Kubernetes Community, 13 Sep 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes Community Meeting 20180913

Description

The Kubernetes community meeting is intended to provide a holistic overview of community activities, critical release information, and governance updates. It also provides a forum for discussion of project-level concerns that might need a wider audience than a single special interest group (SIG).

https://contributor.kubernetes.io/events/community-meeting/

A

All right, good morning, good afternoon, good evening, wherever you are, this, is kubernetes community meeting and will be posted publicly on youtube. So please be mindful of that. You will say or fear or eat or laugh or be recorded and will be posted. My name is Aaron Gupta I work for Amazon I work in the open source team here at Amazon and I show up in cig. Aws often you know doing some fun discussions over there.

A

So please for everyone, please be mute when you are not speaking, and today we are gonna start with so well. First of all, I see that slowly rose from cig auto-scaling is the note-taker. Is it right? Yeah awesome? Thank you. Sorry.

A

So we're gonna start with a demo today, and the demo is answering questions on kubernetes slack with Foucault's flood. You ready.

B

Yep one sec, you sure screen so I'm, sorry to everybody, that's been on the contributes channel. That already knows all this stuff, but here you go again so so my name is what, and today I'm gonna talk to you about the vocal slack pot. It's a slack pod. That's installed on a few of the channels on the kubernetes funk instance. So the goals of this thing are to improve user experience and focus contributor time. And what does that actually mean is a lot of times when a user joins the kubernetes slack channels, they come in.

B

Ask a question and whether it's because they're asking it to 2:00 a.m. on a Saturday or because the question is too big or whatever a lot of times the questions. Just don't get answered, and sometimes they'll post again still don't get an answer and at the same time, contributors a lot of times spend time answering kind of repeat questions in a in a study. I found that 30% of the time people feel like they're spending on just repeat questions and 20% of the time. They're.

B

Spending on answering questions that are already well documented that they don't need to answer. So let me show you how we're solving those with a quick demo. So imagine.

C

B

And this is within my kind of test workspace, but imagine I'm, a user and I come in and I ask like hey how do I run a conformance test, so what happens is when I hit enter without me? Having to do anything else, Fogle will use AI to determine that I'm asking a question find the most useful answer and it will send an end and they'll send it back to me directly.

B

So what you'll notice here is one I didn't in any way invoke focal I, don't even have to know that fact, whenever it's simple messages directly to me within the channel, so I'm not overpopulating the channel, I'm not spamming the channel with many messages so we'll see. Is that there's a couple of different answers that are, you know potentially helpful to me and what I could do is I can rate these channeling. These answers, whether they're, actually helpful or not.

B

So if I click not helpful, message simply disappears if I click, helpful and I still have a question, then we'll record that feedback. And finally, if you click helpful and answer question, what we'll do is we'll post the message to the rest of the channel notifying the channel like hey? This is already answered. You don't have to worry about answering together.

B

So basically, this is a very short kind of description of like how a user quickly interacts with local quickly gets their answers. Their questions answered. So let's take a look at where these questions are even coming from. So a bunch of questions. Ninety five hundred ninety plus ninety six hundred questions were actually downloaded from the kubernetes tagged questions within Stack Overflow. The next section 17 closed 1,800 questions were downloaded from kubernetes Docs.

B

Now what that means is we we, when we are showing somebody at one of the sections of Doc's, we don't want to send them to the entire talk right. We want to only show them the most relevant section. So we did was one scraping, the kubernetes Doc's. We actually divided it into smaller sections, so that one since one we're showing an answer from a doc. We only show a very small relevant section, and this the last part of where these things come from is previous lot conversations.

B

Let me show you how that works so imagine, and for this we'll just do the slide demo instead of doing a live demo. So imagine that Michael and Kay are talking so Michael is asking a question about how to create a kubernetes. Pod kay responds with some information. Michaels happy right.

B

So at that point, folk will not only detects that Michael is asking a question but are also detects that K is helping Michael with with a relevant instrument, so at that time focal will send a message again only directly visible decay and ask her to hey. This is very useful. Can we store this for the future and show it to others? At that point she can ignore the question she can edit the question and answer to provide more context, make it more useful.

