GitLab Release Stage, 19 May 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Auto Deploy Handover part 2

Description

Second part of Auto DevOps handover from Configure to Release

Mostly discussed Review Apps related to ECS

A

Okay, so we have. This is a second meeting on the handover. Subject: I, don't really have something formal to talk about I, don't know if someone added something to the agenda because a little bit my bad a little bit unorganized everything in the previous meeting is on the top, but I do want to talk about some things that came up during the auto deploy to UCS issue now so I think that's a good place to start. Unless anyone else wants to take leave and talk about anything.

A

B

That yeah I mean I, think that's a good start. I think there might be a couple things that I wanted to bring up, but yeah take it away. Okay,.

A

So I'm gonna put it on the spot, because auto deploy to ECS has been merged and great I'm very happy about that. There were two associated issues with it. One was separating the kubernetes template into a standalone template that would be called from auto deploy and there was another issue related to the deployed target. So we wanted to introduce a variable called deploy, target or something else I'm not stuck on the name. I could be persuaded with anything which we had some external discussions and I.

A

Don't know if everything was done in the one issue at in or if yeah and those need to be closed out or if those are still pending. So I'm gonna, give you put you on the spot.

C

Because, yesterday it was a day off in Canada and so I need to catch up on emails, so yeah I just wanted to go like yeah what you just said. So to answer your first question. Yes, it was all done in one issue but, as I said, I need to catch up on emails and in related to Lodi everything related to like which issue what mrs okay, that's something I can follow up with you afterwards after this meeting. If that's okay with you, okay.

A

So what I did see another deploy? Yes after being merged, that there's two new issues that were brought up from the review from the quick review, so one of them was that we need to deal with stop review, apps 3cs, yes,.

C

That's correct, that's something that came up in conversation was like. This is something that's you know that's completely different than compare. It is because what does a stop review at means? Do we turn off the service, the ECA service? Do we turn off the tasks? You know it's just like from an SES point of view. What do we do exactly right? What is the you know? What is the part of the NWACC s API that we want to leverage.

A

So that's actually a very good question, and, and now I'm gonna put you on the spot Shinya. Do you think it's a good idea to introduce our auto expire for review absolute for this process? Yeah.

D

I mean before talking about all this expiration. We should answer to the Etienne's question yeah. If his question is answered code, expression or whatever things is possible,.

A

Okay, so I'm gonna ask your answer your question again with another question.

A

So when we deploy yes, how exactly are we doing the review app itself? Are we opening a new instance on AWS, with with an environment variable that someone put there.

A

Remind me what the environment variable for review obstacles.

C

Change so much, you know what let's go to Target Depot for like target exchange.

E

C

Times over the past few days is that I cannot remember, I, think it was also the verbs platform target so.

A

C

We went for I'm.

A

In the dynamic URL, further review app itself, where you, you have an environment variable.

A

Damn it what's its name: well, the the review, app environment, so that's a different environment than the actual environment. That's being deployed, tecs, correct, mm-hmm,.

C

Well, no, thank you. The rijal are what we see as stilled approach to ETS and they still need to provide a service name, an easier service name, a task definition name and across the name right. So the the requirements for review apps are the same as for productions as for like as if you were deploying to production, you still need to provide those three variables in addition to the regular AWS credentials.

C

E

A

C

It that's why that's how it was. You know.

A

Maybe cuz I think we yeah.

C

We created deployment to us yes at first, just like you know same way for if you and production, but.

E

C

Are starting to raise that you know, maybe there is like a divergence at this point. Maybe we need to create on the fly, a service or a cluster for review, why we want to have why we want to provide underman variables or the production.

A

Environment, maybe there is a different strategy right.

C

To create for for review verbs but.

A

As of today, to answer your initial.

C

Question you need to specify a testability and cluster and service for review. Just like you need to do the same for production, okay,.

A

So, if that's the case and the the user needs to supply the test, to finish your name, the cluster name, everything we need for UCS separate for review, ask for production I, don't see any reason why we can't tear down the service once we don't need the review app and you work one. The deployment is dead, yeah.

C

There is no no reason.

D

Unless I don't know.

