Jenkins Platform SIG, 16 Jan 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Platform SIG 2020 01 16

Description

Jenkins platform special interest group meeting with topics including AdoptOpenJDK hotspot progress report, AdoptOpenJDK OpenJ9 progress report, and Docker image build experiment on physical hardware instead of QEMU emulated machines

A

Hi everyone welcome. This is the Jenkins platform special interest group meeting. It's the 16th of January 2020 welcome to a brand new year. Let's start looking at the agenda and go through our meeting. So here's what I've got we'll talk about open action items. First, then we had an item on the adopt openjdk hotspot transition, progress, open, j9 progress by renaming the agent docker images.

A

Let's see I guess we should probably put the people on this so adopt open, JDK Alex. Do you want to cover that or is that a gym topic.

A

Sorry Alex I didn't hear that say again.

A

A

I'm having a little difficulty hearing, you right.

A

Now, better, yes, that's much better! Okay, sorry! So back to adopt was that something you want to take is: should we just look at it together in the meeting to see where we're at I think.

B

A

That's a Jim problem great and if Jim that's a surprise to you will I'll do some talking to it as well. Okay, but it's all good great all right! So then renaming agent, docker images so like you've got that broadening the platform. Support for docker images, I think, is a general expression of the well. Let's, let's put that one also on Jim. Are you okay? If react? Have you act as the lead on that conversation?

A

Yes, this one and then google Summer of Code, oleg QA, qemu and docker manifest integration and CD pipeline Jim. That's you and hopefully we'll have you show us a demo demo. There.

A

Then, let's see I think it looks like that and then support, Alpine or and newer OpenJDK darker images, and we had and I'll have to take a look at that when I, don't remember, I I think I'm, the one who put it on the list, I'm not sure I, remember what it is, we'll talk to it when we get there anything else that needs to go on the agenda before we start working through the agenda. Oh.

C

Nothing specific, so there was a comment in.

C

One of dr. images about restructuring the packaging pipelines so address it. If it's running platform, support topic, ok, yeah I'm, not sure whether it's an intention.

A

Mark I think it's either here or like broadening platform, support or Jim- maybe talking about it on this one, because I think Jim was one who raised that raised the question.

A

So yes, I think one way or the other we've got it in in the topics. Ok, all right! So then, on action items, I am sincerely sorry happy holidays. I still have not opened that chip for the docker operating system support continues. Online Oh like you've, got the action item on the jet for Windows support issue. It.

C

Will take a while, so I will definitely need to deliver on that some of them later, because we already start to factor in you and your service repair. Oh.

A

Ok good, so this one, this one is is the when SW project in google Summer of Code is that and that what you're, what you were saying did I understand it correctly.

C

Need to start it's not JSOC or it's not okay, so it might be also in JSOC for llamó support and other things, but yeah right now it happens out surgery, so I see. Okay. Thank you. It's not specific to G sharp Thanks yeah. So for me the main objective in this story is to specify which Windows versions we support and hence.

C

Transmitted to which dotnet versions we support and also even API version, we have support because we still have windows, process management issues in the libraries. So it's a bit far from my immediate priorities, but yeah I'm going to finish it. This trip.

A

Garlic all right well and- and this is an interesting time for it, because I think as an example, Windows 7 is officially off support during this month. Bye-Bye Microsoft.

C

Well, it's a good statement, but what if we talk about the embedded versions when they will stop support of embedded versions? Oh and that I don't know could be in 10 years, maybe later right at the same time, I'm pretty sure that there might be users who use Jenkins to run installation on such platforms somewhere they call them exotic, but I, don't yeah.

A

So, whether or not Microsoft supports the platform may not be as helpful as I I made reference good point. Yes,.

C

Or platfrom yeah desktop operating systems are being duplicated quite quickly by Microsoft, but if we talk about it embedded platforms yeah there is a long-term support for them. I believe that even exposed it got it. Thank you.

A

Okay, Alex anything. You want to report on Windows, installer and code signing.

