GitLab Backend Pairing, 2 Aug 2023

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: BE Pairing - 2023-08-02 EMEA/APAC - Refactoring a callback into a service

Description

Discussing the newly-merge styleguide recommendation to avoid Rails callbacks: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/125836

Specifically as it relates to an in-progress MR that adds a callback: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/127497#note_1489574714

We refactored this callback into a service class, draft MR: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/128215

A

um All right, so you you have to use a root group with custom roles, enabled what is this I'm already looking at.

B

Let me show you my shoulders: nothing.

B

You don't want to see on my.

C

B

It's a little bit um so yeah there's some um testing notes. Here for staging and I have basically become stuck on this step here, verifying on production.

D

B

Was suggested to me to use the gitlab.org group.

B

But when I, um when I try just this first like get request against the gitlab or group I just get a 403.

E

A

um So, okay, my first question is: you said: you're testing it on production.

B

It's verification for this issue, so I just need to verify that the fix is out on production now and I just need to run the verification steps to as part of our workflow to ensure that it's done. It's.

A

Fascinating I'm fascinated that you have a verification so that my team does not love that. Oh.

B

E

Always assumed I.

B

Assume this was a gitlab white.

C

B

Sometimes I really don't like I really wish I was didn't have to do this now.

A

um Okay, I think on our team. It's assumed that you do your own verification, so MO I think was the author of this Mr that made this change and so on. If he were on the off team, I would expect that he would verify that it works on production. So it's kind of cool that you should trade.

B

Like yeah, we do that too or I certainly verify my own ones before, but then we unassign ourselves and then the bot will assign the random member from the team too, to verify.

C

Works almost similar in.

B

Probably there we.

A

C

A

That is so interesting, um it's possible that we're supposed to do that and I just never knew about it. So today, I learned um I'll make sure to ask my engineering manager about that. uh So I mean first of all, I yeah. This is these instructions. Aren't that precise I mean a custom rolls license is just something you get with ultimate, so custom rolls is an ultimate feature.

A

um If you wanted to check that in the future, um I always have to look up where licenses are kept. If you look up.

C

One quick interjection is the: it seems like it's a setup for staging. Did you adjust the group IDs.

E

B

So I created a I created a group myself, okay, just initially I created because it says it has to be a root group. So yeah I create a group tried. It got a 403 on there and then I I had actually asked in slack um this. This same question of like any hints on on how to find or create one of these for production, because I've got stuck on that step and somebody else said: gitlab org should work, so I did switch out the.

C

B

Group ID here and um I think I, don't know anymore, but.

A

um I know what the issue is or like I'm, pretty sure, I know what the issue is, which is that you need to be a group owner. So that's part of what's missing here. You need to not just have a license or an ultimate license. Hey sweetie I got some roles.

E

A

uh Not just have a ultimate license, but you also need to be an owner to see the custom roles or to manage the custom roles. So that's what you're getting a 404.

A

um the good news in terms of being able to quickly help you is that I have test groups for all for custom roles and I can just add you to them.

D

A

So I think that uh I don't know what the best way to like share that information in the future with people would be. But I think that is a helpful you know to make people shouldn't have to set up their own test group every time, they're testing the feature right like it's just.

B

A

Should go to test it more easily, so I can add you to that group, um and then you won't be stuck on this step anymore. I.

B

Mean that would probably be enough for me and I. Don't think necessarily it's a bit of a humdrum thing to be going through as a group here. So if you can unblock me with that part, that would be a huge help for me tomorrow. Okay,.

A

Exactly that's on staging and production you're gonna test this I'll.

B

Probably just test on production.

B

Almost certainly done it on staging already, so oh.

E

Yeah, I'm glad that you could have this conversation.

A

Something new really did you do you? Do we have a process for qaing other people's work on production like doing the verification step.

D

No I don't think so. It's documented anyway, yeah.

A

I, don't we've never done that, but both of the other people in this call that's part of their team process, which is really interesting because it's not part of ours, but maybe it should be.

E

D

Should be yeah, it will be really good to have a eye of someone who has not developed the feature to Kuwait on the production, No Doubt.

