solo.io Friends of Solo.io, 4 Apr 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Jenkins X office hours 2019-04-04

Description

Rebasing & Batching pull requests - Pete Muir & James Rawlings
Provisioning Jenkins X K8s Clusters Using Infrastructure as Code (Terraform) - Oscar Medina
Knative Serve with Gloo - James Strachan
Multi-select questions for jx add app - Pete Muir

A

All righty, okay, hello, everyone welcome to another jakey's X office hours for 4th of April tonight, so we've got some. We've got some quite a lot of content. I'd see on the agenda, which is great, we'll see how much we can go through but as always feel free to jump. Add things in. We are looking forward to feedback and collaboration for everyone, so it'd be good to share just to kick us off. Absolutely, let's have a look. We've got a subject, that's being added around rebasing and batching of pull requests.

A

Pete, I'm, assuming you you added this. Yes, yes, I could see you allow me yeah there we go- and this is quite excited actually so and for everyone that just maybe Pete can give a bit of a bit of background. But um this is all tied into the process of our walking. The talk processes where we use Jenkins extra bill Jenkins ex one of the key parts of that is proud. Many folks will know be familiar with and PQ or they tell us about.

A

What's up, what's been up with a problem that the about some of the features around merging pull requests absolutely.

B

So what this is a really neat feature of prowl that we didn't have working in in Jenkins X for quite a long time. We realized we didn't, have this I, don't know before Christmas or something like that, so we only just managed to enable it so I wanted to kind of. Let people know about it if you.

B

So, if you think of the situation where you're you're merging pull requests, one of the things that you want to make sure happens is that you're testing the pull requests and against the current head of masterson occurred tip of Master, rather than whatever random revision that the person submitted. The pull request against right. You can imagine they they built the pull request a year ago and then they submit it and they never rebase on to master.

B

And you know if you test to get just test the code in the pull request, it might well pass, but you know it might well. Something might well have changed in the process. So one of the things that you want to do when you're testing pull requests is always, but one of the things that I think you always want to do is rebase the pull request and the code in the pull request or merged pull requests against the current head of Master.

B

So that was one thing that we added I mean that's a fairly straightforward feature.

B

The second thing that- and this is something this is something the prow would do normally, but because we were enabling Tecton in prowl for Jenkins X. That was a feature we had. We needed to add the second sort of related feature: it's a round, batching of pull requests.

B

So, if you imagine in a busy git repo, like the Jenkins X main repo, you can have you know: 1015 pull requests being merged every day and of course, what happens is that you know someone goes through and does a whole big load of LG TMS or approves all at the same time.

B

You know it's the way people tend to work and, of course you know if you go back to that previous feature, where you're trying to test every single pull request against the head of master, you can essentially create yourself this massive queue of full requests to merge.

B

So one of the really nice things that prowl will do is it will attempt to batch together pull requests that are ready to merge into master and see if you can get all of them to pass at the same time, and if it will, it will then just merge all of those and that avoids that huge queue of pull requests building up. So it kind of takes a strategy of waiting to get.

B

You know series of green ticks against a pull request, so it's not really considered for this merge pulled until it gets at least a clean run running against whatever revision, it happened to be running against once it's got a clean run and it will then consider any pull requests. It's got a clean run and suddenly gets on held, or you know, gets an approval. Something like that and try and put it into that batch.

B

What it will also do at the same time is, it will also sort of it assumes that batches fail more often than not, or more often than non batteries fail. I guess so what it will do is it will also pick one of the other pull requests and run that one in parallel to the batch. It will wait for the batch just to succeed or fail. If the batch succeeds, it will merge it if it batch fails and the other one passes.

B

It will then merge that other one and then try a batch with you know the next set the one was it just merge, so it's quite a sort of complex algorithm that it's only took us some time working out the debugging for when was doing batches and how it was doing badges. But it's now working really well, so I can just quickly show you what that looks like on deck, which is the dashboard that we use that comes with prowl.

B

So this is running. This is a slightly more up-to-date version of proud than we had until recently, and it adds this tied history featuring, which is really useful, particular this demo, and so you can see, for example, down here. We've got a couple of pull requests that that were read suddenly ready to go. I.

B

Think I was doing some reviews this morning and put some LG TMS on them and or retest them whatever and something they become ready to go, and you can see here it triggers the batch, so it triggers the to pull, requests and attempts to attempts to see if they will work.

B

It also triggers these are the single pull requests in the same in the same go, and that's that kind of two-pronged approach that I was talking about and then up here we can see that that pull request succeeded on the j.x repo, and so it then does the batch merge and emergence both those pull requests at the same timing, you'll see a few of those going on I think there's a few more I think it was one of it earlier in the day.

B

I lost it now, but somewhere in here, there'll be some quite easy scene. This is when it's not being busy yeah. You can see another one here where it's done. A very similar thing.

A

You've seen so far well.

