Project Harbor Community, 19 Oct 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: CNCF Harbor's Community Zoom Meeting - Oct 19, 2022

Description

CNCF Harbor's Community Zoom Meeting

A

You can you can transfer yep, you can just wait to tell Junior I told you maybe can help the whole host that okay.

B

Okay, um hello, everyone uh today is October 19th and that's the official community meeting for Harbor. uh My name is and I'm the community manager for Harbor and now I'm gonna share my browser, so we can see what we have for today.

B

um That's what we have fixed today, pre-written into the the file.

A

B

Gonna send again the link to the file. Still please uh address, filter at your topic there, so we can keep a track. What was discussed when and we have a better user experience for the newcomers with that I can see some new folks over here. uh Someone wants to like the Caro I. Think I see you for the first time. Welcome you want to say something like introduce yourself.

C

uh Yes, can you hear me yes, awesome, uh so: hey hi, I'm David uh for the media Foundation uh in the cloud services group we host. uh Actually, we host like Cloud for volunteers and the foundation itself to use and randomly use any projects except wikipediers themselves, Wikipedia somewhere else and when we started using Hardware uh not long ago, but we plan on using it uh more often for all the cloud and hosting images and all those things.

B

Welcome to the club.

B

Super happy to uh to see that you folks are using, however, hope we can be in help um and to to guide you through some. If you have some issues practically, you can see in the uh participants, you can see the most knowledgeable, knowledgeable folks on the planet, about Harbor, so feel free to reach out to everyone. uh I'm, not like the most knowledgeable uh in terms of the the source code and everything, but that can help out as well so don't hesitate. Drop me. A line is back and I'll.

B

Try to to help you out.

C

Thanks I might take your award for it. Okay,.

B

So with that uh two updates, our project to date was accepted for kubecon keynote, so we will have our 45 seconds of Fame on the keynote sessions, so we'll be on the stage uh all together with Linker D content of the uh it's, the whatever uh like the biggest graduated projects. uh We are fifth in a row, so people won't be that, uh like bored already from the projects updates so yeah uh what I've added to that. It's like a pre-recorded 45 minutes, 45 seconds, video, it's about cosine, the new caching layer and the CV.

B

That's that was everything that I was able to shoot in 45 seconds, so uh yeah I'm, sorry, my little one just woke up and he's coughing, so yeah uh second thing: I've added the Oktoberfest uh topic on our Repository.

B

So we can welcome the new some some new folks to join. Just give me a second.

B

I'm sorry uh so I've added that to the main repository I've added that to the community and the website repository so uh I've also sent the Twitter out that if someone wants to contribute back like a fixing to documentation or something about um Community ideas whatever uh so, please take a look at the PRS.

B

Maybe something will come up, but the primary reason for this one is just to gain more uh publicity about our project that yeah to get more like new folks to wish to join us. um My next thing is kubecon Europe 2023.

B

uh The cfp is open, uh I, send a mail to the mailing list.

B

um The cfp period ends on November 18th, so we have a month from now to submit our cfps, um so I'll be super happy to see a few at least few cfps submitted I can help out with all the logistics I can join you on stage. I can help out with the slides and everything on our site.

B

We're gonna do uh in-person kiosk, that's still not open for application, but uh I'm gonna file that so we have someone from Europe um if wants to or if someone is struggling to, keep going to Europe I'll be super happy to share the stage there and next thing I want to discuss.

B

It's I want to book a 300 minutes period, every Monday of your calendar, and it will be asynchronous or synchronous like over Zoom uh to do like a PR, judging kind of connectivity. So we don't miss any new PRS hanging there without um without feedback or anyone actually cementing or something. What do you think.

D

E

D

It's a good thing to do. Yeah. We should definitely think uh regularly on on the open, PRS.

B

Okay, um from your perspective, with.

F

B

Is it should be like 30 minutes Zoom every Monday, or we can do it like asynchronous and I should, like remember, uh remind everyone to dedicate these three minutes. I I'm I.

D

B

Thoughts on this one but I wanna I wanna, hear yours first.