B

We're still waiting on the code signing search, apparently the company that's going to issue this certificate is taking a long time to five the new legal entity so that it's still in the pipeline, yeah.

C

Alex I'm not sure whether you in the CDF slug but yeah, that is a threat in operations which you can join. There is the reasons that some data about that from them mulligans link.

B

I'm in the I'm in that chat, but I saw a conversation between a demurrer. It is, but in olivier in about maybe switching to a different organization. They get dessert to pick your poison. They all.

C

Mollusks, at the same.

A

Rate all right, so the cert continues to be the challenge and will continue to work on it, because it's also blocking the the progress on the automated release process for Jenkins. So we've got multiple reasons. What we wanted to get that signing thing fixed thanks, all right, so Jim should we take on adopt open, JDK hotspot, yep.

D

So updates, coming from upstream with that or downstream whatever mighty support for sent us is being looked at right now for a PR to get into the unofficial, adopt open, JDK, docker image library and once that's in the unofficial, there's a not like an official process but ought to go through some testing itay, basically get into the official one which will help us, because now we you guys support San Luis along with sent to us. It has clef OS, which is the s through ninety kind of port yeah. So that's really good.

D

In terms of the testing side. I have a PR open against the doc test. Testing suite you or I forget when their name, the repo is 100% but I reworked all the docker tests, which was a challenge, because this is like me, diving, very deep into Java.

D

We like they run a bunch like external tests like run wild fly, run open Liberty. They run Jenkins as one of their extent external tasks, so I guess really just common Java programs, I guess and I, reworked all the docker images to be slim as possible and once that PR gets submitted. What we're gonna do we're going to actually add right now.

D

They're only testing I think like a Debian based or Ubuntu based image, and what we need to do is expand their test framework to support multiple bases and basically that will allow the adopt docker image repo to utilize those tests and once they go through the wringer on all those tests, they can get promoted to official images and thus, eventually we'll get Pat to us. So it's going deep in the rabbit hole, but it's going well a little slow moving, but it has momentum so.

A

That's great: that's that's really! Wonderful, okay! So so what you're seeing is that the be existing? Adopt, open, JDK Debian? We we know we need, and there she looks like really good progress on CentOS.

D

Yeah, there's a there's a couple other base images open JDK supports right now, but I can't remember them off the top of my head, but they didn't support. Sent to us and I know that's one of the ones you guys had so thank.

A

You thanks very much yeah any questions from the group to the gym on that topic.

D

Yeah they're there there's gonna be a good amount of more images. Don't be. We supported you VI, which is Red Hat's Universal base image, which we really knack of all their weird ones out there, but I think UVI be really interesting for kind of more enterprise solutions. So.

A

Good yeah I know that I know folks who are interested in a ubi image as well. So that's very good yeah.

E

A

Thank you all right. So, let's talk open j9 is the next topic Jim. That's you again.

D

Yeah, you guys want you jump in here and fill some blanks, but as far as Ana I know the the PR got submitted on you guys got you guys's end and got accepted and a couple of images got built, but I think there was a couple issues that Aaron, who were in the open, j9 team brought up like I, guess, running, I, think Alex. You were working with him on on that, or at least there was some back-and-forth in the glitter chat. I, don't know if I 100 followed everything.

D

B

Have any more info yeah, so it is getting to a point where it's starting to build, but because we're using the open, JDK images right now it seems like they either removed some images like the arm, 30 to be seven or or something similar. It still shows up on docker hub, but when you try to pull it, it doesn't pull correctly.

B

So I haven't had a chance. We just got new silicon at my job, so I this past couple weeks have been pretty rough but I'm hoping to get back and look at it again. Pretty soon are.

D

You guys pulling the the manifest or you guys pulling the specific like arm. You said, 32 /, like open, JDK or so.

B

The way the way the we currently have the multi art stuff set up is we use a ship. It uses a shell script to modify a docker file and then we build that with we build that docker file separately, so I'm sure there's a a different way of doing it, but I'm not as intimately knowledgeable about docker to know if there's a better way right now, no.