C

At least maybe the user wants to turn it off for like competing purposes. You know or.

D

C

I think we should still provide that stop review option and therefore that means you know back to the very initial question. What does it mean from a flow point of view right? What do we do? What do we turn off? The the EWS ccs api is very, it's pretty rich in you know in what it has to offer. So we need to chill like what to leverage.

A

What are our options? I.

C

Talked about that with someone else with someone last last week, I need to go back to to that NWA.

A

C

Documentation, let's see.

F

So if so, the default for environment variables is that their scope to all environments. So if the user has only set one of these tasks, definitions and service definitions- and yes, sorry, I, don't really know much about is yes.

F

So, if said, all these variables at the top like at the star scope, is still is deleting the service still safe. I. Think.

C

The production will take over because it's been defined afterwards. Yes,.

E

C

You're raising a good point. Users have to to define those variables, but.

E

It has to be ambulance basically right. No, that is not evil.

F

Yeah, so that's kind of we don't really have a way to tell once we're in the job if they were environment, specific or not know.

C

That I mean to make that differentiation. It has to be another iteration. You know on top of what what has been done.

A

So here's my dilemma: I think that once something is merged into production, we know long no longer needs the review and, however, if we're continuing development and we haven't merged something into master, yet we would still want to view our changes in a review app just like any other deployment.

A

Regardless TCS, so at any given time, we should have a review app available. We do need to explain to the users that this may have additional billing costs, which is not like our usual review. Apps I.

C

Don't know that being the strategy for kubernetes, but that would be the same problem right. That would be the same situation. Wouldn't it like.

A

If I, if the user.

C

To leave their their review apps on running on communities, they're gonna be built right.

A

Specifically for g and g ke r, like different, you sphere, o GC,.

C

Gc p in general, I would I already don't know the meeting structure for GC p, but I believe.

A

C

A

C

A review up there, you know in a class they're.

A

Gonna be billed accordingly. Well, that's why we have our experts of auto that on the line and they can tell.

F

If billing is something we don't really think about much, but I think we yeah, we should but I think it's we. We are just charged for the nodes so like, because you have a cluster there and that the expensive part is the VMS. That hosts note, so I think that this, the things you have in the cluster don't cost extra.

F

But it's still like you, you still want to clean them up because they take up space and they're like they're they're, expensive in a different way and yeah you, your notes, maybe we're provisioned in a way that, like CPU cycles, counts, like you know, used cycle, counts, I, don't know! This is something that we never really got into a deep yeah.

G

And do we need to write like we don't want to be like another? We don't want to have to another billing dashboard or something for the user, bring it like. What's.

C

It or cc where's Kristin, are you as we're not providing stop reviewer a stopped review. Charles I, don't know much user can still go on there. Aws console des.

C

A

Me never let me create some sort.

C

Of like discrepancies between what we're doing you know, good luck managed yeah. What are we doing on outside and then you know.

E

If they're doing things.

A

Independently on their side from.

C

The AWS console, who.

E

Knows right if we are persisting a few things validity to the ECS deployment.

C

Just like will persist in.

E

C

E

C

Our side, we're gonna, create that discrepancy. There will delete things on the ACS yeah the AWS console, while there will be some related apportion still persisting in our database I. Don't.

B

Think we yeah I'll just just just one in a voice that we've definitely been bitten by that discrepancy. Previous with our eks and gke integrations for sure. So whatever you can do to try to maintain the parity or do the you know, kind of checking to make sure that the node still exists or that it's been deleted. Have those kind of callback mechanisms really recommend trying to you know make sure that that's implement.

C

G

Wait you, like the only parody.

C

I think job nothing.

G

C

Like every night, or something like that,.

F

Yeah I don't think we do any any specific checking, except when the user visits the page like different Cuban. Eddie's pages like you would see, requests there so that if a cluster has been removed, oh you mean, but for the review yeah we we don't do any any specific checks, I think any cron job. That would do something.

C

F

No and then like we don't auto clean review up. So let's, let's be clear about that. We have a manual job because, exactly like you know, you want to keep it there, while you're working on it.