B

You know improve it, whatever waste and finally click store, and when she does that this question answer pair becomes visible and searchable and is available.

D

B

Others in the future, with similar questions, automatically show that so, let's take a look at kind of what the results of this so far are we've been running a little bit over three months on mainly two active channels until recently, in the last two weeks, 37 of these responses have been automatically marked as something that's very helpful to somebody and a total of 189 times since June has has we been marked as something that's helpful.

B

If we look at the breakdown of where these helpful questions come from, Stack Overflow those questions, even though they actually have the most of those questions, they are only been helpful 10% of the time in kubernetes Doc's, which are actually were added fairly recently, but the 40% of the useful answers are coming from kubernetes Docs and close to 50% comes from these previous flat conversations, so we can see that the most of our value is actually coming from our community itself and, more specifically, we could take a look at like who, on the who's, providing the most useful answers.

B

If this is interesting, so in this case, Huber Chen added 10 questions, 8 of which were marked useful, at least once and out of these eight questions. They were marked useful at 12 of 16 times, and you can kind of get the idea. This guy Bob on the channel McAlary.

D

B

Eleven questions for which we're useful and those four were useful nine times so so, what's kind of interesting is what is the most helpful answer? There's actually a tie between the most helpful answers, but I thought this one was a little bit funnier. Is that somebody's basically having a problem with deleting something within a pond? And obviously the answer is just to you know, insert a dash dash into the command so we're? Actually.

B

What have we been doing recently is talking to the kind of the documentation community to the docks people to see how we can work together and then use the kind of responses from focal and the feedback loop to improve the docks and and kind of like improve the search ability of the dogs, maybe provide but more examples and just kind of see what we can do in the future. So the other thing that's interesting is people not only like love it, but they actually even expect it to be there.

B

So one person was asking a question and this users skarbek responded the focal bomb probably already spotted to you, but here's some additional documentation or here's. Some additional context- or in this case edward was asking you know- was asking question and finally, at the end of this question kind of says: well, hopefully, focal can help with this. The the sad part is that the channel that edward asks, this question on was not was was not already signed up for focal, so we were not able to help him because it wasn't installed on his channel.

B

So with that in mind, the short call to action here is invite focal and your kubernetes channel. If you're in the kubernetes slack workspace just do slash, invite focal you could do slash focal and type in something and I'll show you there's a bunch of different ways to add a bot to a channel talk to me if you don't figure that one out I'll help you and we're asking you to invite it to multiple places, so first invited to your user facing channels.

B

When you add it's your user equation channels, we can do this whole, like feedback loop, providing useful answers to people and, at the same time, we're saying you can add it to your like c channels and you can store information such as you know or like whatever the most common questions there are such as you know, when is the meeting time and where is the YouTube channel and then so on, so to add the questions you'll see. Obviously this will pop up and all we're asking is maybe improve the context.

B

A little bit and click store so make sure to add your useful context and useful answers to focal. You can also manually add questions if you like, if you hover over any message in slack you'll, see the like the ellipsis button. Click on that and you'll see add to focal and you'll see a dialog about like adding what is the question-and-answer too bad.

B

So talk to me about integrations, so what we can do is because it like, if part of your project is in some other different workspace talk to me and we can figure out how to import that information and kind of like scrape and pull different sources and put that into vocals, so that we can make that available to your users.

B

Talk to me about partitioning, so one thing that we realize is that, like you know what time you're meeting for signature backs or something is probably not relevant to the kubernetes user channel, so we can do. Is we can create logical separations of information so and kind of finally control how data flows between these, so that a question might only appear within stick and Trebek's, or an answer might only appear within signature Bex and a different answer might only appear within kubernetes.

D

B

Can control how this data flows and finally install focal everywhere else? If you, you know, if we already in kubernetes slack, you can just install it's already installed. So just add it to your channel. If you want to install it to a different kind of open community, let me know the best way to contact me is either through email or you can find me on the kubernetes slack as just under my name.