D

It's fine yeah, that was in the published experiment, though that's the thing I've been hacking apart right now, yeah.

B

D

Yeah, you guys are pulling from the you know the arm 32 one, not the manifest cuz I, don't say the manifest might have been edited to not include are many more arm 32 at least but I mean, if you're pulling from directly from the arch users, which is I guess with docker hub, like the official docker user site, official, docker hub organization or daughter organization. What they did was I think it's kind of cool is whenever you push two official images, um all the multi arch images or all the arch images actually have users.

D

So if you go on, you've, probably seen it before. If you go on the docker hub UI, you guys you can sort by users and if you go to like s/390 user, those are um all the official images that get pushed in the long test. 390, that's a little weird, though, if I'm 32 is not pulling and it's up there well.

B

Then they say that it's not supported that that image is not supported on arm 32 v7, but the tags and everything are there. So I I don't know if they just push it and it's they don't check whether it works or not. But it says there there's a warning right. That said, this image is not supported on the arm, 22 v7 architecture, but there are tags and he tells you how to pull and and stuff like that. So I just have that chance to go back and look at it.

B

I don't know how many people are using arm. 32 v7 with Jenkins I would assume that the majority of people using it are using with our arm 64, but I could yeah no.

D

Iii think you're right I, even on the fictional images, because when I was hacking apart, the Installer I was I was looking like see what the manifest actually annotate. The architecture has four arms I. There was like three different versions: there's arm thirty-two arm. 64 then I saw like a v8 which seemed like another one: I'm not wanted with a cent share, but arm 64 seems to be the most popular, but it would be awesome to support everything so.

A

Yeah I thought arm. 32 v7 was like my Raspberry Pi 1 1 gig of ram that.

B

Which it depends on which raspberry probably have because I could has buried by three, is 64-bit arm.

A

Okay, so my PI 2 I should be more specific. My PI I thought was an arm 30 to b7 my pi/4 and, and it's limited physically to one gig of ram. My PI 4 has four gig of ram and I'm, pretty sure and I know, is the 64-bit arm. Yeah, ok got it. Yes,.

B

A

Yeah I get the.

B

The PI 2 is, is a 32-bit courrier, correct.

D

You said you're getting some new silicon or a new pie's or whatever, when.

A

Alex says, he's getting new silicon. That means his employer has has just received silicon back for his job. There's nothing to do with this.

B

That's why I haven't people? Look at it. Cuz I've been focused on stuff.

D

A

No Alex Alex works for a company that actually does silicon, which is wildly cool. Basically.

D

Sorry I can I always I have a cold puzzling land I can I could try to bring those into work and and see. If you can air I know, there's that the dreaded air I look out for all the time is I figured I would have to go. Look for it, but I know how to top them headaches. There's always the air I get whenever I try to pull down an image and it's not built for s/390.

D

They comes up with some weird obscure error message and like yep, that's 100% like an arch issue, so you might be running to that where it's me, like you said it's made for its tag, that hey works on arm 32, but in reality it's it's. It's not working yeah.

A

So so it seems the action item Alex, given your workload it it may not happen anytime soon, but the action item is: is we need to investigate further what action should I put here? You.

B

Could put for me to investigate the architecture? Support? Okay, because this we're gonna hit the prep work? I, don't think the like! The open j9 is available for all the platforms either or other architectures. So no.

D

I think arms actually lacking so yeah.

B

So we we need to have a way of kind of filtering out different combinations of variant and architecture. So you can put that on my plate that up till to look into what would is actually supported for the different architectures. Okay.

A

Great all right, so, let's take a doc architecture, support and adding it into the build scripts. Is that what I'm did I phrase that well enough yeah?

A

Okay, great all, right thanks, okay, Oleg noted that his network is not stable enough for him to stay on the call, so we're gonna go and we'll skip this one on the rename process. I think I've seen some progress already happening in various places and discussions, but best to leave that for for his him. To summarize for us, okay, broadening platform support for docker images. Jim, do you want to take this one as well.