G

There's a mechanism to get to one flow where I can see having a review up after say a features been merged. You no longer need it for your particular purposes, but let's say QA comes along like hey I, wouldn't see this thing, you know or something along those lines, maybe later, maybe after the new after a flow like, but that might be in like a separate staging branch or something else. So maybe that's kind of it doesn't apply. I'm just I'm not know I'm thinking a lot in front of the group. Sorry yeah the.

A

Skimming really recover, they WMS documentation, which obviously I haven't like dug deep into it. It looks like we have stop and start instances they'll, maybe stopping and instance, as a review. App could be a good first step.

A

I, don't know if that has any effect on budget, but but it seems like a good first step and then theoretically changed for your case. It could be restarted if someone sees a problem.

A

A

And that's way before termination.

C

I feel like if we're, if you are interesting to work on this.

A

E

C

Process I think.

E

You like a following.

A

C

Right after that was something that's quite connected is something we mentioned before is the fact that we're using the same variables same three variables for review and production philippic. That.

E

C

Part of the same, not issues same epic or you.

E

Know so something to consider at the same time,.

E

A

I, remember it's a slum. It's an in slug, something CI environment. Slug is the review. So are we just automatically appending that, on the same, in fact, very room that the trip reduction, I.

C

Remember correctly for review we're doing that, we're appending the stripe or something similar.

E

C

No, no, no just put.

A

The a moment we know we're.

C

Not doing that just.

A

C

The review apps, we are appending the the commit ref, but it's.

E

Just at the end of yum urinating.

C

A

Well, that would work, but that would mean that there will be a lot of reveal apps at any given time which may explode in terms of budget.

A

Am I understanding it correctly.

C

Trying to I'm trying to do it to agree I'm trying to understand. What's here, I mean to agree with what you are saying. What.

D

C

Would it what would it be yeah? Would it be the case as.

A

Long as if there we have lots of commits going on yeah.

C

A

Single merger schedule, so many of you yeah.

C

That's correct, yeah, that's correct yeah and if we're using not.

A

Static in what we're using the come in come in as the pending and it's limitless and we can have many many time. Yeah.

C

Well, yeah yeah.

A

In terms of kubernetes, I assume that.

F

We can have many review ups at the same time, yeah any as many as you want, as many as you're close to can hold it's.

A

F

No I think I think it's a feature to be able to have as many concurrent review apps as you want. It's it's a mainly that you know it's really really easy to start them, so we should provide at least like documentation. Note somewhere that you know you should we don't provide a way to clean them up.

A

Yes, oh I, agree, I think maybe at least for this phase, we don't need to limit anything. We have several issues that need to address cleaning up review apps, so one of them is, as we mentioned on the beginning of the call, is set an auto expiry date. So it's like no activity is done. It just stopped for you up and stopped in the sense of VCS will be a little bit different because we need to stop the service and not just stop.

A

The review have as we do in any other case, and then we had some other ways like you can select in the UI, the review, ass and the environment page and just stop them which will set the expiry and actually stop them after a given period of time, and we also have an API that allows you to so I think in that has recovered, because there's different ways to delete them.

C

Yeah I cannot I.

D

C

Answer this question because I don't know how the way we did what we are going to do. It goes back to like what Shinya was saying just like we need to choose the art strategy to you know what.

A

Does it mean from this point.

C

A

Delete a review up or.

C

To stop a review up I think that's how much that's.

A

Right I think this is the best thing that we can start with, stop and start and stop, and we have there's there's a CLI come in. That probably does that here's, the.

C

Documentation for.

A

Wrapping the instance is the best way to go, and then we can go back to our regular behavior in terms of review EPS, which is set an expiry which will actually stop the review. App Shinya and I had another call on this unrelated to ACS, but the fact that there's a review app that doesn't do anything doesn't really do anything. Just it's just the entry in the database.

A

So there's no harm done.

C

D

C

You be sharing a chat, the pages just sure when.

A

They're uws one yeah.

C

Yeah, it's like it's something. That's that's a bit different than yeah WACC,.

E

Has documentation.

