Cloud Native Computing Foundation Network Service Mesh, 15 Oct 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Network Service Mesh Meeting 2019-10-15

Description

Network Service Mesh Meeting 2019-10-15

A

A

B

Welcome folks we'll get going here in a few minutes.

B

Usually we start about the 5 or actually probably in about a minute. We usually start of a 5 after and Frederick has laryngitis today and so sadly won't be able to lead us in a sauce. Just a fault stand-in.

B

I've gone ahead and put the link in the chat for the meeting minutes. If folks could go ahead and add themselves to the attendee list, that would be fantastic.

B

You folks hear me.

C

D

Yeah I hear you.

C

B

If there's always that uncomfortable moment, when you start talking on a call- and everyone is being politely, quiet and you're, just really not sure if you're being heard.

B

All right, thank you for sharing I, appreciate the person's hearing, the agenda and it's five after so, let's go ahead and dive in so again, please add yourselves to the meeting attendees list and we'll drive into the agenda. So we do have this recurring meeting which you can get to on the community page in addition of CN n CF telco. Your tub also is a group that we work closely with and we should probably take a look at the CN CF networking working group, that's currently being rebooted, so major upcoming events.

B

This actually is now a past event. The October, 2nd webinar I I know that we've been tweeting about that and pointing people, the video it went very, very well. It was extremely well received, but I'm gonna go ahead and release that this is no longer an upcoming of remove s. This is no longer an upcoming event.

B

We also have coming up the open-source summit in Lyon, France and Ivana and rata Slav will be talking they're, presenting an introduction to NSM and there's also a toko introduction to MSM um coming up November 18th through the 21st, that's kubernetes q. Come we do have a talk on the main stage there for five cool things you could do with network service mesh. In addition, an SM comm is there.

B

You should go ahead and please add that to your cube, con registration, we do, you do need to actually be registered and it is filling up very, very fast and we have quite an impressive lineup of NSM related talks, so strongly encourage folks to register for that um and then coming up in March q, con + cloud native con Europe is coming up an Amsterdam. The CFP is currently opening closes November 22nd, although they've now updated that to December 4th, which strikes me as wise.

B

Closing your CFP in the middle of your running conference is probably not the greatest thing in the world. I.

E

Heard that they have moved that? Yes, yes, it was a bit okay, yeah.

B

No I'm sure someone did very same calculations to arrive at some executive. Then people realized oh, wait.

B

Maybe it's not the best choice and then Taylor has actually put together a CFP list. If folks would like to be listed there and then one of the other things that we will commonly do for CFPs for things like cube con is will often put together a Google Doc that the community can collaborate on for crafting a series of proposals to go in yeah. Obviously you don't have to participate in that process. It's primarily there to be helpful and provide support.

B

Awesome cool so the next up we have an update from our social media community before.

E

That, if I can just just throw something in here post them, there has been some some exchange about this, so I don't know. I I was trying to evaluate if this is something that we want to attend. What kind of event is that I think that is kind of generic open source event, but yeah, but.

B

It's a very dirt under your fingernails kind of event. Right so FOSDEM traditionally is an event that is really hardcore hacker and it's also a fairly large event. It's got about 8,000 people involved. Everybody I have not been depressed to fosston personally, but everybody I've talked to who's ever been to Faust. Them has been super happy about the experience.

B

It's it's a pretty cool community, so you know that you know and and the the folks who run the SDN room they're the associated networking room are actually interested in seeing or service smashed represented there, so they are interested in having this um network service patch talks.

B

So and it'd be.

B

Yeah, it's the actually, the software Griffin networking room is actually who's reaching out to us.

E

B

So but I thank you for reminding me to add that.

B

Okay cool, so thank you very much anything else on events before we move on to social media.

B

Awesome so Lucina, how is social media going this week? It.

F

Was it went really well yeah about ten, more followers than we've been having per week, some 23 more followers and followed almost 40 more accounts and tweeted retweeted about 30 times was able to post the NSM con lineup thread. So that was a 13-part of thread.

G

F

Every time bracket and as many of the speakers that had accounts- and that was well received, the speakers primarily liked and retweeted their their reply or their topics. So that was good and.

H

F

Posted on the five cool things at cube, con sent a second thanks again to their sponsors, and that was well-received as well, and I also pinned that thanks again for the NSM come sponsors to the top of ours.

