Cloud Native Computing Foundation Service Mesh Interface, 11 Nov 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: CNCF Service Mesh Interface 2020-11-11

Description

CNCF Service Mesh Interface 2020-11-11

A

Your fingers very good, okay cool, so welcome everyone to our bi-weekly smi community meeting. uh If you haven't yet, please add yourself to the attendee list, um and with that, do we have any anyone new on the call, nothing for sure yeah? Do you want to introduce yourself.

B

Yeah, yeah, hey, hey everybody, um my name is matt yakabuchi. um A lot of people call me booch, so I think you might have seen some of my messages out there. uh So I'm working with uh nginx and so we've just released nginx service mesh. It's now public. It was a little bit um behind a wall for a little while, um but now that we're public we're trying to reach out to the community and and be and be better members, we've got all the legal stuff smoothed out.

B

We're able to contribute both to smi spec and smi sdk go so that's all good. The whole team is is ready to go and from an nginx, bu and f5 perspective. We're ready to engage. So I just wanted to say hello and um I'm going to try and be a semi-regular member to these to these meetings.

A

That's awesome very warm welcome, and it's really great to hear. Do you maybe want for next time or whenever you plan to try and again uh give a quick demo or or walk us through a little bit.

B

Yeah yeah yeah, that sounds uh that sounds great. um We're still trying to work through. Some of um you know our understanding and we want to collapse on certain understandings around the specifications. So yeah yeah I'd I'd love to give a little. um You know five minute presentation on it.

A

Awesome, and can I also ask you to if you click on that link um that both lee and also bridgette now again shared uh to add yourself there just saying your name and.

B

A

B

A

Prefer. Thank you.

C

Awesome cool welcome, excited to have you and congrats on the launch of the service match over at nginx yeah great. Thank you. Everybody.

A

Did you say that the engine x mesh is already smi compliant or you're working on it.

B

It is smi compliant yes, um and so uh we're up to all the latest versions. We're shy on some of the smi metrics. So we don't support every uh resource. That's called out, um but we have stateful sets. Daemon sets deployments, pods and name spaces I think is, is what we're limited to right now.

A

That's excellent and let's make sure that you are listed on on our um web page there right. Yes,.

B

And I do want to get listed. We have a little bit of an embarrassment, though, uh with circuit breaker in rate, limiting um there was a miscommunication or misunderstanding with the team on how those should have been namespaced. We do want to talk to you guys about proposing them going forward. um It always should have been in our organization, so I don't want to communicate that until maybe next week um I could send you guys a pull request.

B

Whatever the method is, but our next release that should be out has that cleared up, and then I want to move forward with saying this is good.

A

Awesome perfect thanks for that yeah! Yes, oh yeah, richard! Thank you very much. She just shared um the link up to add your logo, the fragment in the in the repo there. If you just follow that cool thanks a lot.

A

Yeah, I will move mine the open telemetry towards the end. Unless I see someone near.

D

I like how you have the humility to skip the item I added to the top, which is hey. We didn't actually talk about it at a meeting uh in we. We did talk, but we didn't write it in the notes. So I'm putting it for posterity that michael and teren are spec maintainers.

A

Sorry I I was so focused on what you, what the described there uh around.

E

A

Introduction of nginx that I I didn't, I didn't really see that I, but yes thanks a lot um I'll just make a minor correction um and awesome. So are we good to go to tyrone's.

A

Action sorry agenda item around dsmi, matrix images and charts to github.

D

If anyone else has insight about it, he had it last time. So we pushed it to this time and he isn't here today, so we could push it further, but I don't know if somebody else already knows or has enough context to talk about it.

D

D

A

Then we have from yourself bridgette regarding the upcoming schedule, question to cancel essentially next time.

D

Yeah, so um the meeting uh november 25th would be one day before us thanksgiving and even with the pandemic.

