Kubernetes Community, 30 Nov 2017

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes Community Meeting 20171130

Description

We have PUBLIC and RECORDED weekly video meetings every Thursday at 10am US Pacific Time.

https://docs.google.com/document/d/1VQDIAB0OqiSjIHI8AWMvSdceWhnz56jNpZrLs6o7NJY

Some links from this session:

- Walid Shaari's thoughts on the Certified Kubernetes exam: https://medium.com/@walidshaari/kubernetes-certified-administrator-cka-43a25ca4c61c
- Kops: https://kubernetes.io/docs/getting-started-guides/kops/
- kubeadm: https://kubernetes.io/docs/setup/independent/create-cluster-kubeadm/
- Contour: https://github.com/heptio/contour
- Kured: https://www.weave.works/blog/announcing-kured-a-kubernetes-reboot-daemon
- OperatorKit: https://github.com/giantswarm/operatorkit
- Kubernetes on ARM: https://github.com/luxas/kubernetes-on-arm

A

Welcome to the kubernetes community meeting for November 30th.

A

We've got some sig updates for this meeting big 1.9 update, but first we should have a demo from code fresh Danny here.

A

Okay, apparently it didn't work.

A

People hear me: we get you. Okay, sorry I just had a sound problem, hello, everybody welcome to the Cooper days community meeting for November 30th. We have our usual sig updates. We have a big 1.9 release update, but first we have a demo from dan from code. Fresh, do Dan. Are you here.

A

Yeah it the was giving him a chance to unmute himself. It was here but muted, but apparently not.

A

Okay, well, in that case, Anthony you're ready to start on the 1.9 release, update.

B

A

B

Alright, listen to reload this doc.

B

Alright, so we approved it a few exceptions that are now going to have a new deadline of this Friday, but we verify that they're all pretty self contained and they seem to be on track so we'll, hopefully, they'll go well.

B

We also want to ask everyone to make sure you have your Doc's PRS in a state where it's ready for the doc team to start reviewing this Friday and you know I know that's a lot to ask with keep coming up, but the docs team also is going to be. You need to do work next week and the week after that for this release.

B

So please keep them in mind and help them by getting things in on time, so ready for review this Friday and then the Doc's need to be ready to merge by next Friday beta 1 was planned for yesterday, but we ended up needing to push it to today, not because anything was broken, but because the tests were taking longer to run than we expected on the branch. So we had a green signal on the master, but then we had to make sure it runs green.

B

Also on the branch stores of just now I checked and everything looks good there, so we should be able to cut beta 1 today and then the next beta will be next week, but I do want to mention that we're not including the upgrade downgrade testing in this when I say it's green. So please, if you see any issues going around for upgrade downgrade failures, please pay attention to those we need to double down on.

B

I think that's it for one night.

A

Okay, Wow well, thank you.

A

A

So next up we have a do. We actually have anybody giving us an update on 1/8 I.

C

Put that in there I mean that's Jo Beth says the the release manager. So he's he's basically said in the channel today that that's the case for 25 next week. Ideally, okay,.

A

That's all yeah, okay, one, seven.

C

D

Won't do it because he.

C

Said he's out of he's in another time zone and one six: thirteen was Anthony, Anthony I think you get the award for the longest-running patch manager of all time. Yeah.

B

It rolled right into one ninety.

A

Glutton for punishment, okay, has the in Garfield joined the call I have I'm here I'm here awesome.

E

Okay, you're ready with your demo. Yes, let me share my screen. Well, so, first of all nice to meet everybody, and my name is Jan Garfield I, remember Kane at code fresh, but don't let that scare, you I'm also a full-stack engineer.

E

So I've got my clip in my my kubernetes t-shirt today to represent properly so what I wanted to show you today is basically how code rush works with kubernetes and what we, what we try to accomplish, and so I'm gonna do that by by showing you some actual, you know pipelines that I have set up. So basically, if you're not familiar code, fresh is a pipeline automation tool. It is a tool set up to if you want to.

E

If you want to get containers into production code, fresh is the tool to help you build tests to pull you those containers, so I'm going to share my screen here and then I'll show you how this works.