B

So if you want open it, if you want to install it on any open channel or you can go to focal io, /, OSS or obviously you can install it within any company what we do there's, instead of pooling from things like Stack Overflow, and you know your Doc's, what we can do is we can pull from wiki and JIRA and all these other things, so this actually designed to mainly work within company as well. So with that I know I kind of ran through this quick. What questions do you have.

A

We have time for one question: you know you're already over time. So maybe we'll take one question: if there is any.

A

Okay, all right I guess none thank thank you. Thank You Vlad for the wonderful demo and I always show it always blows my mind when I see how these integrations come along and these organic integrations are very, very helpful.

A

All right cool. Thank you. So much. If you have any questions, please reach out to blog directly and invite Foucault's to your slack channels. The next item on the agenda is release updates. Tim pepper. Are you ready for 1.12, elisa updates, so.

D

I'm tim, pepper, you're 112 release lead a couple of updates right now. We are still in code freeze. We had our beta 2 release this week and the release candidate will be coming out next week, we're working towards a release target of September 25th. But at this point that's starting to trend at risk. We've had quite a bit of a CI signal breakage and we're making progress there, but it's been slow.

D

We've done a lot of good debugging in the last, even just 14 hours, probably, but because some of our tests take a bit of time to run and we basically got seven workdays to the release target. At this point and once you subtract a day or two or three even for CI signal to prove that something was actually resolved, we're running short of workdays to actually accomplish improvements, so we'll make a call on Monday, depending on how things pan out, based on yesterday's debug today's and tomorrow's merges and the weekends test results.

D

But by Monday we may make a choice to slip. The release a couple of days, I'm still relatively hopeful compared to two maybe last week. I would say where we're trending in the greener direction versus the redder direction, but still squarely in the the middle yellow state of risk, so stay tuned. We'll have some updates next week on that front. Also I mentioned that we moved KK to tide on and you may or may not have noticed that we mentioned it last week in the community. Meeting.

D

Here is something that was coming but slight hiccup or two at the beginning of the week, and it seems, like things, are humming along at this point, but just an FYI there and that's my update for the week. Thank you all.

A

Right awesome, any questions for Tim.

A

Okay, I guess man, any other minor release, updates that anybody want to provide.

A

Okay, then, we'll keep moving on that, so no more release updates for this call. There is no cap of the week this week, I was talking to Caleb and couldn't find one. Is anybody want to present a grass of the week this week.

A

Okay, that checksum can be blown away. Then, are there any open caps that we can talk about.

A

Okay, I try to reach out to people, but nobody was available to do this. That section goes away as well. We do have two SIG's that can provide an update, so the first one is say: Windows Michael. Are you on the call? Yes.

E

I'm right here, thank you guys, all right awesome, take it away so essentially on safe windows, who've been humming along. We got to the point where we have finished a lot of our functionality in terms of feature said that we wanted to provide to be able to graduate our windows support to stable for kubernetes. Unfortunately, we've hit some hiccups in stress and performance and scale testing, and we are proposing that we are going to graduate to stable, but with 1.13 and not with 1.12.

E

So we are right now finalizing a draft of our feature freeze for us, so we can actually stop working on features and start concentrating on stability and conformance and and start also finalizing. Our API Docs general documentation, how-to guides and everything else as necessary to graduate to stable so we're putting the stop on new features and concentrating almost 100% on step and getting us ready for for a step for GA with 1.13.

E

A

All right very cool, any questions for Michael.

A

Okay, if none, then we can move on the next update that we have is from signaled and Dawn Jen. Are you on the call? Yes.

C

A

Thank you done and I saw somebody dropped a link to the slide just now. Yes,.

C

Yes, nummy-nummy first, a poster size here: okay,.

F

C

Thanks awesome.

A

Take it away when you're ready, yes,.

C

I'm ready can someone help me sorry, I do not mean writing and because I want to help me to open, though I shall increase in suicides. I can.

A

Do that let's see if I can get access to this, then I can.

A

Okay, let me share my screen here so that others can see.