E

D

Certainly have like family obligations around that time. I know I'm off work that week um because I took vacation then and then the meeting december 9th would be fine and then the december 23rd. One again, I think people's vacation schedules might might be such that it would be very difficult for us to have an effective meeting. So my proposal is, we meet december 9th and then we go back to normal january 6th um and.

F

um Michelle isn't here.

A

D

That sounded good and I'm just curious about.

A

D

A

That that makes sense, I'm just wondering if next week, kubecon week, if we also because I saw usually.

D

Six cancel their meetings during kubecon because everyone's so busy with kubecon.

A

Right but we do have, we do have these topics so exactly exactly so. Maybe if, if anyone uh like, are there any objections regarding the cancelling 25th and uh kensington 23rd december, um and essentially we have then more or less one more meeting this year and then back to normal january 6 right, that's the bottom line.

B

Correct sounds good to me. I mean anything that is pressing can be handled via slack in channels.

A

Sure it's like github issues. Everything goes on just the bi-weekly meetings, um as suggested by brit. Okay, then, let's record that uh 25 is cancelled. 23Rd is cancelled, which means we have one more meeting in 2020., it's more more than enough for 2020 2020 doesn't deserve that much anyways. Wasn't that good um cool! uh Then we get to these uh good news regarding next week.

G

Yeah um yeah yeah yeah uh good, good news. The um what brizzy has written here is accurate, which is to say, I think it yeah it's about a 45 minute long slot on toward the evening on this coming tuesday for well asmr. You know all things smi, I guess the so everyone here is welcome to come and either answer questions or ask questions we'll see who all um shows up to so it would be assumed that uh we would join um of the ones that I've hosted like this.

G

In the past, um it has been helpful to generate conversation to just to have a cup a slide or two or four or or a deck in the background that uh that when people don't know what questions they have, you sort of uh bring up a slide talk about something that's new or that that's going on, and that generates more questions, and so uh I guess yeah. So I guess the thing here is um those that are interested in inserting a slide into a deck.

G

um I guess I didn't realize I was signing up for this until we. uh This just happened, which is hey I'll, go put up a I'll go grab a I'll go put up a couple of slides and put out the link here, invite folks to um yeah.

G

I think my words are failing me today, but uh you know where I'm going with this, uh so some it would be good to yeah. You know just the general just the stump speech so to speak, um slides also, if that we've got relatively new things coming up or or things like nginx service mesh being um sort of yet another.

G

um You know compatible implementation.

A

And just in terms of the the logistics, so all that people need is this cat dot, co, fmyb. There is no registration or whatever just go there and from there they get onto the zoo.

G

There was it was, it only took so many people's names. I filled in as many maintainers names as would fit on the form, and so.

A

G

A

When you know imagine I'm tweeting out hey, you know if you're interested in smi come and join, what url would I use? Would I use that or.

G

uh Yeah, um that's correct yep at that url there's a pre-published um zoom link, so I don't know that entirely how that works. If people actually have to register for kubecon to be able to get into it, but but there's a cncf.

D

Is handling it but at least last time it didn't gate through the entrado system. So I really doubt it's going to have anything to do with kubecon registration. It didn't last time right. They added a little.

A

Bit more structure.

D

Around it this time, but.

A

I clicked on it, but I'm locked in there- and I see you know- join the service mesh interface office hours here and then there is a zoom webinar link there. I'm logging out now, let's see if that's still the same okay, to try again, but it looks good to me.

A

Yes, so even if you're logged out can confirm so we can simply use that to promote it probably beginning of next week. Otherwise people forget about it. Anyways um cool sounds good, that's um definitely something. Hopefully a lot of people will see and come anything else. Regarding.

G

This uh I'll I'll put a link to the deck, probably I'll go grab a couple of things. Lockhee and others have certainly got a lot of good fodder out there. For, for it and I'll ask for feedback and and massive uh editorial review, I guess it didn't have to be much. It really isn't the presentation it's more. Just like hey did you know when nobody has a question, it's just.

A

Correct cool all right, so thanks.