E

If I go over to here, we go so here: I've I'm inside of I'm inside of code fresh, and so the question I think. The first thing you want to do is is probably connect kubernetes clusters now I've actually taken the liberty of already connecting some kubernetes clusters. I've got IBM bluemix in here, I've got one from Google, Cloud, I'm gonna, add another cluster right now from Google's cloud, so you can see I've. This is where my different providers are ikat clusters. Here, I can add.

E

You know if I wanted to work with Amazon's new kubernetes cluster I can do that. I can work with Azure. You know container service, I'm gonna, add a cluster from my Google Cloud account and I. Think I have one here called Meetup.

E

Let's add that so there we go, I can see that that's running and I can get a quick look at all the nodes that are currently up and running there. Everything looks okay to me if I jump back over I'm gonna be able to see this cluster and all the services that are running there. Now it looks like I've actually already got a couple of services running there and I can see which images they're running I'm gonna kill these because we're gonna set them up from scratch.

E

So for people that are working with kubernetes, especially as you migrate new users onto kubernetes, one of the biggest challenges is: how do I generate my my configuration files effectively and expect them to work most of the time because in-between docker files and and kubernetes and all the different Yam? Well, you have to deal with it, can just get you can get lost pretty easily. So I'm gonna deploy a demo service here in about about five minutes.

E

Basically, it's an app called demo chat and it has a node ejs and a MongoDB back-end, so first I'm gonna deploy so to get that configured. I'm gonna hit, add service I'm, going to select the cluster I'm working with, and the namespace I'm gonna give it a name. I think the default port for is to 701 7 I'm, going to tell it that I want to pull just whatever the latest is and set that final port.

E

Now, at this point this is this is actually the meet of most kubernetes service and and replica set deployment configurations. I can go ahead and just deploy that and what it will do is it'll actually use a boilerplate configuration to to get that set up and running. Let's filter all this, so that we're only looking at the cluster that we're talking about, and let's also only look at the default namespace there we go now.

E

We don't see all that other stuff going on so here if I look at my I can see that that's up and running and I can see the image that it's running is the latest version of great, so I can also get the status and see that the service that employment, the replica set all those things are working properly now. The next thing I want to do is I want to get a mine, ojs image set up and let's actually build that really quick, I'm gonna go ahead and punch minions. Let me.

C

E

One and let me move move over this one and then let me go ahead and add a new repository, so I'm gonna be working with this one. There we go I'm gonna. Add this repository, because that this platform is container native. That means you can just import your docker file directly or you can start from a template if you need help getting it set up and running. Well, hey create that was my docker file and it will create my pipeline here now you can see I've got the web hook has just turned off right now.

E

I can turn that on easily, if I, if I want to and run it up or requests or whatever my build is set up here, makes sense, I'm gonna, add a test really quick and just run some NPM tests and then I'm also going to I'm just gonna leave it in the code. Fresh registry close fresh, we'll, add everything to a registry automatically anyway.

E

Now this this UI is like a very simple way of interacting with it, but you can actually just switch directly to the yamo and you can see all the stuff that we just configured so as soon as you're, it's sort of like hey, show up here's the rails. You can just walk along the rails and as soon as you're ready it breaks a class and you get access to do whatever you want.

E

So now, I've got this thing. This thing saved, I'm gonna go ahead and build it, and this will go and build that image now code first going to set up a one-off environment for this, and then it's going to pull data layers from anytime I've previously built this lit this image so that it will build faster. This should take about 30 seconds for the sake of time, I'm just going to skip over to where I already have an image created.

E

There's an image: let's grab the image name, and this is what we're going to deploy. So, let's go back over to kubernetes I'll hit, add service. You can see! It's already got my cluster, a namespace set here. I'm going to expose the port I'll, specify the image and then I'm going to use a an image, pulse, secret and I think these are all the registries that I've configured to use a code fresh I'm just going to use the code fresh one, that's built in that should work. Fine I might have a conflict.

E

If I did this in another account earlier, but I think it'll probably work. Let's see alright. So within a few seconds we can see the image that's being pulled as this one, and it should take just a few seconds for these services to actually come up and run in and then did I tell it to expose I thought I did we double check?