A

D

Jim you're, muted sorry, yeah I'm, trying to figure out what we want to talk about that because really those are kind of tightly coupled well.

A

Why don't we just you? So if it's tightly coupled to key mu, let's get to that one, we'll just we're gonna, delete this one and come back to it so Google's. Let me take the google Summer of Code topics at least give a brief overview. So we've I've put the get plugin ideas under the platform sig because of its platforming nature, and so there has the get.

A

Plugin has two ideas: there is an EDA integration ideas which are out there, so EDA tool, integration, ideas, and there are a few others that are like that interesting things that could be could well be interesting to the platform special interest group.

A

And let's see, I should put a lie: should embed a link to the google Summer of Code I'll I'll. Do it separately after the meeting that way, it's in the notes, it's just on Jenkins style, under google, Summer of Code, hey Mark, what what is a EDA?

A

Oh sorry, electronic I should expand acronyms electronic design automation, although this is the the kind of tools that are in Alex's world all day long every day, saying oh I need to design this kind of silicon, or do this, maybe it's physical layout or all sorts of things like that, where you're trying to design either processors or boards they're, all in generally, under this concept of electronic design. Automation, oh.

D

It's really awesome so trying to like get a summer kind of intern. I guess Google code turn to kind of work on that right. Right.

A

That's the concept yeah. It's.

D

A

There and there there are others good and Oleg's, already started.

A

A

Oh right and Korkis, where Korkis, if I remember correctly, is the is the is like growl, it's real. It's compiled the job if I remember right so yeah, that's that's another excellent! Since Oleg can't join us. Let's go ahead to the next topic, so Jim we're up to you qmu and dr. manifest integration. Yep.

D

So since our last meeting, I really only got to work on it like Monday, Tuesday, Wednesday and then a brief brief reef a little bit this morning.

D

So what I've been doing is I've been trying to hack apart in the published experimental as far as I know, that's kind of where you guys are doing the multi arch builds and I've been utilizing. Travis I know we kind of talked about a little bit about Travis in the past, as the CI CD pipeline I built in a way where it's not tied to Travis at all.

D

We could easily just integrate it with Jenkins, but it will be a little more work on in terms of getting agents for the different platforms and stuff, like that. The one nice thing Travis does for us already is: give us access to s/390 power arm 64 I, don't think they do 32 yet or we'll be sure working, 32, I, don't know so empower so which is really nice.

D

Do you guys want me to share my screen, or do you want to click through to look I? Would.

A

I would love to have you share if that's ok, so I'm gonna stop sharing and have you take the share? Ok,.

D

Sweet you guys can see my ID okay, yes, okay, so I had to part the publish, experimental, I love published kinda unchanged. One thing: you'll notice is I made this that CI folder I thought about, and these are all just changes that I made obviously would have to go through all you guys, as you guys know, more about the whole Jenkins ecosystem to me and how you guys wanted to sign your repository.

D

I moved them kind of out of the main repository I see this a lot on a couple of different other popular repos, where all the CI stuff is kind of pushed into a folder away from kinda. You know bulk of the code and kind of keep them separate, so people don't get confused. One thing I did was you'll see. There's is no published experimental anymore. What I did is I broke it up to be a little more modular where you publish the images.

D

First, you publish the tags I'm still working on the manifest I think get all that done, and you kind of have these three different stages and that's important dude. You have Travis kind of works. One of the things I was trying to do is make it as fast as possible so trying to get things going in parallel and I needed to be a little more modular, because before how I understand it, cart race through code, was published. Experimental kind of did to everything you know it.

D

It went through that loop of pulling down versions, I, think I pulls the last five versions loops through room builds the tag. I builds image, tags them and builds a manifest, so I kind of made a little more modular broken up, published images. I mean we can take a deep dive in the code later, but all I'm really doing is the name kind of implies is going through and publishing all the images down here. This is probably Oleg. Sorry Alex what you were talking about where the arm this is the base image that gets replaced.

D

This is I guess where the whole arm 32 in arm 64 come into place in here. I also made some edits to the docker files. What one big change I made was taking out the use of key new headers I went fully focused on building on platform instead of doing a emulation which can lead to some problems.