B

Okay, so here's one we've got which this was a trigger of for batch four in a batch, so it looks like what happens here. Is that for some reason it's triggered a batch of four and a single and then I guess probably something failed. So then it triggered the batch or something went wrong with one of those. Suddenly one of those goes, those is no longer working.

B

It looks like 362 suddenly went, suddenly started failing for some reason or got held for maybe you're someone on LGG update or so on, pushed some yuko, so it lost the LG. Tm I, don't know so it triggers another batch and then it does the merge at less that's for in a batch, which is pretty good, but you know I've seen like the the JX repo, like particularly say like.

B

If, if you have a sir away at a conference, you know you can get when you're 30 pull requests building up now, suddenly someone's got jet lag, they wake up in the morning. They go through them all, so I think you could easily get 10 or so yeah for us at least coming up in a batch, so particularly useful for the busy repos so that that's really what I wanted to share was just you know a little bit about about what that's doing and how it's working.

B

We are still working on the some of the sort of UI, so both the CLI and the and the cloudBees UI for this. So sometimes these jobs don't always appear perfectly in the build logs, and things like that that you know we're working on improving that I think the batch is, do now show up in the build logs, don't know James, but.

B

Yeah, so it's a good, but it's a little confusing. If you don't know what's going on with it.

A

Brilliant thanks very much Pete nice excellent, so that you have really excited to have that. That's really helped us we're managing than the volumes of pprs and also the the effort of testing. We actually do on certain the JIT repo we're spinning up in doing installs I'm running different parallel streams of tech tests as well. So it's a it's much more efficient.

A

Okay anybody's got any questions on that at all.

A

No okay: next up Oscar we've got even the demo as well. Is that right.

C

Yeah yeah I have a demo prepared just wanted to walk folks through you know, I was playing around with the CLI and when I first started at cloud base and realized that it generates terraform code for provisioning, the clusters that underlying clusters- and so I wanted to talk a little bit about that. I really know to the audiences. Eventually, this will go away.

C

I know that James stretching's has mentioned about the data ups, which will most likely take over some of this stuff, but I think that there's some value in showing this, because some companies may not necessarily adopt a get up feature in Jenkins X right away right and some of them may already be using infrastructure as code as well. They may have some policies that says you can't vision a cluster in GCP or AWS Roger, using that method, you'd be able to see in light, but rather we have to do it biack infrastructure as code right.

C

So that's why I thought it would be good to to show some of that?

C

Yes, so there yep. So with that with that said, I was thinking um first walking users through my screen. Lisi can I share. My screen. James should be up so yeah.

A

This agreement agreed there.

C

We go yeah, so I was sure my desktop here and actually, if everyone can see I recently put together a nice article.

C

To summarize all of what I'm going to show you today, so I don't want to take up a lot of the time for folks today, because I know we have a full agenda, but I will post a link to this article, which covers what I'm talking about now and essentially will what we, what I want to do is show you how to quickly provision the KH KH clusters in DCP and walk you through that, and so here's the article anyway.

C

But what I have here is if I go into Google Cloud now I actually started the command to provision it, and it's doing that right now. But let me walk you through that solution. Okay, so here's here's the the code that got generated from my JX command right, so I originally executed.

C

We'll hear this command right here is our first step right, that that creates the three different clusters: production, staging and dev, and so with that I got the output of my what I call my repository with all that structure right and so that's what you're?

C

Seeing here you can see clusters product of age and I decided to just work on the on the dev one for now, and so what we're seeing here is you know the terraform modified code that I that I tweaked a little bit, especially just to ensure the credentials file, is in place where, when I execute the terraform command right to provision the cluster, and so you can see right now, the terraform apply command is actually provisioning, the node pool- and it has already provisioned the cluster itself in the right location that I indicated, and so you can see I take you from the beginning.

C

When I execute a terraform plan, it tells me what is going to create right. So it's going to create the cluster. It's got a few properties and, and it's just going to pretty much- this looks like you know, a good disk.

C

Basically, as you can see, plus or later on, we'll see how I will modify the cluster and then make some changes to it and and by the way, I want to first say that I have a big disclaimer up here, as you can see that this this approach, basically like I, mentioned, if you're using get-ups, which is pretty soon going to be available I. We recommend you use that. But if you, if you're not, then this approach works too.

C

D

As usually changes cloud infrastructure, it always use tariffs, right orchids and service accounts and clusters and then get ops, yeah yeah.

C

E

A

This is also also be able to use leverage a lot of this through the multi cluster. When we all.

C

Right, right and I think we got some pretty good content that we're planning on for that which I look forward to. But one of the things that I wanted to highlight is that for this only changes or updates to the cluster right. If you destroy the cluster right now using this I I happen to me, you lose the data right. You lose some of that configuration, hence the whole get up stuff we mentioned so going back to looks like we're still creating the node.

