Kubic Project sync meeting
Date: 2018-09-04 15:16:10 UTC
Location: IRC (chat.freenode.net, #kubic channel)
Attendees (alphabetically sorted):
* ereslibre
* fos
* fvogt
* sysrich
* tgeek
Only attendees that interacted are recorded. This is an open meeting, so the list of attendees could potentially be bigger.
Minutes:
* Quick round of introduction. Excerpt:
2018-09-04 17:17:40 ereslibre I'm Rafael Fernández López, an engineer working for SUSE on Kubic/CaaS Platform
2018-09-04 17:18:17 tgeek I'm Jason Evans, a training engineer for SUSE on CaaS Plaform and soon CAP
2018-09-04 17:18:58 sysrich I'm Richard Brown, an distribution engineer working for on Kubic for SUSEs Future Technology Team, and Chairman of openSUSE
2018-09-04 17:19:21 fos I'm Ignaz Forster, a research engineer, amongst others working on transactional-updates
2018-09-04 17:19:30 fvogt I'm Fabian Vogt, release engineer at SUSE and have to leave in 10 minutes...
2018-09-04 17:19:37 fos (also working for SUSE)
* Kubic init (ereslibre)
2018-09-04 17:19:39 ereslibre As for the sync meeting: we are currently working on the Init container (https://trello.com/c/3Ad1BgYP/25-init-container). Right now, this project lives at https://github.com/kubic-project/kubic-init. It calls `kubeadm` in order to start a "seed" node that will host itself and will deploy addons on top of it. Right now, we are facing some issues with flannel and iptables
2018-09-04 17:20:21 sysrich do you think your flannel issues might be fixed if you used containers built against a recent Tumbleweed/Kubic?
2018-09-04 17:20:53 ereslibre sysrich: that can help; but the issues we are experiencing are even trying to list the iptables rules on the seed host, so I fear we have deeper issues
2018-09-04 17:21:39 sysrich oh fun :)
2018-09-04 17:22:16 ereslibre sysrich: yeah, we are still investigating, and Michal Rosteki (another SUSE engineer, network team) will have a look at the issues we are having with iptables
* Tumbleweed, CRI-O as default, merge docker-kubic into the main docker package (sysrich)
2018-09-04 17:22:57 sysrich for my sync: Tumbleweed has been suffering from a bit of a backlog after a major openQA outage a few weeks ago. Things are just about now returning to normal, and the backlog of items includes the changes to set CRI-O as default and merge docker-kubic into the main docker package
2018-09-04 17:23:32 sysrich the main problem left is figuring out how to fit the supplemental packages to configure kubeadm for CRI-O and docker into the same FTP tree or install media, despite the fact they conflict with each other.
2018-09-04 17:24:01 sysrich All of our sanity checks are objecting to both packages being together, despite the fact that in reality you're only going to every pull through one of them via the supplements dependency
2018-09-04 17:24:22 fvogt Which step blocks it?
2018-09-04 17:24:42 sysrich repo-checker installing kubic-onlyDVD pattern; so I'm probably going to split the packages into their own patterns tomorrow, which will obscure them from repo-checker
2018-09-04 17:25:02 fvogt Right, that's expected
2018-09-04 17:25:14 sysrich expected, maybe, annoying, also maybe ;)
2018-09-04 17:25:17 fvogt If the patterns conflict each other they can be added to the product
* Vancouver OS Summit (sysrich)
2018-09-04 17:26:45 sysrich besides that, went to OS Summit in Vancouver, lots of people were interested in Kubic, especially the Transactional Update and upstream-first sides of things, which seemed to be the thing that set us apart from all the other k8s talk.
* Transactional update (fos)
2018-09-04 17:29:20 fos Just some prospect: My plan is to have the zypper integration ready for SLE15SP1 (and with that for CaaSP4), though I'm not sure if I'll make it in time.
2018-09-04 17:29:36 fos I'm currently a bit distracted by other things, but I'll try :-)
2018-09-04 17:29:53 sysrich fos, They haven't finished writing the PRD for SLE 15 SP1, never mind interlocking it. They hope to have that wrapped up by October..so you have time ;)
```
I preferred to be more textual about the contents of the meeting. If you
believe this format is too verbose, please raise the subject and we can
adapt future minutes with a shorter version.
Cheers,
Rafa.