E

D

Lovely, we can propose it as they should be proposing and maybe I'll talk to all about this yeah.

A

Exactly put it in like a retro or something um quickly, Malcolm, do you want your mouth username or your mouth block? He serves her name.

B

Oh uh mellow, please.

A

I'm making you an owner of this custom rolls group, so you should just um thank you. um um Okay, so, oh sweetie did you bring anything you want to work on particular I had an idea of a refactor we could do together, but it's just kind of like for fun.

D

No not exactly actually, there is not much on my plate today, I'm just hoping to work on some interesting problem with other people. In today's session.

A

D

Some documentation updates and also those I'll be working on Millions, okay, cool.

A

Okay, let me share my screen and again, if my internet starts acting up, just tell me and I'll go to my ethernet situation.

A

um This is kind of a fun one. uh So I have you all seen this Mr about avoiding active record callbacks.

E

Got merged today yesterday, it.

C

No I haven't seen it no Okay.

A

Cool, so this is fun, so um I wrote this, but I just agree. It's okay! There was a lot of conversation on it over the course of several weeks and it's basically a recommendation to in general, avoid active record callbacks, so active record callbacks are really easy to add and then really really difficult to refactor, um and there are a lot of reasons for that. um We have.

A

There was an amazing table that Peter lightson put together, so I can find it down here in some of the 68 comments of how many callbacks you have on each of our classes. Project has 870.

C

A

Accounting validations, which I would argue, are fine, like I I, find validations to be helpful and unsurprising.

A

um What's a little more surprising is like you, create a project and then like 18 things get created over here and you don't even know what it was, because it's an invisible process.

A

um So it's unlikely we're ever going to get rid of all of these. So but the point of this Mr is just like: hey when you're adding a callback like ask yourself the question: could this be somewhere else? Could this not be a callback um and so I think the consensus was people are like okay, we can do that and the risk is of course, I've already had several conversations with people who are like Oh. No I got an MR, the Callback and, like the.

C

Person doesn't.

A

Want to move it like what do I do. I'm like it's, okay, like just accept and move on, like we don't have to turn every callback Mr into it's, not a blocker I would say it's more just like let's stop and consider our options, um and so this already came up in an MR. uh Some people even see this Mr and this person is adding a callback.

A

uh Let me just show you the code, it's pretty small! That's what I thought it might be fun um to do this refactor together, because it's pretty easy to understand, what's happening here and then to think about how you might refactor it so they're, adding a callback to the band user model, which is a user, has one band user. We create this record after a user is banned and then at this is the after commit callback, and it asynchronously uh goes through and marks all of their merge requests. As author band.

A

um So and then I guess, there's also an unflag. So if the user becomes unbanned, we unmark all of their merge requests as author band um so anyways. There was a good conversation over here with the author about uh why he put it in a callback and again that's financing the Callback, but I thought it would be fun to.

A

In a separate Branch like in a brand new Branch, see how difficult it would be to refactor this hasn't been merged yet, um but to like, do the refactor that's going to be necessary to fix the issue that you identified so.

E

That's kind of the background to this um so yeah.

D

That's the concept, so do we have like some sort of guidelines when exactly should be at callbacks and then exactly should be not ad callbacks yeah.

A

The in the style guide, Mr, there is a list of situations and callbacks make sense um here, and then there is an example of moving of like refactoring a callback. It's a really simple example. So you.

D

A

Into a service, um but that being said, you know it's easy to write a style guideline and sometimes harder to like do what it says so I think that's also a valuable like experience to be like okay, here's a callback like how would I actually do this refactor and to see what we learned in that process, and maybe that requires an update to the style guide.

A

um You know with with some further instructions or ideas.

D

Are we actually working on this activity to get rid of as many callbacks as possible.

A

No in that, so it says that right at the top here is um or I try to I think we already have so many we're not trying to remove all of them. It's more just you know if you see an opportunity to remove one go for it and, as you add new logic, um you know consider not adding it in a callback if possible, so we're not trying to remove it. So this exercise is not like. This is the beginning of removing thousands of callbacks. That would be really really rough.

A

um This is more just okay. How do.

B

We do this and the.