F

So that's the first thing people see when they go to end service mesh and also posted a reminder of today's two calls and retweeted a few of the VMware open-source blogs. This week the telecom TV interview came out yesterday. I was out of office. So take a look at that today and this week and promote the session at OS se, you reminder to registered anis emkin, promote the individual sessions and speakers at NSF outside of the thread. I'll do a spotlight on each speaker.

F

Maybe do two or three a day for the next couple of weeks a week or two and promote again the session @q Kron we're almost at 500 follower Shore. The Twitter account so try to post something when that hits almost at 300 stars on the github account so post, something when that hits.

B

F

I haven't seen a video of the ONS 5ji panel, so if that exists, please be sure to ping me if I, don't, if you see it before, I, do like to share that and I've already retweeted the O ons you keynote so remove that one and contributors podcasts is that available.

E

Tomorrow, oh.

F

Okay, great yeah.

B

So I don't know what it's going to be rebroadcast, but we do have the we do have it scheduled actually for tomorrow.

E

But we we had the CN CF webinar.

B

The webinar is actually definitely out, we've been doing quite well actually on the viewership on that.

F

I can share that again if you'd like.

E

B

Yeah, so basically, let me see what the viewership is like on the webinar. This point we're at 600 views on the on the webinar video already in two weeks ends so we're doing quite well, they're awesome. This is going well, one other thing I did want to bring up. We also have a LinkedIn page for network service mash that we haven't really done much with I'm curious if folks are interested in doing stuff with that.

B

I realized this, because somebody actually posted a really nice post on LinkedIn about finally understanding when that work surface much was about after the webinar and so in that actually got a huge amount of traction. So it occurred to me that LinkedIn might be an undiscovered country for us.

B

But we can sort of figure that, as we go, I just wanted to throw out that that that resource is out there and if there are folks who are interested in working on it, I can definitely make the keys to that available to other folks, as we expand our social media community.

B

All right awesome.

B

So next up is the announcements. They don't think we have any announcements this week so moving on, we do have an attempt to capture a good beginner problem, so we do have these fuzzing bugs.

B

So, if you're new to the community you're just looking to get your first commit into network service, mesh I'd strongly recommend looking at these because they're pretty clear about what the issue is, we're intentionally leaving the Mun resolved for the moment so that they can hang out as good beginner bugs, if need be, we'll probably just fix them going into 0.2 release.

B

Cool any other good beginner bugs that people are aware of they want to list here.

B

All right excellent status of the project, so we've got a bunch of things that have landed this week. You may recall, for last couple of weeks we've been talking about a shift in the API and in particular we previously it had the split between the remote of the local API and we brought them together into your bringing them together into a unified API, so that we have. You know this vastly simplifies the world, and we talked about that for the last couple of weeks. There's a whole slide deck at the API discussion link.

B

If you want to go and review that conversation and think what has landed this week is we had the break up of it into three patches. The first was the API itself, which broke up the local and remote network service API and brought them together into a unified network service API.

B

The second is compatibility helpers. So basically we wanted to make sure that we had some helpers, that we convert things between the remote and local api's and the new unified API.

B

It turns out that was relatively simple and straightforward, because the API is are structurally almost identical, but this is that patch is going to be super important if you're migrating any code that you have, it doesn't exist in the network service, mist repo, because that's exactly where you're going to go and grab any compatibility helpers that you may want to you apply, and then the third patch was applying those adapters throughout the codebase and again, if you're looking to adapt existing code to the unified API.

B

This is also going to be a super important patch to look at because it applies those adapters all up and down the code base and effectively all in ins up doing is wrapping any servers and then applying a wrapper to any clients. So it's a very structurally simple patch to a thing to do, but again, if you're looking for a step by step, this is how you do it go. Look at that patch it'll be super clear, also landed this week we migrated to using github packagers instead of just errors.

B

The reason we did this we migrated be using these everywhere in the codebase, because when you do an errors new, it will capture the stack trace and also it allows you to wrap errors. So if you've got a returned error and you're going to return an error yourself with additional information, you can wrap the cause, and this ends up making the code enormously easier to debug as you're going through and trying to figure things out, and so we moved made that movement and we also put in some very basic CI.