C

What I did when to just following the tutorial there that I wrote I got the credentials file from the GCP.

C

You know I am setting, so what happens is when we execute the command and VX to create the terror form. It actually creates those service accounts for each cluster right. So each one of these clusters, I notice, that it creates the accounts and, in my case I just downloaded the JSON that describes that account. It has all that stuff there which utilizes in the provider right here, so you can see in the main TF file it is, it is being used as the credential.

C

So that's why we're able to execute this command and in how to create the cluster stuff. So as soon as this is done, what's going to happen is the next step. We we prevent the cluster. Now we can say JX install, so we can install JX on top of that cluster right.

C

So you see there's a separation of tasks there one is to provision the cluster and the other one is to actually install Jenkins X on top of it, and so that's what I wanted to demonstrate here so we'll give it a minute or so and then just walking you through the article.

C

Essentially, we already executed several of these steps right and we've got the credentials so step two.

C

We we added the credentials to our terraform provider or back-end right, because what happens is- and this is common in any environment right if you are working as a team in terraform infrastructure, as code you're going to have a back-end, what terraform calls it back in and that back-end in this case is a bucket and in GCP and jenkins x, when you execute the command it per, did it creates that the code to create that bucket so that now not only I can work on the infrastructure, but James can go in there if we put it on github, but now he can also do a terraform, apply or or work on the infrastructure.

C

So this is actually quite important right. It's not just a single person working on the infrastructure, and so we've already done step. 3 I've done step 4, which is just plan to see. What's going to be created, we that and if I scroll down here we're actually already creating the cluster. So.

E

That's happening right now, our.

C

Next step shortly after that is going to be, let's set the cube, CTL context, so that we can then allow the jxc Li to execute the JX install right and so that this is where we are right now and so when I, actually, what it does is. Basically you know we use the GC cloud to get those credentials in a terminal and it sets the cube config entry so that we actually for the newly created cluster. It now shows up in the in the context right.

C

So that's what I'm, showing here it's one of these I didn't highlight it sorry! Well, let me see if it's done. Okay, so now our cluster is done. Here's the output, you can see the endpoints and the version of the cluster, and so now what we want to do if we wanted to go into the UI just to see, there's my cluster right, so you can see all the settings that are specified in terraform are there. So my next step is I want to connect to it.

C

The other CLI, so I'm gonna just copy this, which is the credentials and I'm gonna, go back to my terminal I'll just clear. This and I know that that is the correct command, so I'm just going to do that that creates that queue CTL entry as I mentioned so there it is, and now, if I say queue CTX and just with them there. It is my newly created cluster, so now I can actually say JX install now, of course, I can pass a lot of flags. I. Think I like to do the default admin password.

C

I'll, give it a secure password like so, and.

E

C

Installing djenka Jenkins X on top of that newly provisioned cluster ready.

C

So let that go take a little while I should have specified the provider, as you could probably tell so. I'll just say like this here.

C

It's gonna do the same thing: typical, install right, I didn't select any special flag. So now you still learn a bunch of stuff which is okay, I'm, not even going to use a custom domain which I have in the past, and we just wanted to go through this quick, install and what's going to happen after I install it is that I'm going to make a modification to the cluster itself? Okay, so the cluster right now and in this is will let this go here. I'm just answering questions here.

C

What what I want to show you is after Jenkins X is installed on this newly provisioned cluster I want to make a change in the cluster itself and maybe add a add-on right with what Google Cloud calls in that on.

C

In this case, I'm just going to show you- and this is deprecated already, but it's the that the kubernetes dashboard add-on so we're going to enable that and you'll see that that does not affect our Jenkins X installer, that's completely separate, so it should continue to be functional and all that, despite the fact that we are making a change via terraform to the cluster itself right, so two separate things there.

E

C

So that's uh if we go back to our step here, you know we've gone through that we're actually installing Jenkins X now and so now. You know shortly after that. I'll show you that modification to the cluster and that's modifying the the code that we already have.

C

So what we're gonna do is we're just going to add this snippet of code there with which tell us the cluster, hey, enable the kubernetes dashboard again, that's deprecated right because we're we already have visibility into this, but I wanted to give you an example of how you can make some of these changes to update the cluster and still have it be functional, as it should be right. So, let's see so I'm going to use the all of the default settings here.

C

Just have it go through that.

C

And I'll choose the static when, although I know folks prefer the server less these days, which I have yet to try actually.

C

We'll make that good, so this will take just a couple minutes here and again typical install right, but the big difference here is that we provisioned our our clusters. Three separate clusters. We can provision one for dev, one for pride and one for stage. Well, you know a lot of people have been asking about that and I think. Maybe while we wait for those James, can you I? Don't personally know you what the big difference would be like the upcoming support for the multi cluster and how it relates to this is any insight.