E

I did now well, while that's going, I can go back in here and I can switch to the yamo view, and this will give you the service and deployment yeah moles that are actually used. So one of the nine kind of value-adds here for your users is if they haven't configured a kubernetes service and deployment before they basically can just throw their image in this will generate the configuration they can grab. It and modify it as needed, but by default we have kind of some smart defaults.

E

It's set up to doing rolling, update it's it's set up to be to you know, give you your ingress and all that stuff very easily, so this should be up and running, and now, if I open this up, we can see that I've got my application deployed now, if I want to finish, I can actually click on this image here, and this will take me to the image that I can see everything that happened with it, which will include again here, I'm deploying one that I made a few minutes ago, and so you can see this one actually has a JIRA issue associated with it.

E

I can see that there's a twist log security scan that was done. I can see that it passed unit test integration test. I can see all those things. I can add. You know multiple tags and things like that. So basically, what you have here is a platform that helps you build. Your containers set up a pipeline, deploy them to kubernetes, and then you can issue rolling, updates very easily. So I can you know, I can scale this up. I can manipulate it that way or I can just let stuff.

E

Launched and I can set up my webhook and let things let things build that way. Now this whole service and then the entire platform code fresh, we've actually built it not only on top of kubernetes, but we built it also on top of helm charts. So if you're using our cloud service, you'll never really have to worry about it, but if you're deploying code fresh in your own infrastructure, you can just use our helm chart to actually get it deployed into your cluster. And then you have, you know an entirely scalable docker build service.

E

You've got integrations into any registry you're gonna work with integrations into any kubernetes cluster you're gonna work with, and that's that's basically the platform. The one last thing that I'll add is that code fresh actually has the ability to spin up your application on our infrastructure. So what that means is that if you want parallelization of people to self-serve tests and pipelines and and different copies of your application, that are sort of like staging environments- one-off staging environments code fresh- can actually do that as part of the pipeline posted to go slack channel.

E

That kind of thing, so that's the demo and I'm sorry I hope that wasn't I kind of blazed through it, but hopefully that was all pretty clear. Does anybody have any questions? Anything you're curious about.

B

E

See that there's a clamoring for it well I'm gonna, be at coop con next week, so love to see you come and visit I'm doing a session with Google clouds, William Dennis and then, of course, we're gonna have a booth and stuff, but we're basically gonna be talking about best practices for delivering containers at scales. So, if you want to, if you want to deploy, you know thousands of times per day. What are this, what its kind of the checklist? What are the key elements of a really good, continuous delivery pipeline?

E

How you do that and that's what I'll be talking about with William Dennis. So let me know if you have any questions. I'll be hanging around the rest of the meeting.

E

A

Okay, thank you very much thanks for that demo, okay, so we are now back to cig updates and our first cig update is Jonathan for cig multi cluster Jonathan, hello,.

F

You hear me: yes,.

G

F

Okay, so hello, I'm, Jonathan, macmillan, Tina's, my github handle and I've been working on the multi cluster project for a while, so sig multi cluster has been I. Don't I think this has been announced before, but just to make sure for anybody who might not have heard that sig multi cluster is the renamed sig Federation. Earlier in the year there was a lot of meeting and discussing in sync Federation about the what we wanted to do in terms of multi cluster and whether Federation was the only solution to that.

F

The only way we wanted to do that and we decided, as a group, to expand the Charter renamed a signal tea cluster and to start working on other efforts as well so I'm going to discuss some of what we've been up to the past few months. Probably so in the Federation. Work has been ongoing for the past few months and is now complete to move the Federation code base out of the k, / k tree that work is complete as a few weeks ago. So there is a new repository.

F

Get up up tom /, kubernetes / federation now contains the federation code. Now that the code has been moved out of tree if federation is reopened for continued feature development, so anybody who's working on Federation go. Look there to find all the issues have been migrated. The teams have been set up appropriately and the infrastructure set up appropriately to continue working in case lapse.

F

Federation I mean thanks to the people, I think movie or fun led that effort I, think some other people worked on into his names, I'm forgetting, but thanks to them, because that was a quite a big effort to move out.