B

So if you try and use just quote quote the open quote errors rather than give up packagers. This see. I will complain to you and likewise, if you try and use fmt format errors instead of using errors, new or error set RF, then it will also complain to you in the sea I just to make sure we keep the emotion consistent.

B

Any questions on that.

B

Cool um other thing that we brought in this week is, you, may have been following the last few weeks. We had a shift in the build system to fill the fill containers and, oh, my god, that it speed up, build containers, I, don't know what other folks experiences have been I would love to hear them. But for me on my local box, it took a cold build from 15 minutes down to a minute. Has anybody else had similar experiences.

B

Okay, yeah I know I I'm I'm, almost the very model of optimal build conditions over here. So you know I, that's good and then the the other thing about. So we migrated to having the sea I used that build containers and we're now to the point where the sea is able to build all the containers in a single job at about a minute which is really quite good.

B

And then, finally, in landed this week, we didn't really have much testing the ICMP Colonel foreigner in the sea, I and so I know. Denise was mucking about with four things and felt it was important that we not break the kernel for it or and so contributed some helm ICMP tests to the CI for the kernel toreador. So many things Denise.

B

Anyone else have anything they want to add. That's landed this week.

B

B

I somehow lost the line for in progress in here. Let me go put that back.

B

Okay, so in progress, this is stuff. That's going on right now in the codebase, so the API stuff continues now that everything is shifted over to the new API via adapters. There's some work afoot to adopt the SDK to use the unified API directly that hasn't quite gotten together. Yet, in fact, it's currently failing all kinds of tests.

B

It should be pointed out if you are using the SDK, the SDK itself is using the adapters, so they should work with the new API just fine, but it would be good to route out the old split API and actually start natively using the unified API. So any any questions on that. As that's going on, you know, there's a lot of movement on the API stuff. Is that it's clear to folks? What's going on there does it? Does it feel comfortable? Any sort of comments on that.

B

Sounds good cool and then, of course, once the SDK is been adapted to using the unified api's we'll want to migrate the examples repo over to using the unified name, guys yeah cool, all right, another thing in progress, so Alex: do you want to talk about Azure pipelines.

A

B

Could you could you stick a link in here so that we can go, take a look and see what does our pipelines would look like you.

A

I'm on the phone, okay.

B

That's completely fine.

A

Hosted you today, yep.

B

Let me go grab that real, quick, hang on. Thank.

A

B

Yeah I know it's crazy late for you guys over there.

B

Let me go ahead and get this link up and I will put it in the chat. Actually, I only can even do one better. I will put it on the meeting minutes.

B

Cool excellent yeah, so um if you could click on the example to clean this or run just so folks can see it here on the call briefly.

I

Do you mean this? Well, yes,.

B

Okay, it's saying the requested build could not be found. Is this perhaps something that is not publicly visible yet um Alex? It's.

I

Possible that I don't care about this yeah.

B

We need to figure out because we do need people to read the general world to be able to follow the links and get access to their builds. So we will lead to figure that out not.

A

Stranger, you know yes, I haven't seen that screen before now. With now, I can see a screen and it shows the pipeline's can we do. This is not.

C

B

Okay, I may have to sportful lengthen, so you can see all this running here. I mean you may want to go back to the builds and click on a red one. Just so you can see what an example of a failed job would look like, because that's in some ways, Phil you're good failures, but it is more important than good success. So it basically says: okay, we've got the sanity check.

B

A

This is a d-bag do background, so it doesn't really helpful to introduce a we are failed. We may pick up and even earlier, build to see actually failed yeah, for example, cloud chest.

B

Okay, so, but effectively, I think know what we've basically got is we've. Let me fix the link to your meeting, so we effectively cooks to go ahead and cook at this little bit and see enough thinks the link of the meeting minutes see what kind of what it would look like you know, alex is trying to move to a place where we can transition one of the nice things about this is that the Azure pipelines folks, are donating a lot of time to CN CF, so that is very kind of them awesome.

B

Should we go back to the agenda.

B

Cool so there's a little bit of stuff as a work in progress on adding go header lenders. Do you want to talk about that? Denise? Yes, sure I've.

C

Provided God also I've provided demo inside there, so please take a look and that's it. You.