A

End this issue and the issue that comes up when you want to remove a callback right, is that there's some refactoring. That needs to happen first right, so you have to make the change easy and then you make easy change, and so this we're doing this a little bit in the reverse order here, but I'm curious to see how difficult it is. uh So it all starts with calling ban on user.

A

So I want to take a look at the user model.

E

E

A

E

I, don't see that stuff.

A

On you all, this is a problem with zoom I'm like everyone, self-top fan. Okay, so I don't see the band modeling user. ugh This is already challenging. So where do we ban a user.

A

Oh ban and reports- maybe where does Banner report that called.

D

A

From block our band, where does blocker band get called delete, async? So that's different. This is like, if we're deleting user, we want to just know if we are.

A

If we are Banning users, normally.

A

A

Or so event, again, I guess I, don't really understand, I, don't know much about how these event blocks work. Does anybody else know this seems like some kind of a state machine situation.

C

Yeah I think yeah looks like State machine.

C

E

A

Using a band bang method, then no bam. Bing e, okay, okay, I, was being called somewhere, but it kind of is um what's the word, it's like meta programming. Are there that's like Library code, that kind of defines the band bang method.

E

Yeah, okay, okay, so.

C

Let's see that should be part of that state machine I, think yeah.

B

And it usually hooks in like um with quite often hooks in with status, C nums and creates the scopes for those as well and stuff.

E

A

Looks like the way that we are user.band, so everything in here is in a spec file right, except for this worker. So we have one place where we call user.bn I do not want to print terminal yet again.

A

You used to have that disabled that I think I as an old.

E

Computer, how old I turn.

A

So if the user is effectively banned, then we do some other stuff, so this band got calls to here.

A

So we got our and then that transitions, the state to band which triggers an event machine after transition, callback for active band.

A

Oh I see so there's a transition for active to band. Where would I find that callback?

A

Is it actually called after transition uh yep? Here we go after transition, the active to band, and then it creates a band user, and this was something that Terry and I. When we were discussing this yesterday realized um we were like what is create band user. Create band user is just like a little thing tactic. Sugar thing that rails gives you, because user has one band users you can call like user.buildbanduser or user.createpanzer. So um that was a fine example. I've.

C

Been looking for that for ages, yeah.

A

I am I, I, think yeah. We we did like end up getting going in like calling dot, Source location, to figure out where it was called, and it was like in the rails internals.

A

um So so this is basically just calling and the transition is calling like user.createband user, and then that creates a band user record where we are adding like an after commit, whatever method.

A

So the question is: um how would we even start to think about refactoring this flow and making it so that we could call this whatever method, not in a callback.

B

um I mean, from my point of view, often the motivation to moving to service class architecture in a rails. Application is so that you don't have to rely on callbacks, because you get this. You know one of the other confusing things about callbacks is like what order are those 270 callbacks on projects going to run it? You know who knows, whereas the service class gives you a linear regression through any task you want to run. So that would be. My first place to go. Is like okay, where's the service class for Banning, a user?

B

Do we pull it into there.

A

Yes, good question.

C

We actually have a band user service class.

E

Your service, thank you.

A

um Where's, this called.

A

It's the base class or band service, and we call the van service.

C

A

Isn't this where we started anyways? Okay, look! Here's the admin controller; again! You don't want to print and.

A

And so then we call band, we call band service. What does band service do also? Is this big enough? Sorry, folks, if I'm popping around- and you can't tell, can you see.

E

C

D

C

Vi user- this seems normal to me, but I.

A

Have a big screen express different.

C

Opinions looks fun to me.

B

A

Or if I'm like clicking around too fast, sometimes so, the ban method calls band service. Dot, execute execute is not defined here. It's defined in band user service, just as a permissions check, it looks for there's a state error. It calls update user, which is I assume defined in the band service class itself. Yes, it just calls user.band, yep, nice, no Bang.

C

A

Tracks an event, and it has success so I suppose.

A

Yeah, so we have this service. What would be the downside of calling? Like you know, user.

C

Trigger that background, job yeah.

E

A

What is the problem with doing that here? Why does that need to be done as a callback on the band user um class.

