GitLab Delivery Team, 21 Mar 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Delivery: Group limit counter feature discussion

Description

Alessio C. and Marin J. discussing first steps for https://gitlab.com/gitlab-org/gitlab-ce/issues/59232

A

hmm All right, so we are talking about issue can't.

B

A

Anymore, oh, no, okay, I can hear you now yeah all right. We are talking about issue 5 9, 2, 3 2. We can introducing the storage counter at Google Group level 4 package feature. So hmm what I'm curious to know here, obviously we'll need front-end and we'll meet you X here so I'll, I'm gonna work with product to finally figure out, like you X like what needs to be done there and so on. So that's gonna. Take some time and front end.

A

I have someone in mind already, but I need to check their availability and that pretty much depends on how fast we can get the back end down. So it might not necessarily be necessary to do it right away. My question for you, though, is if we are building something like this. We all like you and I already know that we need to kind of put a framework in place and package is going to be the first part of our code base that is using it, but we will kind of need to think a bit about.

A

How can we make this? Not generic but able to accommodate all the other things that we are going to be adding right sure. My question to you is I know that we had number of these things in the statistics.

A

Project. Statistics, project, statistics right, but this is now group. So this is a group level. Do we go to the same concept and have a group statistics and in the same way that is created for project or do we figure out something completely different? Ok at group level, I'm quite sure we have some SQL queries with some coalescence operator that can sum up the overall result for a group.

A

Ok I think that this could be a good idea to keep the granularity at the project level, because maybe you want to know which is biggest offender his you have the you have the limits and group level, but you want to know where you have to act for rows.

A

So if we more detail it data, we have the easier we can group them and the other information a group level, but if we stirred them a group level and why right right right, like basically or you're, saying that we should still be focused on the on the project level. That will be our single source of truth and the limits part is going to come from the group rather than the project, exactly right.

A

I'm thinking about extending the project statistics, because I already have a lot of helpers in place for incrementing stuff I think we should just add columns there, some other kind of statistics, and then we can build group yeah. The group thing as a separate class or in the namespace I. Don't know I think about this, so we can leverage limits at group level based on the result that we have single project.

A

This is no that makes that makes quite a lot of sense, because if we think about what else we we are going to be including there we're going to be including artifacts, we are going to be, including somehow registry images we're going to be including we are already including LFS. We are already in its storage right, so it completely makes sense to just group it together all right, and for that, do you think yeah I, just tell you this.

A

If we are going to add also attachments to this later on, because right now, we don't have a way to delete them, but yeah we will be still. Is this something like grow space? And for that case we have attachment at group level and it's the only weird thing, because it's for the epics I think our group level thing and you can add attachment to ethics. So this breaks the simple rule we can think about this later on, but yeah just I mean.

A

But there is no reason for us why we wouldn't be able to just say that even on the project level, we are counting attachments, but the that doesn't bring that basic rule like the basic rule is that in projects we still have all of these items counted. The only difference is going to be on the group level. All of a sudden we are going to have like project dots, uploads plus yeah, exactly right, all right, which is probably what's gonna, happen for a lot of these things by the way now that I'm thinking of it.

A

Yes, we do not have group level attachments or artifacts, yet we don't have root level packages like I, just yeah, yet that's yeah, okay yeah, you say yeah, but if you think about this all the api's that we have in place for.

B

This are based on the project unless yeah I know I, know yeah it's a rival yet yet it's important here, all right. Okay, so can you tell me.

A

For for this specific thing, we'll probably have to create separate issues for the smaller yeah.

B

Tonight's here.

A

I think our main issue needs to become introducing storage counter yeah.

A

Yes, yes, yes, that's the one, the one I linked to so 5 9 2, 3 2, so that one needs to become the the main issue and you need to start creating like smaller items like defining the work. You want to go and do for the what what we discussed um but.

B

A

And, let's think about scheduling that for right now, as in McCann like when, as soon as you finish, the ongoing item that you have and you have free time, you can start working on that. That does not need any front-end. That does not need any UX right right. In the meantime, I'm going to start attacking UX to see what they say about the group level. How do they envision this to be answer and and so on?

A

So while you are actually developing and we ship this, we can already have the UX defined and I think, depending on how fast we get that help. I can then ask front-end, like I, can schedule front-end for maybe next release to have this presented depending on how fast these lives?

A

Okay, we are still considering only counting stuff, not enforcing and is any limit at any uh now for now like we, we need to think about not excluding any decisions that we need to make later on, but for now, I think the main focus very short term is count things and as soon as the UX is there count they on the group level yeah in the in parallel. It's gonna be another discussion on alright.

A

Now that we are actually counting this and now that we have actually have it presented, which limit right, okay, so when okay, so I need to do the package counting and the backfill also because we already have packages so in the first. Very first step is adding count and backfill okay yep. Absolutely absolutely!

A

Then we will like I'll, have to discuss this with with product on which limits are we going to be taking, because this is gonna be complicated? If we have a project limit of 10 gigs that counts everything in, but then on the group level, we enforce it from some different limit. It's gonna be like oh wait. That is it a group limit of thinking we're a project yeah.

B

A

Own vision on this is that the the settings that we have at the moment are okay for certain cases, but for the comm we want to allow owners to lower and only lower the limits based on the upper limits they have so that the an admin instance can set you. The group limits, as let's say what you pay for so you paid for ten gigs total Oh 100 gigs total.

A

So you got that limits in there, and so you, as as a group owner, can set a smaller limit on the things that you create to make sure that you are not hitting too much space right over the overall. Stop is still at the group level, so your namespace cannot exceed so basically on github.com I mean this is not only get about come this. This is. This is very much useful for every self-managed installation as well, because me is a admin of the instance.

A

I want to say that all groups can have a maximum of hundred gigs, for example, allowed and they cannot go or if they need to go over then I as an admin would have to go and talk with. Why or like. Why do you need to go over and do we need to have a separate group, for you prove the increase and.

B

A

Owner of the group is the like: the mini admin like it's administered, the projects inside of the group and says well: I do have 100 gigs maximum that I can use, but by default. Github sets this limits to ten gigs, for example, right, like the group level, yeah and I as an admin can decide. I'm gonna buy more up until hundred gigs I can buy yeah if I want to.

A

So that would kind of make sense, so it goes top down, but it allows enough wiggle room, and that would mean that for project right, like me, is a project owner within a group. I will be getting a certain. You know allocation based on what is available on the group level, but then I can say, like I'm, not gonna allow my project members to have more than hundred gigs. Oh sorry, 100 megabytes or a.

B

A

Something like that, but that is not our story. To example, yeah, to figure out cool I'll, come to you, then, to create the issues that we just discussed. So this is for the backend part I'm going to be using this issue to try to drive the UX part, and you can then start on the smaller issues deliver them. While we do develop UX and then we can move on to the next slide. Okay, good, all right cool!

A

Let's see! Oh thanks! You want I'll chat with you later bye.