B

Know the the basic idea here being that we're not always great and I'm, actually the worst about remembering the license headers in our files, and it's quite ironic that I'm the worst about this, because I actually understand this, really why it's actually super helpful to people downstream and so the go header linter as I understand it would simply fail the CI if you added a file- and you did not add a license header to it, just as a general reminder that you need to include your license, headers so good housekeeping.

B

So Andre, do you want to talk about the network service mesh, the never source manager SDK like refactoring, which, like it's getting very close, yeah.

C

It's very very close the moment to just one in stable test, and it also could fire one master. So I'm mostly finish it and I hope it will blend the twist way, and it includes some additional improvements in different areas like cloud testing tool and retrieving of the logs, so in general, should be very positive on ability, because during this refactor ink I've found and fix it.

C

Few different issues with the healing and if establishing contact with monitoring of remoteness, MDE monitoring of data plane and how aliens are passing through system, so in general, should be more clear and more stable, excellent.

B

That sounds great, very cool, yeah, I refactoring, often I find little bugs as a refactoring. So that makes total sense um so Denise do you want to chat a little bit about the ethernet context issue? Oh.

C

Yes, on last week, we face it this back, we waited to be agent. Buck was related to incorrect behavior with dump request, and we have updated with we agent last week and this new aviation dump request. Working, fine and I have continued to work on this issue. That's all cool thanks.

B

Lip I don't run a slug. Do you want to talk a little bit about the kernel, fording plane, I'm.

C

Sure you happen after merging the matrix support, basically I'm involved more on creating, for example, for for the demo for the open source. Amidst all that we have with Ivana so yeah in terms of features there. There are no more features: developing yeah, okay,.

B

So do keep an eye on things. There's all the refactoring is going by I want to make sure that we don't stabilize the stuff. You guys want to show so do check and make sure that master is staying stable and do also make sure that you know any testing. You want to get into keep the stuff that you care about stable, gets in as well, because if anything goes bump in the night, we're gonna get it fixed, ASAP, okay, so hopefully we won't have any of that.

B

But but we all were all adults we've all been through this before we all know how the world works so cool.

B

Then we've got an in-progress thing that you put together really quickly to these for renaming data playing to foreigner. Do you want to talk a little bit about what you were doing there.

C

Oh I just provided mu, proto and I just wait for an SMS the collective acting because I need to change the same code.

B

So this is actually a question and since we're in the phase of refactoring some API stuff, it makes sense. Do we want to go ahead and do this rename of data plaintiff order to match things? It will mean, for example, having to go fixed examples, because various downstream imports may shift that kind of stuff. My my instinct is if we want to fix this, now is probably the right time to other folks or thoughts. Oh.

C

I think we can provide I can provide to class one two and a same report and another example.

B

To other folks have opinions or thoughts, particularly folks, who are downstream working on things.

C

It's better to do earlier than later. We, if we do so dramatic renaming and refactorings yeah.

B

I kind of agree with that.

I

Just yes, if there are some minority PR that I waited, maybe they can go slower so that they're not yeah.

B

You know something you have in mind in that regard. No.

I

Just think, okay.

B

Yeah cuz I mean do speak up if you've got a PR that you feel she going with some priority because I I'd liked there are a couple things that I'd like to avoid. One is I, don't want people to get stock, and so just pipe up pop up on a slot channel. If you feel like you've got something that you think should going with priority, the other one is I, don't want people to get stuck in sort of having to rebase entry base a tree base, so you know we'd like to avoid that as well.

B

I

Alright have fun those talk with the matrix observability that I'll talk about them when we reach there. So, okay.

B

Next and then again, Denise you've got some work going on with some support: multiple simultaneous letters and borders. Oh.

C

Yes, it is in progress. This also depends on and Sam is developing.

B

Okay- and this is you know- we've had this issue up for a little bit of time. Talking about how to support multiplying multiple simultaneous borders. We still have some interesting open questions about how you handle the order and priority, but I know you took a look at the spread of Slav and other than the order and priority question. It looked good to you correct the the suggestion for how to handle it. Yeah.

C

B

You keep an eye on the PR, because what I'd like to be able to do is I want to make sure that we can deploy. You know both the colonel for tour and the BGP a shoot for it at the same time, I having sort of lis hard configurable switching between them is less than optimal, plus, as we start getting folks working on the hardware Nick solutions. Those are also going to require Anna supporters, so alright cool, so security Ilya. Do you want to talk a little bit about the things in progress there? Oh yeah.