E

A lot again lee that.

A

That was really super helpful that you took took care of that, and hopefully a lot from our end and also the outside will will join and we'll have good conversations there, and I I believe it will be recorded right. So people get a chance to.

G

Yeah, I think so I I'm yeah my right. My assumption is as good as yours that, like they're all recorded and and part of uh yeah yeah, I guess.

D

They ran them like office hours last time and I don't think I published them so your mileage may vary. Okay, I don't think they published the office hours last year or last kubecon.

A

I just imagine that like for me, for example, that's 10 pm, that's still, okay, but if you're like central europe or you know even more eastern or I don't know for asia, it might just be too late, so people might just want to watch the recording right. That's why I'm I'm wondering cool um then thanks a lot again and let's move on to the next one richard you put that down, but that's.

D

I did and I saw john joined. I know michelle wasn't able to join this call today, but john, I see michelle extended it for one more week um seeking feedback is there anything people should know if they're gonna fill that form out to uh offer smi metrics feedback.

E

um I think the doc does have a pretty good description of what we're looking at for that. So I would just take a look at the.

A

Dock. Okay- sorry, sorry, I didn't mean to interfere with your subscribing there, because now that actually makes sense to handle it here. I don't know if you had a chance um in that context. The the um open telemetry issue 199.. So we got someone from the open telemetry community just in there commenting on it offering. So if you click on that 199, you see. I commented on that. I had a conversation with them over on gitter. They still using gitter.

A

I guess moving to slack as well at some point in time, um so I would think that as soon as we have wrapped up the previous one that the google, dr john shared, um we might be able to um get back either invite justin for a conversation or or share that google doctor I'm saying okay. This is where we are. At least that was my thinking there.

G

That's great that that's that's makes a lot of sense.

A

So I really appreciate that they were really super, open-minded sense. Okay, what do you need right? What kind of semantic conventions or whatever, um and then we probably just need to communicate? uh You know here would be our proposal.

A

uh There are a few there, but overall the the metrics are still um in in in hotel are still relatively moving targets, so I believe it's like q1 2020, one where the ga is planned, so we're we're still good, even with our timeline here.

G

Question question michael in your mind, is this: um is this more about compatible semantics and conventions or is, or is this also inclusive of potential in because the word instrumentation is mentioned in there.

A

So there are two parts right, the I would say the semantics are the the easier ones. This is essentially something where, as long as we find a consensus and say we consider abc as the things that we want to you.

E

A

Do or expose or or capture um the rest. There is still this. You know on the wire like what happens with open metrics will metrics be adopted by hotel metrics as the wire format, et cetera, et cetera, where I think we from the smi perspective, we might have an and you know um an opinion on that or an input, but we don't really own that right.

A

um I guess it's more like what is the like working backwards from what would be the idle situation, someone um using hotel as the the standard is able to instrument an smi compliant mesh and gets out everything you know um compliant to to the standard. That would be the idle case. Right um and semantics to me are the kind of like low-hanging fruits right that that's something that doesn't cost a lot right. Everyone is like fine, it's just a convention and the rest is um essentially up to everyone to implement.

A

Then right, if you are buying into that, if you think like yes, I'm gonna, you know write an exporter, I'm gonna write something in the sdk or whatever in hotel. Then you know that's up to you, um but I think that's what we, I think for our smi, implementers or or people who actually provide, as my compliant meshes, want to provide us a guidance or give us a guidance. I guess at least that's that's the way. I see it, I'm not sure if everyone agrees with that.

G

It's helpful that well thank you for thinking about that out that out loud uh um and then I'll progress with another question, though, about where you can do the same thing, which is uh um in your mind, would this be? Do you consider that this might be additive? The like an additive to the existing metric spec, in which it's optional for.

G

Service mesh implementations that are compatible with smi to sort of choose between open telemet, open, telemetry, compatible metrics versus the the metrics that are there today or the formatting of. What's there today. That's a great question.