A

It seems like there's a description here it's outside. If we move it to the band under service level, we have to Handy handle any potential State change failures, as well as shove, even more code. Well, this isn't true because.

E

It's this isn't the user model of the service, but.

A

State change failure. So what would a state change failure be.

D

um I think, because we're checking for the success in the last uh I mean what, if it is not a success,.

C

D

Thing we will still schedule the background job.

A

No I, don't think so. I, don't know where the success is defined, probably in the base service. I think success.

D

I think that would be just the success message or something suppose yeah.

A

I, don't know where that's defined. This is the one thing I.

C

Think that just returns a service response success, the.

A

Risk response: okay,.

D

Because we are logging only if the update user is successful.

A

If, after either my gosh thanks, um okay and I guess one thing I don't know about State machine is when we call user.ban. Would that immediately raise an error if it was a State change that wasn't valid.

D

A

Yeah I guess I'm not really seeing why this couldn't happen here uh like why we couldn't take the logic in here, and there must be something I'm, not thinking ever not aware of. Why couldn't we just call this in here.

C

Yeah I, don't I, don't see any reason there.

E

A

um Because at this point we wouldn't know that has already succeeded, you know what I should have done is I should have needle or still do this I'll pull this branch in you know what I mean.

A

Pull this branch and then build on top of it, because then we can actually see if it works, because I feel like it should be able to just move there. I guess um we won't know until we try.

E

A

E

A

Make sure I have the latest changes a test coverage for new service that.

E

A

Yes, okay and then I'm gonna create a new branch. On top of this.

A

So if we go in and we open this band user.

A

A

We take this, we draw thanks and you open the eex services users.

A

E

A

So we can just call this in here or not in here sorry, NB um band user service is that right.

D

In the bands and.

A

Yeah in here in here, so we we've tracked an event, and then we need to call this guy and we don't have to call this anymore.

C

Just need to change the user.id.

E

A

Perform that um I would assume, there's also like an unbanned service yeah.

C

A

There we go and then there's also a base class here. Oh.

B

Wait yeah, it's a sign, so I will have to like use template method. Okay,.

C

Put it in the update user.

A

Should we does that make sense? I mean we're, not really updating the user, but I. Guess it's not. You could rename it.

C

Yeah or add another method.

E

C

Update Ben I, don't know.

D

A

So in the band service, we're calling Ben and then the unbanned service we're.

C

D

A

C

Just might need to.

C

The base class is relying on the it looks like it's relying on the return value of user Banner unbanned, so I'm, not sure. If we scheduled the worker there did that.

A

One: okay, oh you're right, so maybe what we do instead is we say we create a new method, like somebody mentioned before,.

C

Yeah update band state or something yeah, uh a lack of a better name.

A

Yeah yeah wait it's kind of funny. This is the update band user Flags service.

A

And I don't want to pick apart. The name I just was like what does flag mean, I think it's just marking the merge requests, um so maybe update user merge requests.

A

And then we do that, for this too.

A

Like that, and then we call that.

C

User.Id again, on that other side as well.

A

Oh, thank you. Oh I, don't think, uh and then this is the base service band user based service. We call update user. So if update user, then we can say, update user merge requests.

A

It's like nothing either a lot here. I mean we could yeah.

C

A

User, oh I, guess current user is, um is the person taking the action it wouldn't be banning themselves? That would just be. It would be very kind.

E

Of them, if they did that, but it doesn't really happens right now,.

E

B

I can point out. Maybe we might get some rice conditions here with this option? Okay, because the I think that ah no I'm just concerned that the other, the other callback was after commit.

A

It was after commit so.

B

I'm just worried about what other callbacks might be triggered that that so we're using after commit it might have been guaranteed to run a whole load of callbacks before it got called into the after commit, whereas with this refactor I'm, not sure that we can guarantee that, but I think we can I. Think. Okay, sorry.

B

A

You thinking about other callbacks that would be called after a user is banned like a user model, callbacks or other kinds.

B

Of things yeah well of just anything, that's attached to save on the user, basically, because.

C

B

Because the Callback was an after commit it's like it's, it's fully committed the transaction.

E

B

Yeah I still think actually I still think we'll be safe. Sorry, I I am yeah, don't worry, carry on.