D

So there was no need for I guess the cross I forget where she or Alex and I were talking about on the glitter chat, but every there was something here for the copying over I think that qu headers are qumu binaries over.

D

So if we go back to published images, we're really just it's supposed to the same code as you guys had I, just kind of hunted together to be a little more modular down the bottom. Now you can pass in the variants wish you guys had. They also can packs in architecture which we'll see just in a second. Why that's important and down here I, clean I, basically removed all the manifests and all the tagging, so just two published images.

D

So it does the whole pro boast tag where it's Jenkins version, Jenkins variants and Genki in architecture. So I saw those. These published images need to have a default tag, so it's posting the most verbose tag.

A

So Jim, what are some yet common common values for variant as opposed to common values for architecture, is architecture like s/390 yep.

D

Yes, so if I, if I show you my travesty mo travesty mo, is how Travis knows how to run the build it's just kind of like a little meta file, and in here we have different build stages.

D

My awesome edited to make file a little bit to have more things, but in here you'll have make I'm using make publish image variant which allows you to supply architecture which, in this instance we're running on on Prem for s/390, so s/390 gets passed in so just build the debian variant of s/390 and whatever jenkins version it happens to be.

D

Does that answer your question mark I think.

A

So yes, thank you. Okay,.

D

Any other questions before I go a little further nope.

A

Please proceed: okay,.

D

So let me, let me show you Travis so make a little more sense.

D

So in Travis we have two stages right now, but more the manifest needs to come still, but in here you can see, build and put published images where yeah have you guys ever played around with Travis before I have.

A

Not so this is interesting too, don't be shy, though I I'm, confident there are others who have okay.

D

So Travis is tightly coupled with github, so, as you can see, I actually have a hub repository. This is my fork of your guys's repository and anytime I push it it's configurable, but anytime I push right now. This would trigger a build. So if I go to like build history, you can see all the times I've committed and it's such the kickoff baylin's. You can change it, so it only kicks off building. So certain branches, like master, so that it's not kicking up, builds all the time.

D

That's usually probably what you want production just to kick off a bit of the master. So really only people push it to match. There would be via PRS. So in here this is the first step which is build in published ninjas you'll, see, there's a bunch of different platforms. We can see kind of barely here. The ones for Debian are the first four right here. You can see the variant right here. The next ones are the Alpine and the next ones. I was slim, those are the three multi arts ones. You guys have right now.

D

I know open j9 we're trying to give that supported right now too, but that wasn't in the folder for multi or images I, just based it off what you guys had currently so right now. What will happen is these are all kick off in parallel and start building the images, and they automatically get pushed to my repository over here and least probably should look familiar to you guys, they're, the same kind of tagging mechanism you guys use and the really nice thing about this. These are all basically spinning up.

D

Vms on s/390, on power on AMD on arm actually see sixty-four. So we don't need to make the use of the cue headers at all, which I think is a big advantage and then, after that, I'm still working on it, but it will go ahead and publish tags, I, split kind of the whole difference between building the images and publishing the tags in terms of basically just speed and publish tags. I can show you guys that script basically lets you do like.

D

Okay, hey publish all the Alpine tags, so those are the the tags that go. You know alpine and then the architecture which just to confirm these ones just point to the latest, build of Alpine right available right. It's.

A

D

Lts spilled or anything like that, it's always the latest one.

A

That's what I would have expected I think that's the naming convention. It's already used that unless, unless the the variant or unless the version is called out, it's assumed to be the latest at that point. Okay,.

E

Good, so I don't.

B

I was gonna say so. One thing to note is the publish experimental is for things that have not been fully tested. Eventually, those items should move into ad normal publish script and.

A

B

It would also do LTS based images yeah.

D

I start up, I started actually adding LTS I. Think that says you one. The things that fail down here, I think one of these were LTS yeah I didn't notice that LTS was not in the experimental, so I kind of added that in in my published tags right at the bottom, like I said it does the whole image variants. Then it does LTS Alpine, so it'll go out and find right here, the LTS version, the latest LTS version and then the LTS slam and then the latest tag which I guess is associated with Debian right.