D

The basic idea is people want to segregate development from staging from production, usually on a separate family, so they can't see each other at all, they're, completely black box Denise. You know Chinese walls between them, which basically means you need like three clusters. The terraform study me going right now. Let's you spin up three clusters. If you want on one or two or three, so the idea is yeah, but your clusters were terraformed, then use multi, cluster Jenkins X to orchestrate promoting between dev testing and stage, and even if they're, completely separate clusters.

D

C

Okay, so this would you say that what we're doing here, though, it would still be somewhat applicable after we have that official multi class, their support.

D

Definitely, okay! As soon as you go, we'll take customer go mole cloud stuff to manage right. So you absolutely good right, ready.

A

Go thank you are that in the underground, the multi cluster, so we allowed to these, create already creates multiple clusters using terraform is.

F

That right, yes, that and it normally prompts you to install Jax in the first one I think it defaults on one con, dev of them yeah.

A

And the idea is, then, we can add in a whether that's an environment, controller or something that's actually going to into the different environments and different clusters that can receive web hooks for a guitar and repo. Let me do the deployment so.

C

Like right, which means like imagine, pushing from one cluster in this case there to staging right, so those web hooks help you do that, go from one environment to the other right yeah, sir.

D

C

Yep, so let's see this typically doesn't take too long. Wanna see I think uh we are in this step now. Let's see this should be updating, so my workloads here should be all green, just the Jenkins one. It looks like it's finishing up. Let's just wait for that.

C

Should be almost done and.

A

C

Our negative, like I, said there.

F

C

Okay, yeah perfect, so you know we could act as the Jenkins UI and all that, but I won't do that, because I know it will be up and running all right. So our next step is: we've got the whole installed working right. It's finishing up the the repos here, and so our next step is I just want to show how we can modify the cluster and making a slight change again. This is more of an update rather than a change that would call for a complete resource.

C

Rebuild we don't we don't want that right, so I'm going to go ahead and uncomment this, which is just enabling that add-on like I, said and now, if I do a tariff one plan, what we're going to see is is going to say: oh I, spotted a you know, change right, so notice how this is different, so the tilde versus the flush- it's just saying: oh you enable this. You want this enabled great so I'm gonna, say terraform apply and we're going to quickly make that change story Questor and I'm going to confirm.

C

E

C

Going to make that quick change on our cluster right, so what I want to show? You is well that is executing.

C

E

C

Done I'm just going to go ahead and grab that token, because I want to proxy the queue CTL and then quickly access this URL. Just to show you that that change is actually doing right.

E

C

Let it finish this change.

C

And James, let me know if I'm running out of time, I don't I, don't want to take up a lot of time, but this should be the last step we're going through here just quickly once we see the organ, that would be the final step in our demo. James.

A

Talking up nice to do a demo on key needed service before waiting as well. So that's convenient very exciting as well.

C

And there are other properties in here that we can modify right and one of the things that one can do and if I just expand here, the definitions for this resource, as well as this pool resource, are all under terraform site. So there there are some properties that we can add and modify, and that will not destroy the cluster, but rather just update it like we're doing now right. Just an update to the resource, which is important to note again so with this finish, shouldn't take that much longer.

C

And then again, the goal is to once this is done: we're simply going to go ahead and do a quick juicy tail proxy and access the dashboard to prove that that change actually took effect, which I think I practiced like five times before before today, so I'm, confident that it will that it will work. But are there any questions on this while we're waiting for this someone curious about this whole process. I.

G

Asked can you hear me yeah, yeah, okay, this is this jon hamm working for I'm working on doing some tutorial documentation for the Jenkins axe, community and I was wondering it's so obviously, terraform has a lot of event. Create JX to create terraform has a lot of advantages over Jack's create cluster, but if you're coming as a new user, who's, probably never even created a cluster. Never even done any of this. Before do you recommend doing JX create terraform over Jack's, create cluster.

C

So I think that's a good question right, because what I, what I would answer to that is that maybe in your environment, your shop, you guys are using terraform and that's the mandate per se or if it's, if it's okay, you know and it by all means the CLI is faster right. But if you need to tame and have some control and process, then you want to consider how to approach it right, because anyone can create a cluster quickly using the CLI.

C

The question is: how do you manage it after and that's tough I think that's the the point of mine today is like look. We can do this easily all day, but how do we maintain it and it doesn't become problematic in our department right and that's I- think that's part of the topic here, and so that's that's. Actually a good question.

C

Terraform, obviously I'm a big fan, I've written a lot about it. I think it brings a lot of value into your. You know. Even putting these changes to the cluster, so Jenkins itself would be wonderful right, and so, when I think about that, your process can be pretty neat and but but it really depends what what your shop prefers in a way right, you can still payment. It's just more. What deference right.

D

Yeah, you probably will that quick way to get started so you'd, probably just use Gerry cluster. Most people kick your tires with stuff. Just so you'd keep your tires by doing jokes per cluster and say: hey yeah. This tricky sex kind of stuff is kind of cool. Let's have a production install once you start thinking about a production install. Then you probably Tara for me. So you do.