A

Well, what makes you so sure, because.

E

B

The save I think the save will still be complete.

A

B

A

That makes sense yeah, that's that's a good I mean unless no I was gonna say unless there is a a callback on user, that's run asynchronously. But if that were the case, then after commit wouldn't.

B

Still have the same problem: yeah yeah.

A

Yeah yeah yeah, yeah, yeah, yeah, okay, so so far this has not been that painful I would say um which either means we're missing something huge or we got lucky uh or both.

B

Can we just check sorry Robert's comments again because he wrote quite a lot: I didn't get the chance to read all of it yeah there may be something in there that I'm missing.

D

D

A

When a user gets I, think one of Robert's arguments was that when you un ban a user, the band user record is deleted, um but that's fine, we're not calling anything on the unbanned user. We're just calling a worker uh I haven't actually looked at the worker. Maybe we should do that to make sure there's nothing in there. That's depending on things that we're not thinking of isn't.

C

That worker part of the Mr as well.

A

It is yeah yeah yeah, it is.

E

A

Okay: recalls band user flag service, which.

E

Is done? Oh my gosh. Sorry I'm just scan these are flags there. It is which takes in a flag and then flag merger plus. This seems pretty much fine.

E

Yeah, it's fine.

B

Yeah I don't understand that statement about the related band user.

E

B

It's not being referenced anyway,.

A

Hey Robert had an idea of what the three Factor was going to look like, and there was a problem with that version in his head and then we've come up with a different version, that's kind of where I'm heading right now.

A

Because when you mentioned defining it I mean the state failure changes. This is the big thing, I'm, not sure of yeah.

B

I think what does the ban ban bang bang method? Is that going to raise an exception if it's an invalid State transition? I can't remember how those things work.

E

A

Guess we can look at spec and service, let's.

E

A

hmm I guess this is the future that my team owns I'm, not very familiar with it. Okay,.

D

He is actually uh mentioning that a lot of work for band user is being done only by a activity called callbacks. So probably is scared that if you start using the services, then we may not call the I don't know some. Some functionality is in a correct order.

A

But what is that functionality? Do you know, or do you have an idea.

D

No I don't I'll have to look into it. Let me just check.

D

D

um You are all of this Emma. Yes,.

A

I'll put it in here.

C

D

C

I think that bang method- this wasn't even misremembering something okay.

C

Give a quick look at the state machine, read me and it doesn't seem to have to bang methods. So it's just the name of the the transition.

A

Okay, but then, but certainly it seems like user.ban- is coming from State machine right and that goes to yeah.

C

And yeah to trigger the band event.

E

Yeah move to the end. Okay and.

A

A

So I guess we could write a test in here. We can steal one of Robert's test. Thank you, Robert for doing all the hard work and.

E

E

It on flags like Flags, March requests; no, it Flags merge, requests uh um to the user.

A

We're gonna have to actually create a merge request for this user.

A

My next Great ACT in the style guides will be to say we should use that anymore. Just.

E

C

But I don't think that is going to be a long fight.

B

Sorry, what was it again?

B

E

C

Support you, but it would be a long fight, yeah.

A

E

Would like to get.

A

Ready, no more sad.

B

Examples yeah get rid of let completely.

C

I know it's rather yeah.

A

A

um It just makes it hard to read tests I.

B

Think, oh I see like all all sliders of lit yeah.

A

C

B

Be as well and.

C

Instead of that, we could have it right in the example every example would then, if you don't put it in a let it sort of makes every example more in a an actual example of how to use the code you're testing. So you see all the setup that all the.

B

C

Everything that's necessary for every step, I'm.

B

100 down with the I think one of my pet hates, as well as like so like redefining subject to the method, call which, in my mind, subject, is always the object under test or it should be, and so inside your it. You should call subject dot method with the arguments, but we seem to have quite a common pattern where somewhere subjects redefined to um something.

C

Else, there's a very dry up the method call again yeah.

B

C

Use some variable yeah.

B

Anyway, this is a huge side. Quests.

B

A

Love that side quest, I.

