GitLab AMA & Q&A's, 8 Jul 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2020-07-08 AMA about GitLab Releases

Description

AMA with the Delivery Team

A

Cool hello, everyone so.

A

Get the agenda open.

A

So there's actually nothing in the agenda. So if any wants to add anything now or freestyle, ask then please feel free to go ahead.

B

I'll ask a question that I find the issue linked to Maurin in an issue but I'd love to ask the team, which was what I'm trying to get an understanding of is kind of the the value and the. Why? Behind our kubernetes migration, so I know it's an ongoing thing and something that the team has been focused on for I. Think years, if not months, but I did like I wanted to get us, try to get a sense of the implication of once we move like.

B

What's the what's the value and the drive behind moving to Cabrini space deployment,.

A

Great question and actually very timely because we are producing some documentation putting some stuff in handbook, so this will all be explained like more clearly, more more visibly for you, so the it's been underway for a couple of years.

A

Job will have the exact take that, but the big drive is kubernetes gives us lots of benefits like it gives us a lots of kind of flexibility in terms of how we scale it will give us a lot of stability and things like that as we go forward so as Gilad gets bigger, it gives us a lot more control over how we go add things into communities, how we, how we manage them and have a scale it up.

A

So that's kind of the big motivator there are other, hopefully what we're already seeing that the cuban h pods start up considerably faster than the VMS. So that gives us a lot more opportunity in terms of scaling like we'll, have a lot more ability to handle sharp spikes in traffic, because we can add pods way more quickly than we can currently add. Vms should give us better kind of cost management as well. Just cuz we'll be out our tune, things to more likely fit stuff.

A

So I think from my opponent view it's very much around giving us. It should give us a little bit more confidence in the stability, but also it gives us a little bit more flexibility in how we actually set up and operate get up. Jeff. Do you have other stuff you'd like to hold out there's a super high level yay.

C

um Yeah, that's I, think everything Amy just said: there's also I mean one of the main reasons we want to move to kubernetes on comm is to make it easier for our customers or our self-managed customers to run give up in kubernetes and by dogfooding the home charts and running running, using the same tooling that we provide to our self-managed customers were kind of like working out the problems and making sure that things are. You know working well, so I think it's it's twofold, I think it's like one, it's very good for running the SAS.

C

You know there are just a lot of inherent benefits and running using kubernetes and the other part is also there's just a benefit for our self-managed customers. Cool.

B

It's great I, like the dog fooding angle. One thing you had said amy was this scaling speed like. Does that manifest itself as less performance issues to our users, who might experience interim performance problems while we're scaling give at comm via VMs today, or is it like less downtime in general, because we might have been down or having a problem and we can scale it to alleviate that problem quickly in.

A

Theory, it should be both so one of the good things about Cuba Nettie's is. It should be able to sort of know when it's running out or like we figure it. We need to add more capacity. It should be able to add more capacity quickly and after handle sort of spikes. It should also be able to manage its own health right, so we can set it up. So it will know a certain say: 25% is unhealthy, it can add its own pods and bring them back.

A

So once we have everything migrated today, right and configured in theory, it should be both okay.

B

Yeah that last point is also I, guess, a statement of it frees up the team members time instead of having to manage that more proactively. That's the point of it to managing its own house, got it just like Marin I, see you added the handbook and I. Thank you for that.

D

There is also something that I would like to have there here because may help a lot from product perspective. Even yeah, lots of profit forks are in this code, so there is an issue that I hope in some time ago. I'm going to paste the link here, which is about a nice feature that we can get from a complete kubernetes deployment, which is matching routes with product categories.

D

So if we go down this route, we could hand up having the deployment segregated by product categories, and then we can extract metrics from product categories like number of crashes or number of users, or things like that. So this is very: it's a moonshot right this first, we need to be able to run everything on kubernetes, but once we have it, we can get also a useful business information, in terms of which features are most useful, which features interacts with other features. So I'll just leave you the link.

B

Can I say yes and on that? That would also imply, as I understand, our infrastructure teams set up today. There's not really close alignment to product categories and features and the infrastructure teams responsibility. But that would maybe allow us to be a little bit closer to DevOps, see where the engineering or development teams would have much higher awareness of things like crashes or other things, particularly affiliated with their features and categories.

B

You are not in your head. Obviously oh, but I just want to make sure that, like is that also what you're suggesting yeah.

D

Yeah basically, increase ownership. Okay across the whole brother yeah, definitely cool it's great way of bringing.

B

Cool thanks for answering that question: y'all. Okay,.

A

Thanks for asking, is there anything else anyone loves for us Jackie? Did you have anything you wanted to ask today, no I'm just being a little bit of wire and listening in thanks, awesome cool. So is there anything anyone from delivery wants to especially share about releases this month or.

A

Awesome and that's pretty much everything we've got them today, so thanks so much for joining and asking questions we'll see you next month. Thank.

B

A

You for having us al, but.