G

C

So it's like establishing your process internally now you're starting to say all right. Let's make sure there's a little bit I've tested it out. I. Could you know kick the tires around I like it? We as a team, we're gonna, invest it in it. So let's go ahead and set up a nice workflow for ourselves. You know us because I help you! Oh that's.

G

C

And so just part of the last step here in our demo today, I wanted to quickly show you I executed, I, just grabbed the credentials real, quick, the token so that we can proxy into our cluster and now what I want to do and again, based on following this write-up I had yesterday I'm just gonna click on this link, which takes us to the dashboard. But it's going to ask us for credentials right, so in this case I'm going to use the token that I queried for- and we should see our dashboard here there.

C

It is right. So that's that's kind of the end of our demo. Here we modified our cluster. We have JX installed underlying changes, took effect and and we're we're up and running right. But again these changes are updates. They are not changes to the cluster that will destroy that's the big takeaway here right.

C

So, yes, that is all I wanted to show and again I'll post the link on the chat to the article, because that has a lot more information that you can take your time and and read up on it and, of course, I'm around on slack with any questions, I'm happy to answer yeah. Thank everyone. Great.

A

Stuff thanks walking us all through that they're asking us also a shout out to Garth happens as well, contributed the terraform stuff, as well as a JX, which is fantastic as well. Yeah.

C

Thank you guard I'm, just leveraging what you've done. Man.

A

Community staff, as well as you see anything else in the other people, have ideas as well for anything else around this kind of stuff ideas, issues and I mean all.

D

Right, okay, by the way, just before we do anything else, there's also a blog which we should take out. That Gareth did. That is the next level on from the terraform that sets up a CI job internship, a SAS base, CI engine, so that you can get automated terraform builds in a SAS, because you get this inception problem with telephone that you can't lose Jenkins X to run telephone to reinstall. The cluster with Jake is X.

D

It gets a bit kind of inception, so you often wonder kind of have a SAS that when you telephone, which is independent of you, see ICD system readiness, that's never with a good one. Yeah.

A

Great stuff: well, you made a carry on Jake. Let's.

D

Segue this sequence. So let's talk about okay, sir okay dating sir, so um let me share my screen.

E

D

So I created a cluster gke this morning, I think I'm using the Tecton installed. If you click on the Jake is aekta pipelines docks the jaegers.

D

X pipelines is basically the way of using tech, time and service jenkins as a way of getting to the cognitive service, see on TV, so I've got a cost idiot, that's using Tecton, and if I do like q CTO that pod you'll see that there's no Jenkins server here, there's just some stuff from prow, that's prime stuff as positive as I think I'll pipeline runner, we've checked on and which means the Jenkins explains, and then that's pretty much it. These are just the background, bits and bobs.

D

So there's no jacket server, there's some Tecton awesomeness and that's it. Let me clear my screen you'll also notice I'm, not running any pods in the staging cluster. In fact, let me do my.

A

With increasingly the fun just a little bit and not awesome yeah, that's fine.

D

Now let me just talk a bit about case. Ok, can it asserts? Okay native is an open source project, lots and lots of contributors. It's got kind of three pieces. The first piece on the website Kintla to build, is actually kind of going away. That's becoming Tecton, so Kennedy build is going away, so Kennedy was really well. We've often used ki netic.

D

To me, one of these three things increasing the ki netic is just going to be really about kinetics serve, which is about writing little functions or big functions and exposing them on HTTP, but there's also kinetic eventing, which is about binding to cloud events like calf colon bucket was written to any kind of event in your system. So it's really serving on events. Now the serving is really the core piece of community serving is the bit that takes in a function, change that any function, that's packaged as a docker container and exposed over HTTP.

D

The eventing piece is really about bridging events took HTTP, so it will consume katka messages or sqs messages and s3 events and that kind of stuff and then invoke a HTTP endpoint. So it means as a programmer. You can pick a programming language in and go don't jam or whatever at your functions at the HDP and then Kennedy takes it from there. So I thought I'd start with just something: that's running in chickens X.

D

So if I think Jake's get applications, I created something earlier and I'll just invoke this and then I'll run a quick start in parallel and show you a quick start working. But here's a quick start, but you know it's there's no pulse and if I, just click on that URL it'll take a couple of seconds but notice. Pods are spinning up. So there's the pod spinning up and then any minute.

D

Now we should hopefully get a web page, so I we're using creative serve, which is basically elastically scaling the servants thing again so that went from complete cold to warm hard. So we're now running one pod you'll find do that. One more time.

D

One pod with two containers and what Katie said this is evens a little side curve in your pod. So he modifies your pod so that there's your container and then add a little side cap and then creative. Does this auto-scaling thing that the more requests you get, the Mod Podge sequence? So it's really a way of automating taking functions, wrapping them in the docker container and then auto scaling them up and down you must've, which is really really nice and what they were actually doing.