F

Another important thing that, due to the expanded nature of the cig in the amount of discussion necessary on Federation, the sig has decided to start doing: Federation working group meetings separately from the regular sig multi cluster bi-weekly meetings. Those meetings are currently I believe every Wednesday at 9:30 it maybe every other I'm, not quite sure. We just had the first meeting yesterday I will post a link in this in the community. A community meeting notes to the notes from that meeting.

F

If you're interested I believe the invite goes out to the sig multi cluster who whole group so please join. If you are interested in that meeting, please join that group and please attend the meeting at the kickoff about yesterday.

F

There was a lot of discussion around sort of what to do, how to move Federation forward more towards a stable to a beta and into a stable product, and also kind of addressing and starting to make progress on some of the issues and concerns that have been brought up with Federation in some of the earlier sig multi cluster meetings.

F

Another big thing that the sig multi cluster has been working on lately is the cluster registry. If you're not familiar the cluster registry, is you can think of it in essence, as the part of Federation that, let you add and remove clusters to a list? The cluster registry is basically that functionality extracted out into a separate library, so it lives in its own kubernetes repo case last twelve cluster registry we've been working on it now for a few months.

F

As of a few weeks ago, we now the project plan and the API design for the cluster registry clusters. Api were submitted to the COO were approved by the Sagan submitted to the community. We both we also have that allows us to set up. We have milestones in place for an alpha beta and a stable release of that API and that that pool really looking for people to contribute and to try it out.

F

If you go to K slash cluster registry, there should be some developer documentation, we're working on getting user documentation out, hopefully pretty soon, ideally before to con, so that people have some opportunity to start playing with it does it's announced, and it's talked about their if you're interested.

F

This is intended to be sort of a basic building block of multi cluster solutions, something that you run and then you can build controllers around it to handle all sorts of multi cluster problems, and there is an intention to, at the very least, consider or to move Federation's also targeting this API. So there is no internal clusters API.

F

Another thing update is that makiel Jindal's working on a tool called cube mci, which sets up a multi cluster ingress with a.

F

He's also looking for contributions to that effort. The intention is to make this universal to support, nginx and then other ingress controllers. The first step here was to make it a more congi CP. The eventual goal is to make it work everywhere. Currently it is a command-line tool that sets this up. We're looking at and investigating the goal is to eventually turn this into a controller that targets the cluster registry.

F

I think that's really about it from what we've been working on, and the thing to remember is: if you have any multi cluster questions, concerns problems you want solved or things you think are interesting. Please send a note to our mailing list or join our meeting. The group is very much his focus.

F

Thing is working on Federation, but is also really trying to solve multi cluster problems and the best way possible, and we know about a lot of them, but if you have ones we might not have heard about them, and we definitely want to know what you want to do, how how you want this to work, especially now with the cluster registry being so young. There's a lot of potential for this space for us to really evolve this space so come and join us if you're interested.

G

A

Okay, well, thank you very much and next up cig apps, Kenneth.

D

Hi so I'm Catherine's, the new cig, APSCo, II and I'm, going to give you a little update on what we're doing in sig apps. So who are we?

D

One of our goals is to define, install, manage applications or communities, um talk about and iterate on, higher tools. These are things like meta controller, helm, compose rag and all the things that use the apps layer to make kubernetes easier to use and manage the application lifecycle. We work to make sure the apps layer, which is kind of a workloads API supports most workloads well, and then we work with other SIG's from the nucleus layer.

D

These are the storage networking to make sure that they build the right primitives that allow us to make the actuator work. Well, it's also a forum where people who work primarily on workloads, collaborate with and get feedback from, people who work primarily on higher level, tooling, higher level, orchestration and tools like helm and compose composers of work, and our goal is to be a friendly place for people to come in and start their journey in the case.

D

So a few announcements, as already said, I'm the new co-lead for so gaps along with Matt and Anan Michele's, stepping down, which makes everybody a little bit sad, but on the plus side, she's stepping up the join the steering committee and help us develop the governess. We know the project and we're all grateful for all the hard work she put into helping build the maintained mistakes so round of applause from Michelle.