C

A

Okay, so I think in terms of what we want I'm just going to recap: the decisions, our review up, so there's no limit to review out having you review ups we can have and the way that we're doing it. What by appending the commits reference is fine, no need to change. That's what we will do is in case of expiration date or the mr is merged. We will send send a stop instance TCS, okay, in the future.

A

We can think about how we can improve this, and you know if someone has an X amount of review apps available, we can send them some kind of notification. Hey watch your budget or something like that, but we can think about that in the future.

A

And the second issue that came out of the review from auto the point ACS check it out. Oh, it was the need to handle the deployment target when we want to do kubernetes. Plus you see us at the moment. We support either or we don't support both and, of course, there will be a day in the future where we will need to support. Oh I believe.

C

More and more that's in order to do so, you know we need a UI.

C

We need the UI that says like in a way in a way that okay, if we want to like choose mark, was talking about that right. Yeah I may want for review to aim that algorithm for that for pollution. All that work and yeah, of course, I think we could. We could just.

A

Like drive, that with.

C

Cr variables or whatnot.

A

C

I feel like which is neither UI with, like you know, different with the targets that are available to the project and the different and the element.

E

And then just like say this target that I'm.

C

Young man that I get that I just want.

A

C

I, don't want to fall into the trap of like relying too much on things like that's it so.

A

I'm, a big believer in that, but I also think that our persona is a lot of DevOps engineers who don't really like UI. So we need to think about them as well. Ui is great because it's very visual and you can get a lot of warnings and notifications, but I do think. We need to think about people who want to do it as a script. Worm yeah.

C

I mean you, I use one side of a one side of the coin right and also think.

E

A

A pea, I write.

C

A

C

E

More relevant to the person.

C

That you are describing.

A

Okay, so going back to our favorite subject of naming.

A

So right now we're talking about deployment targets right, deploy, targets.

C

E

Sorry I did that on Friday.

C

But I need to get back to it.

C

A

Underscore DevOps.

C

Underscore platform at its core targets, while Mark was saying that you know there was there.

E

Was no such thing as.

C

A best practice when it comes up like using a prefix.

A

C

Tom said the opposite, and then he made a comment somewhere I need to find a threat against like, but he said that, yes, they have started. You know someone or group of people started like perfect, seeing some variables with auto devotes, but I. Don't remember what you know. What was the motivation behind that? I need to get that's like thread so.

D

Yeah they were, they.

E

D

C

There was a case of you know using as a prefix, so.

F

The main reason for using a prefix for for something like all the DevOps is, if the like, so firstly just to do namespace the features that belong to all the DevOps for others to to like the clarity like when people come, and they give us a piece of fun environment variable for debugging stuff. It helps a lot, but also some of the words are just really generic like platform targets can mean so many things.

F

So if you put that as a top-level CI variable and you subject, all the you know all of the use of good luck to that as a reserved feature for Auto, DevOps or or four or four like this semi nice feature of deploy targets, I mean like we have. We have a fair amount of users but like compared to the total users of CI, its small, and so reserving that for something like you know something that can mean architecture operating system or you know any other thing.

F

The reserve that, for you know, cloud deployments is like you know something to just think about, so not something to just rush. But ya know it's that's the main reason why we we we prefix things so even just prefix it with CI platform target would have would have been enough, I think. But if you prefix it with cin, you call it see a platform target and you remember the association with operating system and architecture.

F

Then the name is actually kind of confusing because it makes you think Oh am I selecting the architecture that my runners are gonna use. Maybe.

A

So I wanna I want to pop in a question.

E

A

Think that this use of the deployment target or whatever we want to call it, is a temporary solution and I think it's great for our first iteration, because right now we're asking the users okay tell us where he wants to play tunes at kubernetes. Is that you see us is the GCP whatever whatever italy in the future. I would prefer for us to detect it automatically whether or not we can query AWS. It's off Eli's telling us here's the here's. The deployment target I actually found the actual call that that it can return kubernetes.

A

We are ready to text by ourselves and I assume any other cloud provider has similar way to query where it's going to be landing so really really far in the future down the line. I would like us to detect it automatically, and maybe that means just populating this variable behind the scenes for the user and not asking for it.