D

I should have explained this I'm, actually using something called glue from solo I can, if you serve, is available in two flavors. One is with this guy and one is with something called glue: glue from solo dot, IO I went with glue just because glue is really tiny, and let me show you how big glue is.

D

If I do you get nine spaces, you see, I've got one layer, space can I, could serve I'm one day space glue and if I do get called in the glue, there's both small pots oops sorry I missed the main thing there. We go so there's three pots in the grid no space and then in the kenneth who serve in space. There's a couple pods and if you try this yourself, it literally takes like 10 seconds to install, because these are these are six tiny little pods and these pods are doing ingress.

D

Cluster low, bouncing occur native pcs. Doing the scaling up and down and glue is doing the ingress it's using my proxy and he's doing there, the auto-scaling and all that kind of stuff, so that's glue and connective serve. Let me talk briefly about how all this works with Jenkins X. So let me create a brand new, quick stir who actually before I, do that we've got a new command.

D

Jake's edit, the chlorine and Jake said I deploy, can work either on a project basis or the team basis, and if I'd be minus T, it's going to edit. The team default, so the team default I can set the team default to be default, which basic just means use of kubernetes deployment rather than okay native. Why?

D

If I pick connectivist they're going to set the team deploy kind to became it, so that means every quick stir I create from here on in is going to default to use connective serve rather than using a cumulative deployment and the community service. That's.

A

Doing that us using the helm, values I'll.

D

Show you that in two seconds so like we should create a quick step so I'm going to create a quick stir. The quick answer is yes: James I'd like to show you a quick stir and then I'll show you the result of this time. Let's call this office node office, node and I'm gonna do a node.js example, but this works with all of the quick stats. Get me posit Ori that there and it be scrolled down for this bit. You might spot a new magic line.

D

We've modified the home file that but other than that, we'll just create a normal Quick Start, and if I do the normal gate get activity which gives me the pipelining me now and they will do the normal. You know: do a release, release of doctrine, a Teresa how's that promoted to staging a blah blah blah. So one that swing. Let me go into this in QuickStart called office. No, and there we go.

D

Let me quit my open vs code and let me show you the code, so you see the pipeline is running there with Joe gazettes pipelines and Techtron and I kind of stuff. So that will obtain your time. You see it's reasonably quick, quite quickly so on and if you with node as well, he's even faster than using Jovian stuff, so that pipeline should be ready to promote very soon so he this is the quick start. We just created it so normal, no js', quick start.

D

It looks very familiar if you've tried the QuickStart before there's a slight difference in the hound chart. So if you look inside the vines, Yammer series, magic line, can you deploy to our force? So there is a hound flag of whether you want to use KNAT to do deploys or you want to use traditional Cuban, ities deployments and services, because this is true.

D

This is going to enable care native serve deployments and what that means is we create a pony quest here as the normal dinner public, where staff and the pull request is the same kind of pony quest is normally if I look at the pony quest, it's a normal promotion for requests and look at the day if it's normal stuff, nothing's really changed, but all the magic has happened within the Hound show. So that's a normal promotion, nothing to see there the magic happens inside the help of the templates of the chair notice.

D

We've got that service for Cuban eight service, normal South. We've got the Cuban ease of deployment normal stuff, but notice. We only do the deployment if it's okay lately. So when you use a service, if it's not connected when you use a deployment, if it's not connected, but it is k native. We make this other thing, and this other thing is a community service.

D

A declarative service is the kinetic serve thingy, which looks a bit like a deployment, but this is the CID that does auto scaling up and down so we'll see that promotion is just about finished, and so, if I do.

D

If I applications can see, is it ok? Not quite so, if I do, SVC is the name of the CID and if I do you'll see, there's one there's one kinetic survey right now, and this is the domain name and the staging repository is probably chicken.

D

The stage we play is probably dopey right now, if I just do that, we should see the master promotion sugar in a second, so masters, not quite good, with only commercial that we have anything more to domesticate. Oh there, it is okay, it's just happened. So you see those the office. Note thing there, so that's the kid native serve object and there it's now got a domain name. So now, if I do jxk applications we'll see, it's doesn't have any notes. Yet the notes up inside that we've got a near wall.

D

Now, if I do my watch, the pods again get old and stay tuned, it's gonna, be you see. The pods are just the other, the other parts of terminating because we're not using the service anymore.

D

So the old demo is going away because we're not using it anymore and if I click on this one there we go there's our demo, so we've created a QuickStart, the same normal way of making the QuickStart, but it's using Kennedy Serve right now, so it makes this CRD and this 70 then generates pods dynamically under the covers and if I wear it like 10 minutes, you can scale this back down again now. The other thing that's kind of interesting is you can choose to switch and traditions whenever you want.