A

Sure the first one it's introduced, tokens I, think discuss it last meeting.

A

There are several comments, after of you, I think I can fix them in a while, but we also decided to merge it after and it's like SDK like in a semi fracturing from injury and the second one is really just require rebates, as far as you know, enable security and, inter demented.

B

Cool, so, if on, if you want to talk about metrics and observability, yes,.

I

And the moment merging prometheus integration is blocked because it first requires merging the the other PR that exposes both names in the cross, connect because for misusing them and the other PR is blocked. Immersion by the agent dependency show, but I'm trying to resolve for a couple of days. It's breaking the CI.

I

B

Deviation 2.3, it.

I

Was yes, it would 3.3 and 2.3 I have the same issue as I had before it's when I update with golden rate or demote, pals DC enemies are there and then they make format is failing, I can't reproduce locally and if I expand by removing that fine and if I remove it gotten away they're in the CIA accident, then the different note on the right face and it's an infinite circle. I tried one workaround.

I

B

I

Forgot which one, which is, let me check- apologies.

B

I had originally out of your PRS back before it just started, copying and pasting no.

I

It's just so, it should be here.

B

I

I think it should be 16 17, okay,.

B

So yeah rich out of her over slack and let's, let's see if we can get you on stock because I very few things actually bug me. But the thing that one thing that does about me is people being stuck so I try and figure out how to get the one stock.

I

Okay, I think I've reached out last weekend, but we couldn't find any solution discussed the update of to not free, but still we didn't hear so well.

I

B

I

Walking this one because no meet need is merged. Yep.

B

No I totally get that and I make sure we got some block because you've got cool things. You want to show off your shortly so yeah.

B

Let's definitely follow up with that on slack and we'll go ahead and figure out what, if anything, could become there because I'm very excited to get the metrics and observability going cool so on SR v6 support. So this is stuff that our team is working on. I think Arden is currently on vacation, which is good. It's good for people to take a little time off and the PvP fix that we needed for SRB. Six has finally gone in upstream and so I'm currently talking with the VPP agent guys about getting a stable, reece pin.

B

That includes that fix and be bpa trend, and so hopefully, once we get there we'll be able to guess our v6 support working so cool anything else in terms of in progress that folks want to discuss.

B

Alright cool anything else that folks would like to discuss that isn't on the agenda.

E

I see weighing here.

H

How's, the guys tell you hey.

H

Thanks man, I mean to make the breeze meetings would be tied up with them some other stuff lately. So I'm yeah kind of just pause. My way and yeah I changed you guys trying to help out I'm, actually looking at so most of those to fix those issues in.

B

Trying to make beginner-friendly issues and those of the linter with the buzzing tend to be good. For that. The other thing is the the kind gentleman from the fuzzing community became and started buzzing things I think once those start getting fixed, you'll, get enthusiastic again and start buzzing worse off. Okay, cool yeah feel free to reach out over slack.

B

You know, or talk there or here or wherever about kind of what brings you to the community? What you're interested in, because if my experience has been that you know it is absolutely God's work to roll up your sleeves and help with things like blessing bugs, and you know documentation of that contribution it, but getting people lined up with stuff they're passionate about and should be a long-term sustainable path, so figuring out. What really excites you here is gonna be very helpful.

H

Yeah so I mean I, don't ride a little girl. We maybe Alexia and Python shops is a bit of a loner, so just bear with me a week into, and my engineering team, as always came to that's.

B

Fantastic, the suggestion I would make there is don't hesitate to ask questions on slack because there's there's a lot of situations where you take something out and if you just ask someone can point you to exactly the answer in two minutes.

B

No-Go is an excessively Pleasant and logical language, but it does have a few myths in it. They say you gotta, wrap your head around yeah Kenny.

E

And when you, you send me a question about enabling I'm. Sorry for not answering you just easy know couple of days, so maybe maybe you can post this in the NSM dev channel so that you get like because it's more or less than VPP common. That needs to be there, but probably people will be able to us or.