C

A

Feel, like that's my true dream: I mean yeah and I I. Think for performance reasons we kind of need, let it be, but I have this suspicion that sorry I'm still on the side, quest I can't get off of it. Is that like we, because we're so into let's we're creating more data than we need for a lot of test scenarios, um we're not recreating it with Let It Be but then like? Sometimes we have to do, let it be with reload but I. Don't know, I just feel like I.

B

Think one thing that I've seen is like cargo colting, let it be because Let It Be has a very specific use case of like this is an expensive object to build like project, um whereas something cheap. Let is better because the aspect cleanup cycle gets hooked in better, but people just let it be everything, strings, yeah, but booleans. You know let it be just.

C

Because yeah as well had that on a on a review yesterday, let it be five.

A

But our dogs do say, like always let it be I mean. That is the documented thing, and maybe it's an example like I agree that we shouldn't use it for an integer, but I. Do we have this interesting thing, I think we're such a large team that works on the same code base. uh People tend to like really respect the docs too much it's like they should be reading them and then, like maybe changing them, but I think people read them and they're like okay, like this is what I should do leads.

E

A

Some stuff- maybe we can add, that to the docs you don't have to let it be.

D

uh um Query I had that. uh Why are we not thinking of putting this worker in the after transition block.

A

After transition- oh, like this one or after.

D

E

D

There's a there's a block for after transition, also right, our active colon band Yes here.

A

That's a good question: I guess the thinking I have is that this is a steep machine. Related thing like it's kind of just a callback in different clothes.

D

Yeah, this looks like a fallback to me as well: um I'm, also not very much familiar with the state machines actually, but I think this is a I feel this happens once the transition takes place, it comes and calls back uh executes the code block that we have mentioned there. I.

C

See someone is.

D

Actually, using run after commit in the above after transition block.

E

A

D

Just uh on line number 488 in uh someone.

C

D

Actually, using run after comment here.

A

I think I can see you, but they heard a voice.

B

I think, basically, if we're kind of hiking uh kind of moral stance on this change that like we don't want to put something into a callback, then if we put it into one of these callbacks, which they essentially are the after transitions, such as callbacks in your goldbacks.

D

At the end of the day, we are using the callbacks and we are there.

B

Is yeah I think yeah, it's probably just a that's.

E

B

Thing by a different name.

A

And somebody did bring that up in the style guide. Mr. There was a pretty extensive conversation.

A

Where was it about uh State machine and she yeah uh and the conclusion was kind of like yeah State machines can be good but also like, maybe sometimes we would. We would want to use something else because they can be confusing in a lot of the same ways that callbacks are confusing where it's like things have like. You call ban and then like 18 things happen and you it's hard to figure out where or how um but anyways. So this did come up already and it's interesting that's coming up in this conversation already.

A

Maybe we should address this more explicitly in the guide.

B

uh I think it's one of those things where State machine has a use case for me that just Banning and unbanning users is not necessarily the best place like State machine is not the state machine for me is like a linear, workflow type. Things are a better bit for that.

A

B

E

A

Because a user can be unbanned, do you think it's not a linear process or.

B

Does it fit well? No, maybe it is I suppose we have a lot of other states right, deactivated, active yeah. You.

A

Do but they can, you can bounce around, but in different states like.

C

A

Not always going through the same flow, no.

B

It doesn't make sense, actually yeah yeah.

A

um Okay, let me we'll move this one next, uh so we have a test here um where we create a merge request, it's not the author's not banned. We ban the user, we reload it and we expect the merge request author to show up as band. Let's see if that works.

A

I, don't remember if you do anything special to say like run background jobs immediately or if that's the default, Behavior I.

C

Think that's implicit in our tests. Yeah.

A

It is okay, um I have an extra end.

E

E

B

Pretty sure I have had to put that sidekick in line thing in somewhere yeah. So I don't remember where, though,.

E

Yeah I don't know if it's a specific type of job or.

A

Something all those flags by the way did you know that if you misspell, if you put a flag in a test and it's misspelled, we just ignore it so.

C

I figured that out the hard way you do. Oh, no prior to write, sass and I actually wrote SAS instead of software as a service so and I.

E

A

C

Not seeing it I wrote s-a-s-s instead of oh.