D

So if I wish directory I can also do J edit deploy- and he says, o urk native right now, let's go back to the default model and it's modified my values jung-wook. So if I do a git diff we'll see it's changed this line from true to false. So if I now do.

D

And I'll go get push and then I do it. Watch will see a new release, 511 sugar and, if you've enough time to bother watching it, it will now make the next version of this quick.

D

Starts that will not use connecting so the K native resource will be uninstalled and it will replace it with a service and the deployment again. So, in other words, you can switch between connect either all defaults deployments for any Qwikster at any point in time, either by changing the release chart for Charlie you release or while modifying that flag in the environment, so you could use care native on pull requests in your preview chart that you could switch the deployments for stage of encryption and switched to you can use you either old.

D

So beyond this I think you should always just pick one for the lifetime of the project, so stick with either kinetic or use deployments trying them in different environments. You never know if something might not work in one of them. Yes, you can switch. That's it! That's me.

A

Last question: oh, that's called a question before before I ask a.

H

Quick one is this: I think it was Vincent Behar just had a blog about Osiris and yes, it scales down automatically. Is this the same thing, but this is.

D

The same idea is the same idea: it's just a different service to achieve the same ends. So a cyrus is really cool as well. By the way I'd like a cyrus to be an add-on in jacket X. So then you've got the choice. Do you want to use a cyrus ordering one use Kennedy Serve the nice thing about kinetic serve?

D

Is it's really designed from scratch to be this autoscaler, and particularly with the event in the eventing thing adds on to Kennedy Serve, if you're, just using just kinetic, serve and kinetic serve and Osiris are very similar, the nice thing about the Cyrus. Is you don't have to change your Yama at all, so any deployment will be scaled up or down with a cyrus with kinetic serve.

D

You do have to change your custom resource definition from deployment and service to care serve, serve but they're, very similar they're broken and when they're, both fairly simple they're, both fairly lightweight and I, think both I've got a a good position. I think Kennedy Serve is slightly better if you're going full-on functions of service or service because of the eventing piece right, because you can bind cloud events like.

D

Which a Cyrus doesn't really doing that Cyrus is just a module scalar, but yeah I mean there is the horizontal pod auto scaling as well. So you got three choices with given ease. You could use a horizontal pod or two scalar with your deployment. You could use a side which kind of alternate orders kinky, so you might notice by the way it's already torn down my or my pods, the pole of there's one morning and he's just made it.

D

It just made the deployment because I'm just about to promote the just a few more it's there. We go he's starting the deployment now and it's gonna tear down the old one. So if I look at the pods now I wish he could hide the terminating so.

E

D

No non-kinetic serve deployment. You see that on there we've switched from Kennedy Serve to a different, though it contains yeah.

A

So so what's so, he makes about glue what's what's the rolling that glues doing in this is that the the API gateway? Yes,.

D

So glue is basically what glue is a lot of things: it's basically a service machine awesomeness and in particular one of the downsides with this year. Right now is history's massive I tried to do this demo with East. You know minutes work, as histories has got a gazillion things in the box, because this year is doing a lot of things and it's doing stalls. You know. 50 containers takes up a lot CPU and these doses and everything in the studio is calling all glue, has a really tiny footprint.

D

So if you just want also scaling the services, there's something called super glue which will we're looking at for handling a facade to in familiar to cluster, so that developers don't see that you're using many different custom. Super can find most requested, but we're using just glue with them in which is basic. Just an API gave it. So it's a very small in by path. If I get me that really just takes care of the English side, things and then it plugs into Kenya T.

D

So so can, if you serve, is the thing that scares up and down and then glue is the giveaways in groceries and.

A

Yeah big good fame both have a go and have a little look at I projects. They're open sourced, they're, superb they've got offering lots, lots of features around prove the API gateway fans like retries and auth, and you know all the all the great stuff you get with api gateways and also they're great people as well. So this good community around there so encourage folks to go in that. Take a look at what so they were doing.

D

I wasn't thinking, I mean I'm a I'm gonna document up this demo I have no chance of it. I'll probably get next week sometime the.

E

F

That I found kind.

D

Of I found quite interesting is we've managed to take all of our bill packs and enable them for connected syrup or deployments. We should be able to make a cyrus an app as well, so you can choose to add the cyrus in a similar way, so we could oversee race or kinetic, serve or deployment and just easily move these things.

D

One thing we haven't done yet is turned glue and committed serving to an app and jenkins X, so you can just install that by the way the current installation instructions are fairly easy, but there's a couple of little tweaks, you have to do to get the external URLs working, so we're hopefully going to make that really slick. So it's just to be. You know one app to install and connect to serve and glue together in one app.

E

A

Lovely stuff, I love, it I'm sure everybody else as well. This is superb and we've got seven minutes left Pete I noticed you've got three things on the agenda: I mean it for each one.

A

A

B

Do you want me to do James, whichever.