A

So in the long run, I think it'll be less confusing. Oh yeah.

F

Populating populating the name automatically like so so that that's the forward like that's why, starting with a CI variable, is also a forwards compatible approach, because because you can start with it like this and then you know move towards populating it automatically. What was the other one? You said: I had another thing to say, but I forget.

A

No, please I can't hear myself something.

B

Yeah I guess I mean to me like I, feel like diploma target, is actually a pretty good naming scheme like just that, as a name that makes sense to me. I think that the the auto-detection would be really cool, though I think you ever mentioning to build a know, just innately where, where it's going, the spirits merely user from having to include additional inputs, it sounds great, but it's.

A

Definitely somewhere.

A

Do we have any concerns about the other composable Auto DevOps things connecting to ECS? Besides the review apps.

F

Like like dust and stuff like that, so vast does deploy a specific environment on like when you're on the default branch. This was added recently so like in the last year.

F

I think so, then that job would need to be port it if you want parity, so I would need to be ported to to to EECS, but I would say it's also like not really MVC at all to have a default branch test, so so that what that means is that you know when you merge the master, there will be a job that runs so just to be sure that you didn't, you know, add a bug that somehow got missed in in the review phase.

F

Oh and there's this Cuban Eddie's active condition that is used on browser performance testing that should be switched to something else. We're using that as a kind of like a way to to say. Oh all, the DevOps you know, browser thing is like has a deployment like there's a proxy, but it's not really like that. One is definitely a Cuban Eddie's agnostic, so that one should be switched and you would and then you would get it for free.

F

So probably one of the first things I would start with as a follow-up would be to populate this deployment target populate the kubernetes variable as well, and you can just do it right next to where she, a Cuban, Eddie's activist populated.

C

The motivation just behind that the motivation behind getting rid of the kubernetes I think it's just to make the whole life of the divorce more about mistake. What.

F

C

F

So so I would we can't really well. We just introduced the CI Cuban Eddie's active in 30 know. So we can, you could say it's removing it right away is almost like it never existed, but so, but I think we should keep the variable for now and but stop relying on it in jobs that are agnostic and use. The presence of the deployed like the platform target, for example, as a as an indicator. Instead.

C

A

Looking at the dust scrape.

C

That moment the does default brush deployed.

E

That's the one, that's a very Covenant.

C

Is focused there.

E

Is the auto deploy.

C

Comment: that's the script. You were showing me the other day. That's the Lord, the big fashion script that you worked on right. It was Auto doc, deploy exactly yeah. How would you see these being translated? Oh, like you know, whatever is being done in the script? How would you see these being translated into something? That's that's good. Okay. We.

E

Have like lots of things that are.

C

That could be translated, I, don't know exactly what download chart does, or it ain't your namespace. But to me it feels like it's not something. That's really like you run it right.

F

Yeah so I think that Auto deploy like the steps that are there the reason it's it's broken down into these steps still in the CI jobs, just partly just because it used to be like that, but it's also to just give users on insight into the cut like different sub comment, so they can compose this. But if we want them more generalizable like strategy, we could we just have it or saying oh command like these. These steps don't actually have to be broken down.

F

There's no reason for it like functionally so I think if you did the ECS one it would just be. You would probably just that. Well, you replace the script part pretty much.

C

Yeah as far as I can see so far as far as I can see today, the equivalent of this it's just to register.

A

C

New task definition right the way ccs works. It uses like a task definition, which is which are in nutshell, specs right, specs for the entire environment. We need to. We, we.

A

C

Task definition, so that we are pointing to the new.

G

C

Of the docker application image right there, because I'm not with it going yeah.

E

When I see this list of auto deploy.

C

Comments what I think of the transition into ECS, I, think of life I'm.

E

C

The test technician right now.

F

C

Like next step is to see what else needs to be done? Yes,.

F

So, like the this list of commands, it's very helpful to just think of them all as just Auto deploy deploy like that's really that that's how I would think about it.