H

E

H

I'm, just like and they've got another use case, I made you love news cases, so one of our things is reconnects by V pin into the stalkers in Africa and what you're trying to do is instead of having a boundary based, B pin like a new secure instrument. Example it what inbound weekend right where we Dowe we in a connection into and the telco and then we use it as a pipe two-month in shape and have apples on in the CNF, as as we push traffic through this, the pods. So it's an awesome.

B

Oh okay, that's interesting, so you were essentially running the reverse of saris, exactly yeah.

B

I, don't know what time zone you're in a lot of folks.

I

H

I'm Jim, cheapest, juicer and I'll. Send you guys a picture of the of the use case, and maybe we can use that as the discussion excellent.

B

Okay, all right.

B

Otherwise, I'm inclined to yield back the time, but it folks there's something else they wanna discuss. This is absolutely the time for it.

E

Do you want to share because we have some time here? Do you want to share the picture that you're trying to implement I think that Taylor might be interested yeah sure? Let me just.

C

Try to share my screen.

C

Yeah, basically trying to recreate them the network topology of a 4G network I'm.

B

Not seeing it sure, just yet yeah.

C

Just trying to find it.

B

Sharing we're all good, no.

C

So is it showing yeah.

D

C

C

Yeah, basically, this is the topology of the network that I'm trying to recreate with metal surface mesh, so I I'm, basically recreating the picture that is widely available in the Internet, but I'm using GSM concepts. So yeah you can see. I have I have chosen some of the blocks to the endpoints. Some of them are clients.

C

The idea is to have a single client and a single endpoints and not have a points that is both a client and an endpoint, so it was successful actually until I rebased, like this master.

B

If I want to make sure we get that cleaned up you, it may just be that you need to use the adapters if you were base this morning, for example, yeah.

H

B

So so ping-mei will get you hooked up with the adapters and that should fix you plus as soon as I can as soon as I land, the the native API stuff in the SDK I'm gonna go ahead and shift over examples as well so but I'm delighted that you've gotten this working, that's actually really exciting.

C

Yeah, basically, it's just a skeleton of this picture for the images of that of those blocks I'm using my opine images. So it's it's it's up to whoever tries to implement that to change the images, but the clients and the endpoints are configured to to create those net. Disney.

B

North America, which is.

I

C

Some big round boys, someone yeah.

B

That's exciting. It's also probably a really good talk proposal for something like ons North America, which is open again see if decent.

C

Yep, hopefully also oh, you.

B

Want to see exciting telecom operators coming in and saying here's the skeleton of your 4G with that resurface match that's exciting.

E

So Taylor this is also partially meant to help with CNS test bed and the the idea about you know having something to be demo'd based on the packet towers and etc. So what we wanted to actually show here is how you can use natural service mesh to interconnect all those boxes.

E

We don't have the content of the boxes yet because we are not that deep into the telco stuff, but we assume that, if someone capable of understanding how the open air interface or some of the other open source projects that will implement these things can you know take over and then they can feel that the right services, insight and input put the things in place. But this is like just saying yeah if you wanted to enable these solutions based on anything. This is how we will do it. So that's more or less. The idea here.

D

That's great and this.

H

D

Interested in following happen- and we can see- maybe we could actually implement this on the packet locations with the sprint 4G connectivity.

B

Very very good.

H

This diagram is really excited for me as well, because.

B

H

Of the stuff we're trying to do there around the PCRF and the packet get right.

E

Okay, I didn't so who who said that, but I was like okay, so if we, if you have anything that that is open source and confuse some of these boxes with some geo content, would be we'd, be glad to do it further and try to put something together that that can be, you know, can do some real work, not just you know, caving beans back and forth between boxes.

H

Okay, yes, for sure.

G

B

That's awesome anything else folks want to talk about before we conclude.

B

And in by the way, Bratislava don't be shy about putting things like that on the agenda that would have been dead center of in progress right. What? If you oh I.

C

Was I was actually disappointed that it broke just before the meet I'm.

B

So sorry, like I, said I'm pretty sure it's just a simple matter of grab the adapters in the way you go, there's a literally an SDK compat now and- and so if you go look at the PRS that are actually in the meeting minutes, the one about how to apply them is probably going to be super instructive.

I

B

Right well anything else.

B

All right talk to you guys later yeah.

C

Bye, bye, Avon.

E

I

Mike Cheers.

I