A

You think we've got time for I guess: we've.

B

A

Feedback from Sumer.

B

Well, let me let me do the let me let me start on some of the app stuff and we'll see how far we get so.

B

Okay, let me show my screen right, show I, guess just just going on. Okay, so James was briefly talking about apps there, but apps are what we're working on to replace add-ons as the way to extend Jenkins X and that and the simple differentially the differentiation is very simple. Add-Ons require you to add code into the Jenkins Xcode base, apps, don't so with apps we're basically trying to take all the things you can do in add-ons and and make them so that you know you don't have to make any modifications.

B

So one of the capabilities that we wanted to offer was the ability to ask make it easy for the user to answer some questions when they install the add-on. You know Jenkins eggs, always pumps. You do want to do this and you add you want to think of that and so on, and that's what I'm gonna show now so just to sort of very briefly on apps apps are basically just helm, shots there's really!

B

No different difference, you can do J, X, add up Jax, leta objects, get out, checks upgrade app and you can actually just pass any help chart you like in and that will get installed properly. What apps do is they add a number of extra things on top of helm, charts in stuff, like secrets, management, the ability to ask questions some nice stuff around upgrade the thing I wanted to show. You today was around asking questions.

B

So, if you're writing an app- and you want to ask some questions of the user, what you can do is you can write a JSON schema. I've got one here and what the framework will do is it will take the JSON schema, you just call it value schema JSON and for each property in the schema. It will ask that as a question, so this implements most of JSON schema. All of the validation and formats are supported. Almost all the basic types are supported. Array types of support. A properties are supported.

B

There's a few kind of like sub schema validation, things a sub scheme of things that I didn't actually implement yet, but we can implement those if we find we need them. So to give you a demo.

B

This is a reasonably advanced use case. So I've got this app cheese I want to add it. So I can do JX, add app, so this is going to pull it from the Jenkins X chart museum by default, and that's where we publish all the apps too. If anybody wants to write an app, we have a github org and we can publish from there to that to that chart, museum for or publicly available open source charts.

B

It also supports writing your own apps that you can run and publish your own chart museum and you can pass potentials into the commands so I do the add what its gonna do is prompt me these questions. So the first that's not quite right. Let me just do slightly improve that so a little bug around my rendering, so let's just fix that with the updated version. So it's going to ask me.

B

Firstly, the first question: I've got up here, so this one supported architectures I can do a question mark and it will show me the help text which is specified in the description and because this is of type array- and it has this enum, which specifies some values is going to present this to me as a multi-select list and I can go through and I can choose the ones I want. I've got a second question in there for support operating systems. So what it's doing here is it's using these?

B

These values from the defaults thing in the schema as the values to select by default and then giving me all the options from the enum Center and it will then do the install and what I can then do is if I do so. All this is doing is basically just installing a binary plug-in, which I think I demoed last in the last office hours, so binary plugins are managed by custom resources in community.

B

So if I do cube, CTL plugins yeah well just to take a look, you can see that the temp, the what we did was we generated a values. Memo from these questions, and then we use this to apply for the template and and generate this value. Did this plug-in store animal? That's going to be deployed into the cluster? So what the the question is? Do they generate the values camel and then that's just used in the standard? You know home way to install the chart.

B

So this is quite nice because it allows you to not only put some default values into the the app that you're building. We don't have the user override them on the command line, so that was that was cut five minutes. The other thing I can just show you very quickly as if you development, you can also do this from the command line. So if I did from your local directory, so I do JX, delete cheese, just leet it quickly and for djx ad and if I just specify the path on my local machine.

B

I can remember the path, complex, go path.

B

So I can actually just add the app directly from my local file system as well. So, if you're building apps, this is really handy because it allows you to test out your local code. Obviously you know if you're going to install a container or you're going to download binaries for the internet, you've got to work at how you publish that container and you've got to handle that I, think dev gods and some of the scaffold stuff we have would could help with that.

B

But this at least allows you to kind of validate your questions and things like that without having to publish this and go through the full publishing cycle. So in six seven minutes that was a couple of my demos done: I, don't have any time for questions feel free to ask questions on.

A

/P that was brilliant, well played together.

A

One last question: just so: we didn't get many most answer, four questions through that is it working doing every two weeks? Should we try and should we be thinking to try and do this every week? Can people afford the investment and the time to join this every week? Anybody got any thoughts on that I.

G

Think it sounds great I, wouldn't.

C

I would like that yeah definitely.

A

I think we've we've got I think this powered showed that we've got a lot there, that we look in. One of these reasons, for these calls is to get feedback as well and I. Think so I think we look. Maybe we're packing quite a lot into the into the into the hour. So maybe let's try it see how we go on this one for next Thursday, then and didn't see and take it from there and keep going people. Okay with that.

A

Yeah all right see you all online time so much for the great demos they brilliant great stuff, like I, think.

E