GitLab Package Group, 8 Jun 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Package quad-planning: Milestone 14.1 review

Description

No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).

A

Hey everyone thanks for joining the package, quad planning meeting this is week four, it's about the right page. This is week four, where we go over the plan for the next milestone, so I shared a link. I could share my screen actually to the milestone planning issue. That's the first item.

B

C

A

You could see the the screen right. You could see the planning issue. Okay, so I know we have a couple of things that we should mention that it's in july and late june, we have a couple of time off going on. We have a new, develop ruby developer, that's starting!

A

So I tried to keep the especially the back end, ruby issues a bit light, so if you feel like the things are at risk or that we're not capable of getting all this done, please yeah, let me know- and then generally I try to prioritize issues for the container registry.

A

The cleanup policy, as well as online garbage the code changes needed for the dual migration path.

A

I've tried to prioritize a bunch of dependency proxy issues, knowing that we're going to be prioritizing bringing that, together with the package registry in the coming months, and then just uh some front-end scaling issues to take advantage of the graphql work that we've been doing and then a follow-up to some of the community contributions. Okay, so that's high level I brought in haley and juan. I asked you originally in the epic for the dual coding, which issues we should bring in hillary.

A

You mentioned these two as being the most important, um but then john said we should bring in all of them. So I added in the remaining issues from that epic, as stretch goals.

C

Yeah, I think that's a reasonable way to do it.

A

Okay, I just didn't want to like overwhelm us with p1's and then had heard rsa do ratio if we didn't end up getting to them, but ideally we could get to all of them. I guess that would be great.

C

Yeah, I think I think it's possible. I think I, the word devil- needs to get prioritized and get done. I think we can start a little early as well.

A

Okay and nothing is missing or should be worked on aside from from that right. That is the most important thing in your mind,.

C

A

Good, okay and then the third item I have here is the percentage based rollout for the cleanup policy. I know for this milestone we're going for one percent in 14.0 and then we'll look to see how it performs and bump it up. I saw that um and then another change that could go out with this is the cleanup policies are not showing the nil values, that's a nice usability improvement and then including the cleanup status on the list page. So it's another nice usability improvement as we roll out these changes, david or nico. Did you?

A

What do you think of the set of issues for the cleanup policies? Do those make sense? Did you expect to see anything here? That's not.

D

Seems so good to.

E

Me uh no, it's all good those two issues on the ui, the nil values and the cleanup status, and they were brought up multiple times by support. So that's why I think it's a good idea to focus on on them and that's it.

A

Okay, yeah, I know I agree. I've I've seen a bunch of question confusion about this recently, so that would be great to tackle so then on to the dependency proxy. So we have a a fairly common bug that comes up from support and from customers where the feature is not working when single sign-on is enabled.

A

We've done some investigation into this, but it would be great to resolve that. The next issue is uh adding deploy token support for the dependency proxy. It's another frequent customer request to add support for that um and the next one we were talking about in slack recently, we were. We only have one metric right now for the dependency proxy and that's if it's enabled, which is at the instance level. So we don't really know how many groups are using it and how they're using it.

A

So this issue will track um how many like how often how often our pull is being done. What's the hit ratio- and things like that, so this will be really informative, as we can consider expanding the functionality of the dependency proxy.

A

Any questions on on those.

A

Okay and then nico, I put in this issue for converting the package details page to use uh apollo and graphql.

D

Yeah yeah one question: why the detailed page on p1 and the list on stretch and not the other way around.

A

I made one guess of what you might say: the priority was, I thought details might be slower performance, so I thought maybe it would make sense to do that. First, yeah.

D

Yeah on that regard, yes, so we would solve more issue details yeah. I actually asked him the issue to swap the two, because you know in my mind at least, cancel for the details.

E

D

Actually right, we will solve some stuff by moving to the graphical implementation and details page. So actually it's a good call. Okay, no! No risk of.

A

It looking weird or a bad user experience.

D

No, the ui is going to be the same. It's just I'll refresh.

A

The data- okay, so yeah that good. I was hoping to clarify that. Thank you, um haley. I have the increased storage driver visibility for s3 this, like a follow-up to an issue that we I thought did we finish this or is this still open.

C

That is still open.

A

C

um With the work we've been doing, with s3 with, uh we have the wrapper all the um uh actual calls out to the s3 api so having making that rapper, um more generous, um more like making that wrapper include a logging function um would be pretty simple.

C

um That would do do a lot. I think I think our, I think the the difficulty this is like. How do we get the logger into s3 um properly, which it shouldn't be too bad? Do.

A

You think that we should prior still prioritize this, given the amount of dual code, routing work that we have to do.

C

C

We could make it a stretch goal uh that is like related to a customer, though.

C

Yeah, I would move it to a stretch goal like we really just need the hammer on the dual, the dual code routing. I.

A

Think so, okay.

A

Plus that will give us an even ten that will that will feel a lot better, just kidding, okay, that's it and then uh steve and and david matthew, the person contributing the debian repository reached out in the issue and asked if we can help contribute the documentation.

A

So I was hoping I think it makes sense to schedule it because otherwise, that will be hard to get to. What do you think does that make sense to schedule as its own issue or um yeah okay. I.