D

It's always the Debian build which is latest that's.

A

My wreck election is that latest has been that the default and first operating system that was supported was a Debian Vera Alex. Is that consistent for your understanding as well, yeah I believe so? Alright,.

D

Good I just want to make sure that that the kind of default image was always Debian and made it in a way where this is flexible enough. So if down the road we did switch to Debian slim, you basically can just swap them out right here and kind of change. What late is actually points to yeah? So this is this.

D

Basically, as far as I got I guess, the major improvements was really kind of building on on platform and taking advantage of Travis's platform, but, like I said, this is not tightly coupled to Travis at all, I mean you what I imagine what you could do in Jenkins is have Jake agents for s3, 94 PVC and for armed 64 there's one other kind of questions I had was what is what is kind of the whole build infrastructure right now, as it currently stands? What what are you guys doing? Yes, you know kind of trigger.

A

Builds current Kearney, actually I, guess Alex. You can give a better story about current build infrastructure. You're closer than I am I'm comfortable, describing it somewhat how if you want to go ahead, Alex yeah.

B

So, just as a historical understanding, the reason we did the qmu route is because we could not get agents for an infrastructure yeah in order to build on those other platforms, so that that is the reason we went with the QE Mir out. One thing that we I think we have a little bit of a concern with is in order to publish images. We'd have to put credentials for specific repositories on docker hub, and so that was one of the concerns we had about using.

B

Another service was in the Central's outside of our main infrastructure that we control fully but I'm sure that could be revisited. If you know, if we can't get these type of agents, then we're gonna either need to do something else, or you know figure out. If we can pay for some sort of service or something like that, so I mean that's kind of the historical stuff. We only have right now: AM D, 64, based agents, so.

D

B

Yet, and of where we're at right now with infrastructure,.

D

What I can show you how I manage the whole Travis credentials in here? You'll see up the top I know you guys had some sort of login token mechanism, which I'm still trying to integrate, because I knew I just thought. One of the big things I lack right now is checking whether the images are up there. So body like trigger a build which I made some changes to publish images to get latest working right now so with Travis. If I do I had like a the new change, I.

D

Can do push and travis has a github hook, which should.

D

D

Yeah so just triggered a new-build, so right now it's gonna be spinning up workers, these ones just kind of popped into existence. Right here, the the one thing to note, which was a little surprising to me. We have an enterprise, Travis and there's no kind of like throttling mechanism, but in the free Travis it's not giving throttle, that's necessarily in enterprise.

D

All these things we kick off all at once, so you're not kind of waiting around for these workers, but in the free Travis you'll, see that only five booted up and then as soon as one is done, it'll you know trigger the next one, the next one. Next one next side, thanks one this still cuts down on time at arm sixty-four takes the longest takes about like 20 minutes, to build all the images for each variant.

D

So that says she pushes up the build time to about like 40 minutes total doing this way before I had it do like in stages where it was like. Okay, just build Debian, and it would take 24 minutes for the arm 1 and then build Alpine, and it would take 24 minutes. You know so on at 24 minutes 3 times an hour plus just for the ability images, but this gets it down to about 40 minutes, probably doing it all relatively parallel, which is pretty cool but yeah.

D

This is just on the workflow you would do which would Travis you kinda do a PR to say: hey we're now supporting X or you can trigger it via web hooks or you can just come and manually trigger it here. But then it starts kicking off a build. One thing I want to show.

D

You was my secrets up at the top I'm still working on that whole github token thing you guys have and whoever did that I would love to talk more about that, but right now, I, just added a docker law gun method and you might be seen like: okay, hey he's using environment, variable stock or username password I'm trying not to explode my username and password. So what you can do is Travis has is really nice.