D

Well, um I find both study is good, but I think we should start with a in person first, because until it's um until it's elaborated and then kind of a settled, then we can switch over to an unsynchronous mode, um but I think you know to get the though rolling we should do it synchronously it's. This is my my recommendation or my take on that.

B

I share the same. The same uh I think to build the habit of doing it without actually uh think of it. I think we should do it like a few times over over zoom and to to build a habit, and then we can we can like do it asynchronous and just discuss on this meeting every second week, if something pops out that we cannot review or needs more attention in terms of review, reviewing foreign.

B

Miner the rest of the team.

A

um We actually have a regular uh Awakening meeting to try out the issue and review a charge of PR's. um um We uh so I'm, not sure. The the study minutes that you proposed um um has some overlap with our meeting I mean um so.

A

So my my personal opinion is that we can do it uh asynchronize. Firstly, so then.

B

um Do it what I'm sorry can you repeat the last one.

A

Oh I see my personal opinion is that we can do the the review. I may focus on PR uh asynchronized.

A

Firstly, so then, then, yeah.

B

The thing that you mentioned, that you do that that's internal to VMware right, but that's.

A

It yeah yeah yeah yeah, all.

B

A

That's that yeah.

B

That's my point: we shouldn't be doing that internally. Only in VMware, because in in this case, for example, Vadim cannot join I cannot join you yeah now.

A

I I agree: yeah.

B

A

That so so um yeah, so my my point is that the um if we book another a half an hour to review PRS, it will have some overlap with our internal meetings. So we can reorganize these two meetings, um so we can confirm with journey, to leverage most of our efforts to real PRS. um Maybe we can.

A

To cancel our internal review meetings and just attend the the that's 30 minutes, but that's.

B

The best thing yeah, okay, that's the best thing that can happen in my personal opinion,.

E

B

uh I know that I spoke with Jolene about this. One I spoke with you in the past about this one and I know that you, when you discuss that in in in your language, it's much faster and more productive in general yeah, but that do not include me do not include with him yeah. That's not good. I mean it's not optimal for the community, it could be optimal for the internal VMware engineering team, but it's not optimal for the community.

B

So uh we should focus on community more and do these changes and I've already synced that with Jolene.

B

You're dropping out actually.

A

I I started in VMware, it's just so out of the uh yeah uh maintainers I mean the harbor maintenance, so uh adjust the house in the China time zone um within uh uh we're, not work, but uh um I want to mention that all the attendees are the harbor maintenance. So uh um but the yes.

G

A

That we should we should uh evoke more maintenance outside uh when we are to join the meeting, to discuss the issues and fears.

G

B

Go ahead, I'm, sorry.

G

I think because of the amount total amount of the pr can raise, the body contribute. uh Community are very uh less than five every week, so I think we can. We don't have to book an enemy meeting to uh review this class I think we can talk about it in a community in its community meeting.

B

Fair enough, um but I was I was about to propose this one as well. um If we do you think if we can accommodate like this 15 to 30 minutes in our community meeting, but the thing is we, we know we don't know if someone will pop up on the community meetings and wants to discuss something else and practically we won't have the time and uh we we have a problem with this one.

B

If I show you some some examples, we're not paying much enough attention on on PRS and in reviews, and that's the reason why I'm I'm trying I'm bringing that topic. My intention here is not to create you more work or do you have just another Zoom call?

B

uh The thing is we should work towards nurturing new contributors and the way to do that is to address the PRS properly and in time, and the way to do that is to provide meaningful and more reliable feedback when when there is a PR or issue, let's focus on the PRS first.

B

um So, as you said, they're not enough they're, not a lot in number, which is great because we can then we can book it half an hour and then we can squeeze it to 10 minutes I'm. Okay, with this one I don't want to spend like a half an hour of these using PRS. Don't get me wrong. I I just want to provide a meaningful feedback to the interviewers. So uh back to your your remark, uh someone when is that meeting scheduled, which is the internal one with the China maintainers.

A

um It's two: uh a four pm internal time zone. Every every Monday.