E

C

Oh and I could not find the error, and I was just like stuck on it for like three hours and I was like I I put the flag in there. Why is it not working? Why does it keep telling me I need to be in SASS and then and then I finally noticed it and felt very, very dumb.

A

That's amazing um that that's also sad but like.

C

Three hours on a.

A

E

A

um What's Happening Here undefined method, band user, I.

B

Think, where you, where you copied and pasted that spec from that's probably a named subject.

C

A

B

A

My gosh, thank you. Thank you. Thank.

D

A

um So band user is defined.

A

Whereas band user defined oh.

D

It's just stupid.

C

A

We don't need that then, because we're not depending on that object, I was just creating the band user record, which you don't care about. We just need to make sure I believe we do want to make sure that uh if we say recall band user oops, not that.

E

A

Let me reload the murder requests I, think that should work yeah. The misspelling thing I saw that there was an Amara at a Robocop rule, If You, misspelled aggregate failures, because that's really hard to spell a lot of people have to spell this.

C

It is um Auto completion, the most that and vulnerabilities. Oh.

B

Yeah, then get me started, that's my team's work vulnerability. So how to I'm.

C

B

To come up with a new name.

C

Someone suggested some some vlns yeah.

B

I think that was I think the v13 v10s or something was.

C

Something like this.

A

Almost worth the v10s.

C

um Just like M9.

A

What, if I'm nine M9.

C

Is for maintenance? Oh.

A

My God I never heard that I know, accessibility,.

C

I made that up last week because oh you did not smell it,.

A

E

A

Like that and then I like I, only know of kubernetes and then accessibility, oh 11, why.

B

Localization, that's the other one right.

C

Internationalization yeah I.

A

Never I never knew that's why I would like it. You know and I thought I didn't ever just considered it I just called.

E

A

Idea, uh okay, I think we might need to run sidekick jobs in line here, but, interestingly, they didn't have to hear in order for this to pass. But.

C

um What does band user do in this case? Because if it's just creating that user, we don't have a callback and then.

A

uh Well, it's calling.

C

Subject: band user describe class.

A

Skype class new DOT execute. So when you call band user it should it should call this.

D

um Is it because it's an asynchronous job, it's not that's! What I was.

C

Wondering worker flag on the.

A

E

A

The um it, what is true is this is for band service, not um band user based service. That's fine! That's fine! This is Beyonce risk, so we could put it I mean this is my main method of debugging is.

E

That it's not good.

E

A

Let's try this see ing getting in there. Are we even in the right.

E

A

It is interesting if we have to run them in line, but they didn't in the model class.

E

B

Well, I think you wanna, you wanna, put a buy bug into whatever the worker calls right.

A

Yeah yeah well I'm, just gonna see oh I see well, let me just try to perform it and see if it errors out.

E

Or perform now artwork.

A

Okay, so you put a, we can put a bug in there and see if it gets in there. Oh, we know it's getting into that and.

E

A

B

uh I think even oh wait well that one.

A

We can go inside here and the band user flag.

B

Service I think that would yeah that's like. Definitely not gonna get cold. If we, if we don't have inline soccer.

A

Okay, okay, so I'll run this again and if sidekick job is not executing, then we probably will never get to the pry um and then I will be interested to investigate oh shoot. I have a meeting. I have a coffee chat in two minutes, but um oh yeah I shouldn't get in there, so I think we have to call sidekick in line. Do we just do that? Like this thing.

B

Spell it right there.

A

Okay, the thing with them, love them so much but autocomplete, not so good. um Psychic in line is that look great.

E

um We'll see I guess.

E

E

A

Yay look we're inside.

A

Nice, okay! Well, we didn't pass, that's fine! We can figure that out later, um okay. Well, this was fun. um We didn't complete every Factor but learn.

E

Some things, so you all so thank you for joining.

D

Thank you. It was a very good discussion. Yeah yeah.

E

B

C

D

C

Isn't cute? That might be enough in that case, make.

A

Sure the job's in queued, oh okay,.

C

Yeah, we don't probably actually need to run it yeah.

E

A

A good idea: okay, I'll.

E

A

That later, thank you thank.

C

A