E

F

That that that should be our target interface, if we want to share the interface between platforms, maybe auto, deploy, deploy and persist environment URLs! That's like good luck, generic enough, because you still want to persist the environment URL, if possible, yeah I actually did notice that they're, not an ACS I hadn't thought of that before. So that's, probably another post and we see I think to look into but get one thing about this default branch that lik I think you need to solve what how to clean up before. Even thinking about that default branch. Oh.

C

Yeah yeah because.

F

That one is auto clean and.

C

If you like, you know what we grab the whole like a review. Cleanup yeah we'll have like more tools to work on dust, yeah, yeah.

F

Yeah, the test on video apps should already be possible, like that's how we do test on merge request, that's not a separate deployment. They target the review app for that branch.

E

C

When I see the dust and Amanda and the spot deploy job I can see, I can see the IBM and ul that points to a dust and then project CI a product ID. So I don't understand. It looks like you're yeah.

F

There are two test jobs actually yeah.

C

F

Of them is the one on master which I think don't target like. Don't don't think about that are all like. You know until you figure it out clean up and stuff, but there is vast that runs out side of master, which is just called a study lab see. I am oh, that's the original one, the one that's been there for a long time and that one you can.

F

You should be able to just get working right away. Yeah.

C

F

Long as long as you have the environment URL, you should be able to get that I think it just needs to be able to call your application.

E

C

C

After the ring you up, yeah.

F

Well, yeah this one, this one like this one depends on just having an environment URL. It doesn't depend on having clean up or anything, because this is just a CI job that that hits your environment with security scanning.

C

I have a question about MB emanuelle in general, it's always been a bit confusing for me how it was possible to have a URL, that's like set with like a very particular pattern that includes project a project IDs, like you know, the cube request based, oh man, so.

E

A

Something that's that.

C

It's a neural, that's dynamically formed right, but then you pass on this timeline. Dynamically formed URL on to the you know the GCP api and then it's some sort of override it does. It does like how does it work from from GCP GCP is going to like use that URL or IE just like how does like GCP and this dynamic URL formed? How do they correct together, yeah.

F

So if you let's say you may be open, Caleb C I'll share my screen just for context for every.

C

F

Sorry yeah, so it's it! This is the deploy, template and the way it works is we have this cube, ingress based domain that we assume so all of all of this environment. Url magic works into the assumption that.

E

F

User has set up a wild card base domain, okay, yeah.

E

F

So this is like we assume you set up a wild card base domain and that we are able to do whatever we want to underneath there. So the CI project, ID environments like this, is just to avoid collisions.

F

We could we could let them do other things, and then there are extra. We also give them the ability to provision extra endpoints and in order to helps itself, but this is the primary one, I think for production. We don't actually even have do. We have.

F

Okay, we took, we still have yeah there, it's just the CI project passed like so it's the same same idea. We just assumed that you have the wild card and that that's all like so III, don't know. I presume the same should be possible for each yes, but maybe you would have to provision the wild card domain in a completely different service, but, like you could still, you could still do this like this. Is this isn't really cuber Nettie's dependent at all? This is just like.

F

We assume that, because you're deploying a web app that you will want to have something like this and.

C

C

D

Domain, that's.

F

C

C

F

So ingress I'm not too familiar honestly with details of the ingress but like, but there's we don't create.

F

So it does create a service endpoint that that will respond to this URL and then it's up to you to actually be in control of that URL like when I, when I'm lazy and I don't want to install all the required things for all of their webs I just said to example.com.

F

So it's not a required requirement to have something that you control.

F

So I think but I think this is the approach you would probably want to go as well like for just because you know I either either this base domain with with subdomains or path based routing to different apps, but that's kind of less clean, because then your application needs to know that it's routed at a different path. Yes, it's not quite as clean where's. This one is kind of at the inker first level. You you bind to just the domains that you're given or you don't even have to bind to them.

F

You can bind to just any IP or any host. So this is a bit cleaner for, like Auto for the auto purposes of things.

C

E

A

C

Screen to the to show you why you know where am I worried, you know my questions can be found. My screen quickly.

C

So this is the the idea IWA son is.

E

C

Panel, where have where I have my cluster my service, as well as my task recognition, based on what I've read and learned when sending up my ECS tasks for testing purposes. The only way to to.