B

So it's uh you say: 2 p.m or 4, P.M 4, 4, pm 4pm is yeah yeah. So practically that will be okay, also for me and for redeem I suppose in it's okay for his time zone as well, and uh so we can extend that thing practically and we replace it with the opening Community priyaging one.

B

um So um David are you in the are you in Europe or.

C

uh Yes, I mean Paris time zone in work. Sir yeah.

B

So if you, if you plan on actively joining us on this stuff, I think it would be okay for you as well, so that that will be like 10 o'clock uh or now with the time change, maybe to be nine o'clock. But it's like it's like doable for everyone, uh premier league and since we don't aim for now to include uh anyone from the U.S time zone that that will work.

B

So I really want to take a decision today, uh because we are I. Think I can show you next time, I'm not prepared now, but we are kind of lacking behind getting attention from new contributors. So uh we should work on this one.

B

um So, let's, uh if you say that one, if you say it's a four o'clock.

E

B

E

B

Can we extend it for externals, like by external side, I mean non-vmware folks.

A

um I think we should we have to uh confirm with jolly I I've I've.

B

Done that already, if you want, have a chat with him, but I am Keening on getting that done next Monday to be extended on uh on outside VMware.

A

I I I'm, fine I'm, fine with that. No.

B

Okay, all right I'm, I'm gonna have one-to-one uh tomorrow with Jolene, so I'll.

E

Bring the other thing and.

B

We'll have it up: okay, good good.

E

B

Right uh I can see we have a new update added uh demo for the application chunk.

H

Yes, so can I share my screen.

B

Yeah sure thing, foreign.

H

In the previous Community meetings, I have shared the uh proposal for replication trunk, so today, I would like to share the demo for uh support uh copy by trunk for Hardware replication.

H

So uh can you guys see my screen?

H

Yes, we can okay. Thank you. uh Firstly, let's login to cover, and then let's prepare our a fake image.

H

Okay, so I have pushed one image to the library project uh and the the size is 100 MB and then I have added the demo dot go how about dot IO as the registry. So let's configuration configure the replication rule.

H

Let's just push this image.

H

Okay, you can see a new checkbox has been added in the replication rule edit page, and this checkbox is not enabled by default. That means um we keep the same behaviors with previous version only if user want to enable copy by trunk. He needs to check this checkbox uh as the compilation Let's uh copy it uh by by blob.

H

So, okay, let's copy it.

H

Okay, the like skillshing as a job is succeed. Let's see the logs uh the the this image has one blob, so you can see the try to copy the blob and the blob copy completed and then, let's check in the demo.go power.

H

Okay, the the image has been pushed successfully and then, as the compilation, let's uh push another image and with the same size.

H

Okay, the V2 has been pushed that's replication, yet by trunk.

F

H

That seeing able this this checkbox copyright trunk and a notice is, uh if you enable copy by trunk, that may increase the number of API requests, because, uh if you copy it by block, it may need only one request to the Target registry. But if you break down the blob into a multiple chunks that may uh that may send send the multiple requests to the Target value stream.

F

Okay, let's do it.

H

Okay, let's watch the logs and you can see the uh difference between the copy by blob and copy by trunk. uh If you enable call you by trunk, uh you can see, we separated the blob into mutable Trunks and the trunk numbers depends on the trunk size that we set it in the harbor uh we by default. We have a value, it is uh 10 MB and if user want to override yet the they can that the environment variable in the hardware container.

H

So you can see we it's copy the trunk from zero to the to this value, and this is the blob size and when it's, when the, when this trunk down it's copy, the next chunk. The next chunk is, after the previous trunk and range plus one so I and the next until we transfer all the uh all the trunks.

H

So we can check it in the heart in the Target registry.

H

And you can see V1 and V2 and the.

H

The digest is same.

H

uh So the the advantage of copy by trunk is, uh if, if your network is not stable, for example, uh if you transfer uh 100 MB data, uh that may that may that's, you may uh meet the failure uh so, but if we separate it by trunks so every time the request only carry about 10 MB later so the success rate is maybe more uh higher than uh probably buy whole block and and another uh Advantage is.