C

Alright, this is mass update from a signal. It is last a piece of kata ago. So this is a I'm going to update would happen in a queue. Sorry, sorry, can you go to the next one? Please Thanks, so we finally have the finished initial signal: chatter and the PI's merged, and also we have not the we goes through the new room meeting as expected and a defined by steering committee, and we, how do we still hold the meeting with me every Tuesday, 10:00 a.m. Pacific time and the English slides I.

C

Have that our notes, and also the link to the YouTube videos in additional and actually we also hold the sponsor and a hold and our bi-weekly resource management work and every Wednesday, 11 a.m. and they're also implement. We also have I think put in a size. Actually, we also have the asian-pacific signal meeting, because there's also a lot how work in well with the different winters. So especially on the one, the earlier stage on the company rent time implementation and the CI implementation. Can we go to the next slide, Thanks so doing in the q3?

C

We reviewed our original signal, the scope and update. What too, is under this tape? So we categorize a file area like the work to week we talked before it is to include of the know the lifecycle management, application management on the node and they'll include of the content of runtime. All those kind of things and also resource management include of the device, management and also computer resource management and access EQ memory, and also includes storage management collaborate with collaborate with the stick storage.

C

We also enters the also the secrete here and ISO nation on the node, and then there it is the instrumentation login on an older. Can we go to next one so now I'm going to quickly update what we delivered with the any program in q3. So the first one I want mention that we made another called hast ID progress, understand octopod, so there's the runtime counts, propos'd and which it is being accepted by the SIP architecture and crude I read a signal and in the q3 we deliver.

C

This feature rental cars, API feature as damaja feature and also implement of the rent hannah arendt on class API therapy and and there's the treaty to support on the north side to support the rental car and also there is the we have work, we're close with the cattle community and also contain your community to different to integration with the caja and also send fax solution, especially we email, heavily ways the company community and to help them define continuity, Shin API version to you so which is better integrate with other than the Box solutions.

C

Can we go to next slide? Thank you so like earlier Sega, Windows Update and we also work were close in this quarter. We also we're close with the Sega windows. We help those working together base than those sick of induce the community and define windows so for the ga o-- and the document in here, and we also loop or instead of only because the support of the windows container, it is a 2-pack. Unity is actually it is a support or new kinda form windows plan.

C

For so we working together and expand their the working community here include off the six storage and the sick at work and in the q3. What deliver on the node most here it is on a big note. We added the Cuba stickers for windows, the system covenants as like as the Cuban aid and the container and time so. Basically, on the note we already have all the node and a system really need a specs for windows plan for and also there's them.

C

Several simple regression fix on the windows container on the node and beyond that, there's many proposal made by the Siegelman those folks and metal foot box, though there's the DNA's kappa-b, need your support for Windows EMI and also there's the proposal for Windows the see my support.

C

We need I'm least here, because I want to call out of the stick Network attention. Please have asked you to review those PRS and a standard speed-up of the velocity on the windows is support and there's also cover off the pr proposed to the artist invoke ripple for automated windows testing framework, and I also called for the sega past- have I'm here to review those PS and we go to links then in the trustee, or we also spend a lot ham to reverse our testing in passing work.

C

There's the announcement for the changing the node III test and, and it's been reviewed by the architecture, and then we introduce the new set of the type to have out the node ringing, the features and the node the related of the component tests. We also build in our community dashboard. We have the specific testing dashboard so so, either container random can easy to plug into that dashboard and also user or vendor the kubernetes.

C

When you can either group at Lowe's, say I related dashboard and a figure out working out the container and family plan to use think the posts, continuity and the crowd it is in product writing states, but each one of them have the distinguished of the feature support. So that's why we we see what this can help our community to figure out how to choose their container runtime in their production and we go to next one.

C

So we also made a lot of other progress. I only highlight a couple things so that the progress made that I made it for the user names based report and the current we move to target.

C

We talking through the careful okay and the owners, many discussing to provide a bus tractor, a restriction on the resource class to support the more device and the color resource class it yet, and the considered complicate API so still under discussing, and we try to looking for more than GP other use cases more than GPU support to to to make this and broadleafs the process, and also there are a lot of progress made. The belief is in a heart rate and to ensure better scalability, and this one is retargeted to the one concerning run.