D

You are to manage your secrets where it's like: okay, hey you can enter them here, they're, hidden and in the output of all the logs. About the cop in here you'll start, seeing, like you know, just normal kind of like you know, docker builds right up to the top he's doing a docker login somewhere right around here. If I can I'm not belong and right now, it's somewhere in here, but you notice that you'll see secure right here. That's because this is you know: James Crowley IBM is my user.

D

So my tan, we talk our username and also that you know the repository or organization name, so it will actually blank out any uses of passwords for usernames, because I have my secrets managed over here. So that's one nice thing about Travis. They have that kind of built in nice, but I understand the other concern where it's off. You know off Prem, so you don't really I get I, guess Travis could be malicious, but I mean it's pretty widely used.

D

Is that kind of history question on terms of I know? It wasn't really a question but actually kind of at least concerns with the whole passwords and stuff I mean.

B

I've used Travis for another project, I worked on so okay, it's just the other. The other thing is, we always try and eat our own dog food, so use genome yeah, so trying to use Jenkins to build Jenkins to do all the infrastructure that that that's one of the goals so that we verify that all those use cases are handled. Yeah yeah, so you know we have used other services for like I.

B

Think one project use is using app there right now because of the it's a Windows c-sharp based project stuff, so yeah I mean it's definitely a possibility. No.

D

I'm totally that's why I made it flexible, do not be tightly coupled with Travis at all other than you know, I guess in the it's not really tightly cobbled, they just the whole architecture thing pulls from it.

D

This is the only time you're really seeking Travis variables or anything like that, but you would need to get the architecture from you could do it from you name if you really want to, but like we mentioned to mark before, getting access to s/390 and power is no problem in terms of you know giving you guys access to I guess servers, so you can make them into agents I, don't know how you guys necessarily would get arm, maybe like an army of.

A

Alarm boxes arm is actually already available from from Amazon itself. Now. Oh really, AWS now has armament arm machines, I'm, not sure if they're the exact architecture we need, but they just recently announced it so so I'm now we're currently as your base so saying it's on, AWS isn't isn't as much help as you might have thought, but knowing that one of the major cloud providers has already started making machines with our market picture available at least gives hope.

A

D

But anyways I think that's kind of what I want to share is moving away from the team you headers, making things a little more modular in terms of you can run them all at once kind of speed up, build and stuff the one big thing that I was really confused about in the beginning, but this whole process is in the original script. You have this whole concept of getting these latest versions from Jenkins, which I guess tails the last five releases.

D

That seems a little odd to me and I. Don't know if you guys are actively trying to make changes. I'd. Imagine you guys probably want to move to like more like a push model where I guess you see I release on Jenkins, github or up on the website, and then something triggers a build so that way, you're not just building five. At a time you really just building, whoever just got pushed. Is that something you guys be kind of working towards or wanna support, I'm.

A

Not familiar with this particular segment of code and I, don't know if alex is either it may be it deaf. It predates my involvement with looking at these scripts, so my apologies I, don't have an answer for you. Okay,.

D

That's fine, I think.

B

The reason that this was done actually I, don't know why this was done. I was thinking if I'm, pretty good, yeah I think that well, I think this was done, because if the docker image changed and there was an error and the actual docker image rather than the Jenkins itself, they wanted to update the most recent versions that people would be using. Okay,.

D

Okay, yeah I I saw a couple like I guess in the published one right, you're not telling just five I think you tale the last like 20 or something like that, and then you guys have checks in I'm trying to fund somewhere the last 30. You guys have checks in there to basically say hey there's this version already, you know, is it published already, so it would skip that. It just seemed like a lot of redundancy of like okay, hey.

D

We pulled the last 30 versions, let's just loop through maybe let's say 25 and they're all already published and the last five aren't. So let me just go ahead and build the last five would be nice to have more of like a push model where, because that what you guys is like lag time where, like say, Jenkins dropped. Something.

D

Are you guys immediately building that newest version or LTS that Jake is just dropped or is that even a goal for you guys to kind of stay relatively in sync of like hey new jake has just dropped. Let me just kick off a build to kind of build the docker image for that, so.