H

uh If, if, if you, when you copy the whole blog, um a network disconnected uh happened in the in the period of the copying, uh maybe in the when you have copied about 90 90 MB, and then you need to re report and repost the all the whole block.

H

That means you may need cause about more more than one times uh for that, but if you uh copy the blog by trunk, uh for example, if my, if the final trunk uh failed to to copy, you only need to uh copy the this value trunk and that can save uh much more time than if we copy by blog.

H

uh Due to the, we still have some issues to fix for for this part. So today, I can't share the compilation when the when we, when we encourage the network, broken issue.

H

Okay, so this is the part for demo, the um replication trunk. Any question.

I

Yeah I have two questions too. um The first one is the default trunk says 10 megabyte if it's configurable through the homo diamo file or.

H

uh No currently, we we don't want to expose it to the hardware configure.yama or the Harvard Helm configuration, because um uh because we uh in the current state we we don't want to expose it for user configuration, but um but user can override it by inject the environment variable in the hardware container.

I

Okay, so if the user want to change size for the uh for every block for every trunk, he will have to change the environment variable right.

I

Yes, oh okay! um The second question is well what happens between two? How registry One support um replication by trunk and the other one doesn't support it. I suppose.

H

Yeah yeah yeah I.

F

H

um uh In the truth, uh the the the the the talker distributions back defines uh that's to the as the docker as the image rest registry. If, if he follows the osei is back, uh he should support the copy by trunk, but uh for the, um but we actually we. Currently, we only support enable the copy by trunk for both source and Target are Harbor registry.

H

uh uh If I mean if, if you're, if you're, if your targeted registry is not hard, uh we will not enable we will not enable this.

H

uh We will not not sure, for example, I choose the docker registry, uh you, you cannot choose the copy by chunk blob, because we we will validate that and we only enable it for both Hardware industry. But uh if user really wants to test that, he can enable this by uh Hardware API, but we do not ensure that can work.

H

We will only ensure the uh work for both Hardware registry.

C

D

You okay, um the question: why not make? Why is it an optional functionality? Why not make it default? You know like. Why do you open up the user the option to select if it's much better so.

A

D

A

We do not want to change the default Behavior for the replication, so the default behavior for replication is to replicate the whole problem there. So um we want to just introduce another option for user to handle some uh situation like with high latency in in their Network, so um yeah.

H

And another point is, uh and.

D

Okay, the second question: if.

D

If, if the spec the support, uh loop and chunk, why not supported? Then you think if not supported use block, so you know because I think it's you know it's over complicating the the users with the choice. So if we, if we make it Dynamic, so let's adapt the system dynamically to the situation. We don't have to ask the user to explicitly set it or not, because because it will work anyway in both cases you know if, if there's a probing, not testing, if it's supported or not.

H

uh Sorry, you mean you mean dynamically. What what? What do you mean dynamically.

D

Well, uh dynamically I mean we can we can probe and test if uh chunk is supported and then use Chunk in this case and if it's not supported use blob, and this way we don't have to expose the internals of the system to the user to make the decision, because my opinion for the user, it makes no difference, you know. So why would someone explicitly enable this and one why we should one disable this?

D

You know if, if it would work automatically in a sense that there's a probing that can decide uh or make a fallback to long time in case no, then it.

A

Will always work yeah, but uh based on the uh distribution spike, it can now support this kind of scenario. I mean that if you decide to replicate the whole block, you can't fall back to replicate by Trump. But if you decide to replicate by trunk, you can now go back to the the whole flower. This is. They are the two different kind of scenarios: um unfolder. We Define two phases for trunk replication, the in the phase one. We just want to support Harbor to Harbor um in phase two.

A

We want to extend the capability to mount the more other kind of adopt adapters like a dog harm, uh GCR or a and Etc, if, if they support the way to uh version of a distribution Spike, so then we can uh extend the capability to um other kinds of adapters. Yeah.

D

B

Thank you. Anyone questions on this one.

B

Is that um I have one um we're aiming to get this one into what two seven.

A

B

Yeah, okay, and do we have.

A

Yes and we have already submitted in the community, so we have a down two run: the review to the proposal, um um but uh um yeah. Yes,.