A

C

A

Urls to go to the.

C

Ec2 be easy to instance, that that is attached to your ACS instance. You go to you ec2 and each instance is linked to a load balancer. So you go to the load balancer that you have set up previously, and- and this is here you know- rails, dog, res docker, alb and so on. So on so on.

C

This is like the this is the URL to hit right so on one on one hand, I have this URL, that's defined by the load balancer that I've created in its you know in situ, and then, on the other hand, I think this dynamically generated, URL I just want to reconcile both of them. Maybe there is a way to override that that URL here you know some sort of like link. We run out of a cname record from a DNS point of view. So I don't know yeah.

F

I think you don't need to specify it the way we've seen the way we do it in the in the CI mo I. Think it's enough for you to just out with the artifact. So if you have, if you haven't like an API call, you can do it too. Just like fetch this. That is probably enough.

E

E

C

Have any other questions so far.

B

Okay, yeah I- guess maybe I'll click over the my item here. There's this a proposal for scripting the auto dev ops release process. Since you right now, it's it's a little bit of a manual process to do this, so Tong is proposed to create a set of scripts to automate it, and the goal would obviously be to get the entire process into code. So you can see how it works and then also that you know either team would be able to to run it or, if we're doing the full handover.

B

That it'd be really clear and hopefully easier for the release process for you to take it over. So it's a proposal right now to get that into code I, just.

C

Was like yes, creating all this release manually every time, every single time that something was hitting master right, that's not convenient there is this one keyword see like the kid. Let's see, I am all keyword called release, which is what we want and may.

A

I need to read that.

C

Unfortunately, that wienies keyword was has been disabled for a while or behind the feature flag. That said, Sean Carroll was working a tree art in aping. It and I think it was part of 13.0, so whatever now I think he must be. It should be available, if not maybe for the next three days, but that's something we should involve either Sean or Jackie.

B

Okay, yeah I, think I mean feel free to comment on the issue. If you have ideas or a counterproposal, definitely you know, what's what's in caps like that, the issue and I can move it forward. There it'd be great. Thank you.

C

B

It should be yeah, it's you know we kind of have our yeah, it's so under May 19th and then at the bottom of that part of it. I have my yeah here. I'll put it in put in chat as well.

B

F

So just a quick note about the environment, URL, so actually I looked and it doesn't seem like the artifact is see I like she I picks up on that at all. Let that's just a way to pass it between jobs, but this is a way if we don't have a way to retreat like preset it, then this is a way forward to support. Reading an artifact like a URL from an artifact. You.

F

Just you'd have to add the artifact parsing yourself.

A

But just from the little bits that I comprehend, maybe we can leverage the damage that we introduced that allowed you to pass things between jobs.

A

It's a report tied between jobs, so if it's like dynamically created, we can get it, but I think we need another discussion on this topic anyway.

A

I have a kind of silly question unless anyone has something additional to say.

B

Okay, so I've got more I've, got more to say but yeah just just yeah I'd love to hear you had a question so.

A

When you get started as a user with a lot of us, it's usually when you don't have anything set up your bye-bye and you get a button that says, enable auto the laughs and then the magic happens. Does anything need to change at that point? In order to the point is yes,.

C

Ci in variables needs to be said.

C

The process will kick off if the auto DevOps platform targets CI variable is set, it will go all the way to either to.

E

C

A production step, jobs, sorry, but the deployment we fail and there will be an error message saying different.

E

Like deployment.

C

Fail, please enter your AWS credentials as a variable, so.

A

C

A bit of an error handling whenever those credential variables are missing. I.

A

Actually think, but I might not be correct that when you start off with other DevOps, it tells you that you need to go to and set the kubernetes data in order to get started. I think that's I changed.

C

The message as part of the auto DevOps to is yes to include.

C

Less I feel like.

E

C

It's not, you know, it's not a fully. You know.

A

C

Great job, it's just only a basic the the message, but I feel like needs. It more needs to be done here at this bar. You know in that section is not also the work section now.

A

We have Dimitri online, something he can take a look in seaman.

A

You wanted to say something: a nice and I bothered you. Oh no.