C

Some of the work is already merged and some students and the disgusting and the development we also english history. We also promote it across heightening states sharing to the data and also we fellow me out ship agreement under the our container idea and implementation. So the proposal merge and the implementation is under development, so that all updates for the q3 found a signal.

A

Alright, thank you done. Thank you. That's very helpful. Any questions for gone on signal.

A

Boy, no questions. You know we are placing through the agenda today: Thank You Don. Once again.

A

Two announcements we're already to the last part of the agenda here. So there is announcements from Paris I see that that is you're on the call. Yes,.

G

I am hi everyone. Yes, we came all right great. This is your weekly kubernetes steering committee announcement, parent myself, your and George are all election officials, so you'll be hearing from us throughout this entire election cycle. Listen up tomorrow is the deadline for all nominations. This includes the entire process, which is all the way up to BIOS uploaded to github, and that is the deadline is 11:59 UTC and we will be strict with that. Also that seemed line of tomorrow at 11:59 p.m. UTC is also the deadline for the voter eligibility forms.

G

What this means, if, if you are not listed on the voter SMD file, which I will include I, see that's not a link in the agenda right now, but I will include a link to that. If you're not on that file, that means that we do not have you listed as eligible to vote. The steering committee has decided that 50 contributions are hired in the last 12 months is the threshold.

G

However, if you feel like you have made more than 50 contributions, that would include things like non code or non github events that relate to upstream community. Please please go ahead and fill out that form we will be uploading. Excuse me updating the voters MD file with folks that are approved and then send out other communications to folks who are not next. This is polling. Ballots will go out this coming Wednesday September 19th to the emails that we have on file for those eligible voters if you do not receive an email by Thursday.

G

Also, please definitely check your spam and/or bulk, because this is coming from SIVs, which does do bulk emails. So if you have any kind of testy filters, please definitely check that, and but if you do not get an email from sibs, please contact community at kubernetes io. We can definitely send you a new ballot. So don't fret to think that you are not included in the election. We can definitely get that to you for get that to you.

G

Well, of course, remind everyone again on this call next week, as well as all of our regular channels, which includes kubernetes development, mailing lists, discussed, kubernetes, thought, I/o and slack, of course, and that's it for us and we'll be doing this again on a weekly basis, all the way up until October 4th you'll.

A

Election officials, for keeping it straight and honest thanks so much all right. Well, I guess we are heading towards one of the shortest current IDs community meeting today. The last part of the agenda really here shoutouts. If you know anybody you have done. If you know of somebody who has done something great and I want to say, thanks use the hash out outs channel in the slack. um So what I didn't gonna do is now a readout from the shout outs channel, essentially from last community meeting up until now.

A

The first one is from M ze thousand shout-out to ash thunder and whenever a for incredible help with CI signal, the next arrow is from ash. Soon their huge shout out to go anywhere a once again for lighting up the right fires when and where made it for one crowd way to go. The next one is from just August's shout out to Doug ahem James Ben tell their stats and anyone I might have missed for working the weekend to test our release.

A

Engineering, tooling, ahead of the next era, cut shout out from misty Luke Perkins for adding per heading for heading anchor links to the docks, so people can share an end page section at any level without having to go back in the talk to find the link next, one from neo light one two: three thanks to Timothy and Claire and Fabrizio Pandey, who helped her helped with debugging or released blocking e to e testing for SiC cluster lifecycle. Next one from M kuma tog.

A

Now we have why we 112 o beta to today's images, are all fat manifest. This made all other architectures. First, our citizens, thanks James Doug m XD, Lucas Caleb miles, P, pepper, mint, hello, Paris and FK mix Tyler from Paris, shout out to a make me qualms for helping contrib X with our communication platform discovery and doing the hard work.

A

Perfect example of chopping wood and carrying water last, but not the least shoutout from Tim pepper, huge shout out to bent Hildur for working late last night and right back to it this morning on diagnosing resolving bill pipeline issues in support of 112 release. So that's it from my side. Is there any other last-minute question people ask discuss here and go release team I! Think that's about it! Now we're gonna wrap up we're gonna, give you 28 minutes back to your life. Enjoy everyone thinks.