D

Helms someone is coming now we're not sure exactly when, but it's going to be sometime in February and the details will be forthcoming. We should have more details for you by the time that Rubicon sig apse is taking on mentees and I. Believe Paris gave a presentation to the community about the mentee mentoring program. That's worth coming in to kubernetes, so we're gonna be kind of a testbed for that and we're going to take some people into workloads, someone to helm and I believe charts is going to be another project.

D

That's going to get, hopefully a few interested mentees to get mentorship and rise up to them. I think the rank of reviewer.

D

So, since the last time 1.8 has come out and we did our retrospective- we we looked at like what did we? What went well and what didn't go so long so well was. We did have more crosses in collaboration. We work more storied who worked North API machinery, we worked more networking and that was kind of a good thing. It led us to make sane decisions and kind of accelerated her a release process. We've talked a lot more about charts. We've had a lot more communication from the community about charts.

D

That's worked out well, but we still probably have some issues that were working on with CIO charts. We have some solutions there and getting good maintained, errs and updates for charts. The general consensus is that Michelle and I'm Anna Matt have been doing a great job managing to say and that everyone felt like released communications during the money release process went, went really well, we didn't go so well were release notes, and hopefully we do better with the 1.9 release.

D

That's something: we've been kind of striving to do so helm, releases time to Cooper, Nettie's and causing that and its issues hasn't bitten us. Yet we do hope to get the next home release to build against kubernetes 1/9, and it's been a little bit smoother than has been in previous releases. So far, reviewing a pr's of Cades in the helm has been difficult when we still haven't solved this.

D

Yet we've done some work on charts documentation, but we still probably need a better best practices guide for how charts interact with kubernetes what the best practices are. There's been some stuff with our back that was recently released that which explains how to incorporate our back into your charts. Well, we probably still need to do a little bit more work there, so discoverability for charts documentation.

D

There will be some interesting updates related Kubik on, but we did do some work to improve upon that and the client go is hard, so client go is still hard, but it's a little bit better. There's ongoing work going into 1.10 that'll help make client go a little bit more friendly and we're actually kind of reaching out inside of a sig to try to develop more SDKs than just client go like leveraging the Java SDK, the Python SDK and trying to get contributors there.

D

So 1.9 we got the workloads API to stable, so that's been ongoing for like six months and we're. Finally, there demon said deployment replicas set in staples that are going to go to v1. um Cron job is still in beta, but over the next year we're gonna look at getting that to go to b1 um and we did a lot of work on improving the EE test integration unit tests. Hopefully this bears out metrics and reduce flakiness having collected them yet, but I think it probably has.

D

We still have some work to do on making sure tests find appropriate owners and that we're able to triage and execute on fixing tests when they do break more effectively. So helm is up to 250 contributors. As we said earlier, the our back Doc's have been updated, which was kind of a big deal. People were really kind of struggling around that and two-point-eight supposed to be out by the end of the year, with support for 1.9 compose, had five minor feature releases, volume, detection and conversion improvements.

D

The documentation is improved, a lot yeah, mochi editions, CLI, parameter updates and labels have more control over PCs going forward. We're interested really in participate in the active working group, which you guys haven't heard of it. It's all about how the application definition is defined or kubernetes. So what is an application? What is a package?

D

What are the standards around that? Can we provide a common way for all of the tooling to interface with the system in a sane way, so that it's not everybody's doing their own thing like compose? Does what it does helm does?

D

What it does kou be cuddled does what it does, having a set of standards around that for the system for workloads, we're kind of excited about storage, aware scheduling, which kind of fixes some interesting issues that have always been there in staple set and allows us to leverage local storage using PBS and PVCs, which exciting for some users of the system that have very high performance needs in terms of storage and we deferred a lot of things going to be one saying our focus was going to be stability for workloads.

D

Our focus is still stability but workloads, but we're also going to go back and take a look at some of the things that we said we would defer to later and see which ones we can start doing now and then another biggie is Hellman 3.0. So it's kind of design open, there's a lot of discussion about what helm, 3.0 should look like and what should be in.

D

It will be more details forthcoming at kubaton and, if you're interested the cig, apps deep dive is a good place to come to hear about all this stuff and I. Think that's a wrap.

A

Okay, well, thank you very much shut up date.

A

So, okay catch up in there. Okay, so we have a few miscellaneous things.