B

No, no, no, it's totally fine, no I, yeah I definitely think this has been great terms of kind of the ability, for you know, to kind of share like at the technical level, I think at the process, level of the handover. I think we might still kind of be getting there. I guess: I guess is there any? Is there like a standard process of like handing over product areas? We just kind of come up with our own checklist like okay, like this?

B

Is these are the areas that we need to hand over, and this is the time line. I guess anyone have any ideas.

A

Sorry, then, are you going I.

C

Was gonna say like based on what was said, I feel like we just need to have like you know what we want to tackle and there.

A

Is clearly a priority.

C

Of things, first, like the the one we want to leverage.

A

C

Of the WACC, a SES API to either stop the task, go start the service or stop the service, or you know so on that, then we can work on the stuff with you up and then.

E

We can work on that, but then we won't.

C

Wait so once we have, this is priority. This list, with pro ati ten set up I think we can just like you know, follow up on the handover. That's from my point of view.

A

So um as far as I know, there's no handbook page that talks about handover between groups so I think we pretty much needs to figure it out. What I would like to see I wouldn't want to just set a date. I mean I, think we need to target to a specific day when we started talking about the handover. We talked about two to three milestones.

A

What's important for me is that the team feels comfortable to take on the responsibility, regardless of how many milestones it takes, and you know, since we started being active in this area and we started developing issues so you can see. This is why we started coming up with questions and things and I think that's really important that we don't take the step off of the gas from everything that we want to do, for expanding Auto, deploy I, don't want to go into auto review, apps or anything else.

A

At this moment, just focus on potty deploy, I feel like it's huge enough on its own I. Do think if we can set up a checklist. That would be great because then we would know we covered not sure we're, never gonna cover everything, but at least we've covered them the basics and we have another hand over that we're doing to verify for emerge trade and what I wanted. What I wanted to do there and I think is a good practice here is that we start doing the code reviews together.

A

So if someone like a community contribution goes in, someone else can start being a maintainer there and then slowly we can learn the ropes.

B

Yeah I definitely think that would be a good first step that we get.

B

You know like chignon etn as being trainee, maintainer zon, these projects, the mentorship you kind of have it seems like a really good, first step for sure I.

B

Guess I guess my other question was just based on this, based on like what you're hearing about kind of all the different components and things like that is it? Is it affecting your decision of wanting to own all of it or are you you know, because it's it's kind of its kind of a lot there's a lot. This community's focused as well so I'm just curious.

B

If, if, if your interest is more in the east, yes and kind of expanding, that and kind of the different deployment targets or if it's like owning the entire thing, I think we're more flexible on that and want to support, be collaborative so just curious. If what.

A

You're thinking.

B

A

Have a good answer for this moment, but I do believe that there should be like one specific group, that's DRI and we do want to expand on this, to lots of targets and in the in the past, Victor /, Daniel, / I were very confused about. Where are the issues need to go and you can see like a bunch of issues were like? Is this yours as it's mine? Since yours is mine, and previously it was like okay, if it's kubernetes, it goes to configure. Everything else goes to us and I.

A

Don't know if that's a great way to continue, because it's very confusing so I do think that at some point we would like to take full ownership of auto deploy, but have you as advisors, if that makes sense,.

B

Okay, yeah I mean I, definitely like I, wouldn't want you to get in over your head, initially so I think staying on to kind of support or would make a lot of sense for sure I mean it will make you aware as well that we are kind of going to we're in the process of building a completely different way of deploying to kubernetes, so that will probably be where we will focus generally in the future.

B

So I think this makes sense to hand this over.

A

In terms of what I'm gonna just everyone, what our next steps are, do you feel like? We need another call next week, or should we spin out a little bit more, maybe in two weeks like how what does the team feel like they need.

C

E

Two two weeks from.

C

F

Think we can get on calls as needed if like, if there's any specific things, that we can't cash out asynchronously, but yet do we.

B

Okay, yeah and then I think in terms of an action item getting folks as training maintainer zhan. All the project seems like a good thing to do, which is something I can do so. I can take that on great.

A

So, thank you. Everyone for joining and I guess I'll see you in two weeks.