F

Mean it's it's it's essentially written. um We just have to make a few small updates and then it's ready for review. Okay, uh and we of course have to put it into an mr, but it's uh I've worked. I've used the documentation to test it. That's already written it's in pretty good shape. It just needs to be modified with some items such as mentioning that it's behind a feature flag and whatnot.

A

Okay, perfect thanks all right and then stretch goals. uh We mentioned all the container registry issues and then for the package registry. I have a few scaling issues that would be great to get to. I know: we've been wanting to investigate the latency of the npm endpoints for the package registry, we've kind of been putting that off for a bit. We had great success when we did the same for maven, so be nice to look into that.

A

There's a composer endpoint, that's causing high latency as well that dan called out when he was looking at the error budget. Dashboard we're hoping to take a look at this and then there's a data issue um right now. Any any activity with generic packages is not being captured in the usage ping, so we're shortchanging ourselves on users and packages published in all of our metrics.

A

And then finally yeah nico, we mentioned converting the package list page to use apollo and graphql.

A

Anything any other. What do you think of the stretch goals? Anything that you were hoping to work on that we we should consider swapping.

E

uh There is this issue I try to to fix it on the one for nugget packages, where the archive is really really really big, so the fix I was uh hoping that it would work didn't work because it's not um it didn't target the root cause of the bug still need more investigation on that, but I'm pretty sure that um there is.

E

We removed the file on object, storage, meaning that the file is on object, storage under one key, and we change that key and instead of using the move or copy api or whatever the api is it's downloading the file and re-uploading to the new key?

E

A

It's using too much memory or something.

E

Yeah, and so I think we just run into memory issues and the background job just gets killed by the monitoring and the the whole sidekick gets restarted.

E

So that's uh I think that could be a problem that is not proper to nugget. Only. We could have the same with all packages that need a background processing like ruby, gems and so could be good to focus on that bug to fix it.

A

Okay, I think I'll swap out number two here with that one because you, even though this composer callback is not working properly, it's it's not at the same scale as potentially uh background workers processing uh those packages. So I think we could swap those.

A

Or I'll at least add it about higher in priority, and then we could, if we can get to it, that we get to okay, I added this section I think a milestone or two ago for feature flags to manage. I know we're managing the helm, charts feature flag should I also add the one for the cleanup policies in here as well.

E

um I'm not sure what this section.

F

A

I don't know if it was useful or not. I I feel like what happened was uh in 14, maybe nico had we were had to manage the future flag for the settings change and we wanted to call out that that work existed.

A

But if this isn't use, if it's not useful or it's duplicated, we don't have to capture it here.

E

Okay, let's skip it for okay, it's not super clear. What this item means does it doesn't? Does that mean that we need to enable the future flag in 1401 or in 41, or does it mean that we need to remove it or clean it up? Yeah.

B

E

B

A

What do we have to do for the helm, charts feature flag in this case.

E

um Wait for the implementation to be complete.

A

So I'll remove that.

E

It's about there on in the same state as the debian mvc feature flag. We are waiting for the work to be completed.

A

E

So I'm just going to remove that section for.

E

A

Okay, so if you had to say given um vacations training, a new developer and other work on a scale of one to five, how do we feel about the p1 accomplishing the p1 issues?

A

You could just do a show of hands if you want.

D

And then we do a show finder with a scale from one to five, with the numbers on the hand.

A

Okay, see four four fours are good.

F

I think from the ruby backhand standpoint, the only one that I'm worried about at all is the sso bug for the dependency proxy, because there's just some unknowns there of like what's actually going on so it's hard to say how much work is involved in the fix until we understand what's actually happening, but otherwise everything looks good. From that standpoint.

A

Is there anything we can do to do you risk that I could try and reach out to the customers in there see if they can.

F

It's really just the investigation, um that'll be the initial work, probably like the first day or two of work where we work through reproducing the air and understand, what's actually happening in the underlying code that we need to fix.

A

B

A

The last thing uh there are two issues that were in 140 that were stretch goals and just one was for a link to. It was first support for symbol packages for nougat, which is how you test your code. That's a fairly popular issue, but steve. Do you want to verbalize your comment.

F

Sure um yeah, I was thinking about starting that one. uh I did an investigation right before I took some time off, but based on that investigation, it is a pretty sizable amount of work, maybe even up to a three in weight. So I don't think it would be finished in time for 14 and it might just be better to schedule this for a future milestone. In that case,.

A

Okay, yeah, that's what I thought and the other one on. There is another dependency proxy issue. This one is failing to pull from ec2 instances.

A

This one has some customer interest as well and some support, but maybe we need to reschedule this as well.

A

We could see offline if we could look into it. Maybe there's a solution, but if not I I could reschedule it as well.

F

Yeah that one I'm just not too familiar with it, it does look like there's a possible fix labeled in the description, um so in that case it could be a little less difficult if that's actually uh a true fix, but yeah. It would just take some time to look into what's going on. Okay,.

A

Okay, that's all I had for this for this meeting. Does anyone have anything else, they'd like to talk about.

A

Thank you for joining and for your attention and I'll talk to you soon. Bye.

D