B

For the official repository it is built on a time basis, and so that's another reason this check is in because we the way that the current early set up is done. It's it's not fully automated. We have a project that is fully automating that release and we would be able to kick off.

B

Docker builds and stuff like that. But currently it's it's it's not that way. So that's another reason it's a time-based thing and that it has to go pull the most recent versions. Okay,.

D

All right, thank you. I was a little confusion. I guess you probably do want to wait a little bit anyways to make sure everything's stable and all that right. No any major bugs.

A

Or actually that that is not typically an issue in terms of concern. The typical is we want. We want the docker images very very soon after an LTS or a weekly as available, but yeah alex is right that the initial release of the Jenkins version, so two dot 221 as an example, is generated at an unpredictable time and done right now by a human being Kousaka because of the code signing.

D

Okay, all right, so you see I, just those were like the little you know, kind of areas. I was a little confused about going in and I didn't know if it was something you guys were working towards or there's other little gacho's, like you guys kind of mentioned so I mean that's that's kind of it.

D

From my the new code and I like after talking with you guys, I, don't know how useful it would be since you started the whole problem of not getting not having all the architectures I guess I could add back a little team. You stuff, but I, think I. Think moving the building on all architectures would be a good idea, but maybe that isn't practical right now, at least well.

A

I think it's an excellent question. It's when we need to hoist into the infrastructure team to see all right. Are we at a point where we could, where we could turn on other agents in the in the CIO Jenkins that I/o infrastructure and you noted that there are several different, several providers that might be willing to provide access to s3, 90 and PowerPC 64.

B

It looks like on a sure they do have arm 32, v7 and arm 64, but know as 390 or a PowerPC. Oh cool.

A

Okay, so the arm so as you're as you're very good, all right. So that's really encoding, encouraging I. Don't.

B

Know what their full support is, but it looks like you can create vm's and stuff on those architectures I mentioned so Oh. Was it arm 32 and.

D

64, just a support both.

B

32 and 64 Oh awesome so that could at least do part of what we're looking for and then we could maybe do Travis for the s/390 and PowerPC I think we.

D

Have resources to like give you guys access to a beefy VM, and then you guys can make that into an agent. If you guys want so I mean you sound, it sounds like I mean it's, it's a sounds like we almost kind of got the platforms right, if you guys are able to at least utilize the whole adjure shutup forearm yeah.

B

Look I'll look into it further and do a few tests and stuff like that, so those are publicly facing servers. Yeah.

D

Yeah, it could be set up to their hosted at Marist College, it's part of the Linux one community cloud and for vendors. We work with or open source projects.

D

We can basically give a long-term box to you guys because, right now how the community cloud setup is I, think you have a 90 day, trial or 180 day trial, I'm, not trying I, don't think you can buy anything there by a 180 day like limit, but for I get. You know like I, said vendors or open source communities. We can give access for a long term a cak indefinitely. So.

A

Thank you very much Jim. Thank you. Thank you. So, in terms of action items, it looks like you're going to continue your evolution of the continued evolving. The publish scripts yeah.

D

I need to refine it a little bit and then finish the manifest steps and then I mean I'll keep making it. You know flexible where we can use jenkin agents. You know I mean where you basically just you know pushing a workload to another machine. I won't tell you couple it with Travis, so I'll give you guys that flexibility, so I, just don't I'm gonna continue testing on Travis, because I don't have access to you guys, built by fine. So.

A

Alex is that reasonable to you, you, okay with that.

B

Yeah, any improvement is a positive I. Think yeah.

A

Thanks very much Jim, that's great. We have. We have actually hit our time. I was I'm. I had one item remaining that I'm in a skip. We will defer to another time unless there's some urgent thing that we've got beyond that I'd like to call for an end for today's session. Any other topics we need to be sure we recovered today.

A

All right thanks, everybody Alex! Thank you very much Jim. Thank you. Thank you. What great great results, it's so cool to see as 390 and and PowerPC. That's! That's! Really! Wonderful, thanks, very, very much and we'll meet again in two weeks. All.

D

Right, thank you for. Let me share thanks.