A

Justin has an announcement.

H

Good hello, everyone I just want to say that, like we are halfway through a double yeses, reinvent conference and I, think the big news that came out of that, which is pretty much relevant everyone here, is that they announced what they call Amazon, PKS or elastic container service for communities. I had to Google that to forget what that was, but the ejs is sort of like gke for for Amazon. So it is a managed kubernetes thing, and that is one of the many announcements and like we welcome you know it's great to see.

H

Amazon getting on board I was in there or AWS. Sorry are part of the CN CF. Now it's great to see them supporting that more natively on their cloud. I think it's in preview and I think some of them will be it's coupon next week, there's also a whole bunch of other announcements and things that are going into criminals itself to support some of those things like nvme volumes and there's lots of exciting stuff and I think we're still only halfway through reinvent.

H

So, who knows what the rest of the days will bring, but yeah I wanted to like welcome AWS and say that we appreciate all the stuff they're doing and point out like, but it is exciting to see them getting on board and that they have an interesting product that probably some we have of use and will be used by a lot of people here.

H

B

Oh you're, on mute.

A

Sorry, a couple of nebs when items here nobody's name is attached to them, so just notice the notes coupons next week. There will not be a meeting next week for that reason on the coop con Europe, which is going to be in Copenhagen this year. Cfp is currently open in case you'd missed it, and then with that, since we are way ahead of schedule, we have Caillou for a announcement retrospective on the OpenStack summit.

G

So Steve going here, that was my fault, shall we say so George just asked me to give a quick update lately. So this event was two or three weeks ago now: I'm just catching up on a few things, so the most notable things I put in here, the per sec user survey.

G

The latest round of that which gets done every six months, has a question on what paths and container tools are you using to manage stack applications kubernetes again only continues to gain in that category. It's mostly questions so I think 50% of responses so they're using kubernetes, and in that particular question answer in the default selections.

G

They actually break up and shift out separately, which is of course itself of kubernetes, offering as there's another 17% that I like 67% of respondents using kubernetes on the OpenStack clouds, and then the flipside of that is, of course, there's also an increasing number of people using kubernetes to actually deploy OpenStack services as well.

G

There are a large number of kubernetes related talks that were recorded I'm not going to dive in too deeply so recommendations or anything like that kind of themes. Week against source plenty of people sharing lessons learned on running queue by OpenStack, flying up the stack services with you and also more notably breaking out a spoke of no sex services like the bare metal base or the storage and networking business, and just using those with the cube in isolation.

G

And then we also had to call forum sessions in the OpenStack community, so these are basically equivalent as I understand it to the deep dive sessions. We have a coupon next week, whereas more face-to-face discussion and collaboration rather than someone presenting at a room. So one of those discussions about improving the go for cloud library and also the OpenStack communities provider which leverages it heavily suppose. These links I should mention how to ethan parents with notes from those sessions and kind of what was discussed. What conclusions were made and the other one was.

G

We have a cig OpenStack on the eddy site, it's kind of a governance hack, it's mostly the same people, but there's a cqb Nettie's on the OpenStack side. We kind of have to do that to get a space at each conference and so on, but just how we handle that collaboration also making sure that communities are aware of what we're doing as well was the second floor. Incision then.

G

I

There's any questions on that or I'm happier.

G

Talking people on slack about it later, that was the quick update on the OpenStack stuff.

A

Ok well, thank you very much because they're gonna be the fastest community meeting of a rabinin. Does anyone have something else that needs to be announced at this meeting.

I

Yes, I do Terrace yeah, just added something to the agenda: the first contributor office hours contributor being upstream to compliment the office hours that for users that George does went off yesterday at 9:00 a.m. Pacific time.

I

We had a total of four to five questions, so we will pick this up again in January variants that very similar format to George's office hours. So it was cool to kind of hear from people on you know how to contribute, what tools to use etc. So it's helpful for people who you know want to talk to a human instead of redox, which is always an option and that's it. Okay,.

A

Well, thank you very much anything else. Otherwise we can close the meeting ten minutes early, okay. Well, thank you very much see everybody either at coop con or in two weeks. If you're not going to come, come.