B

A

Ahead, go ahead.

B

Yeah I was I was curious. uh We have the proposal um already approved or.

A

Yeah yeah, yes, and we have already demonstrated in the community meeting.

E

A

um But the country we we in kind of one problem in the distribution side, and then we we are um working on debugging on this on this may be a risk for this feature in 2007. um on.

A

We maybe have to update the some source code for this distribution side um to fully support the trunk replication um yeah. So uh so I have to say that there there should be a risk for this feature in 2007.

B

uh Does that mean we have to update and fix the proposal as well, if I get it right or.

A

I mean uh in the social level we encounter some a one problem today um when we may debugging the the source code.

B

A

B

Is it architectural problem like from The Proposal itself.

A

Or no, no, no, some incompatibility between the it's a idea- and this was called uh it's an absolute Upstream problem.

A

um We we probably have to update the Upstream source code to fully support the strong replication, um yeah and I will talk to the distribution containers to uh to to work with them to to have a talk on how to modify the source code to support the the way to uh API.

E

A

So there is a risk for this feature in 2007.

E

B

Is is a is Roger aware of this? One I mean: uh do we update the the robots.

A

Yeah so far, not um I.

G

A

You uh uh let him know, and tomorrow.

E

Okay, gotcha right: okay,.

B

Anything else on this one, someone wants to add something: I have added three more things: I'm, sorry for that.

B

All right, if not um I, saw one of the uh tests is failing. It's called similarity test case, uh even though some PR that are not related to any part of this code and I saw these PRC over here that are removing that text.

B

uh Anyone can comment to this one because it says the singularity has been renamed to uptime or wherever the name is, but I can see. Only the removal of the test.

A

um Yeah we will replace the The Singularity with the new I, don't know a ABP trainer on yeah, but the temporary. We we want to deceive disable. This case.

B

So the the thing is right: now we don't have that app Diner or whatever the obtainer obtainer.

E

B

Called in and we just want to remove the singularity one not to fail.

A

Yes, so then we will have another case for of app trainer.

E

A

B

Talking about tests I want to share something. uh Give me a sec.

B

I've added the so-called fossa, so that's the the security scanner I think I added the budgets on the on the on the on the main page and that that's not the important thing over here. But the thing is: uh when you have time. First of all take a look at that PR, please and then second, um we have quite a lot of things to to fix.

B

That's a a mandatory scan license scanning from cncf, so we have to accommodate it and then work on maybe resolving this or uh we'll see how to proceed from all this. So far the the license can is passing so we're good. We just have some findings that we have to figure out.

B

um Yeah I have to add this one um I have a a mask for home interiors, please, watch and and monitor the maintainers channel in in is like I'm posting some stuff there and I think nobody's ringing that channel. uh So please take a look at it and last ask from my site is: if everyone can think of some cute first issues, so we can be more welcoming to the new to the new members of the community.

B

That's primarily asked for the maintainers to figure out I know it's superintendent attempting to to have an issue and just fix it, but better create an issue Market as good first issue and and leave it into the open for a few weeks. So we can have uh contributors able to join us.

B

uh Let's ask for everyone pretty much I know it's super tempting to fix it like in two minutes, because you know what to fix. But if you spend these two minutes explaining that an issue, then someone else can can pick it up and enjoy working with us on the community.

B

um That will make our life much easier because that will be bring more folks, yes, and that remains more less quality, maybe PRS. But that means more folks able to join us, and some of these new folks will be able to become a maintenance at some point which we mean. Let's work, let's put it this way, so yeah I, hope you understand uh what I'm saying I can elaborate more, but I don't want to waste time. So uh without I, don't have any other topics.

B

Anyone wants to add something for, for today,.

E

B

Right, hopefully, quiet. Thank you.

D

I have no topic for today.

B

Was that I didn't get it.

B

F

B

D

I have no topics for today: okay, okay,.

B

All right, I'll take the The Silence from from anyone else, is a mature agreement that we can end for today. All right, thank you very much have a rest good for the rest of the day and have a good evening for the channel folks and talk to you in two weeks and bye-bye.

E