A

I I'm unsure, but I I would say that open telemetry, given that it's you know it's an emerging standard and so far, essentially the only part that is going ga by end of year is traces right. Both metrics and logs are kind of like still work in progress coming next year.

A

I do believe that you know looking at different specifications or future standards within cncf. It would make a lot of sense. If you know we, we support each other right like if, unless there's a really good reason, why? Why should we not? um But I also see that kind of like backwards, competition compatibility?

A

That might be an issue right, so I don't really have a strong opinion that other than I would wish everyone going forward would simply commit to hotel and say like. Oh, we we gonna, implement that standard, but I think we can only give a recommendation there and saying like if you decide to adopt hotel in your implementation, then here is you know, for example, the semantics that that we suggest to use, rather than coming up with your own. I don't think that we can like make that a requirement or force anyone.

A

G

One thing that'll be informative in this, in that thinking is understanding how many of the implement the service mesh implementations today are in fact compatible with smi metrics. It conformed to this spec, and so so yeah. So.

A

Do you think we should take an action item to to research that to have a closer look at that or what's your.

G

I do well yeah well or uh yes- and I was sort of saying it, tongue-in-cheek, with respect to uh the smi conformance effort to go over and verify that also like oh yeah. That might.

A

Help I think that is super useful if you could, if I don't know if it extends your scope or if you see that anyways there, but that would be awesome absolutely.

A

I think that would be super useful for everyone.

G

Or just yeah informative toward, like hey, is backwards, compatibility really or how big of an issue is it does how compatible our meshes today exactly yeah.

B

G

You know sorry go ahead.

B

Oh, so I sorry, I would say that we have some freedom here being alpha and it is a moving target. um There is a real nice story to tell to align with open telemetry, um and I think you're going to find some freedom here, because I think a lot of data planes are going to be decoupled from the presentation of smi metrics versus open telemetry anyway. So to. I think this is more of a presentation problem rather than any like data sync or data sourcing problem.

B

So we should have some freedom to to move between either implementation, where many data plans probably have the freedom to move between any implementation. So I shouldn't see us saying it's mutually exclusive, but I should I would like to see a should to say collapse on open telemetry. If that's right,.

A

And I agree and if I could also ask you to comment on that issue 199 uh and in addition to the google docs, that's mentioned there to the smi metrics, essentially sharing where you are, where you know you're thinking there is given given where you are. You.

B

A

B

A

uh The the issue is 199 on our side. I think the tracking it differently in optimal metrics. But if you comment there uh justin from their end is the liaison and then I'm trying to facilitate things from our end and um I'm not sure I think he's from lightstep, I'm not entirely sure um but yeah this and then the google docs. That's um in the previous agenda item that trump pointed out. The google docs around smi metrics feedback.

A

Cool, so we have seven minutes left and I will shut up what now other than moderating and handing over to lee for the smi conformance work streams. So.

G

So last time we met, we touched on. We touched on the topic of smi conformance, which is, uh which is an effort to verify uh that I give an implementation, um adheres to the specs to the smi specs and there's kind of a lot in there.

G

It's a fairly well advanced effort that is functional and and performs some tests the it needs to perform. Probably or uh um let me step back and say so, there's a lot to discuss. Last time we left michelle had suggested that we, you know that we call a meeting and- and I think probably her ex- that we call like a recurring work stream um to kind of advance the discussions and give some dedicated time to it. And uh for my part, I thought that I would have done that by now.

G

I thought we would have already had that meeting um since it hasn't happened yet um and there's a few things that are happening within the service mesh working group. I was trying to find a spot on the cncf calendar to host that call, or that discussion and thought and thought of leveraging the service mesh working group to help advance it. um Do you so two questions? The smi metrics discussion that was held two three weeks ago was that how was that scheduled? Was that done through it.

D

Was in this time slot, but just on an off week, oh I see.

G

Got it okay that might actually work yeah? I should yeah okay.

D

You want to do it every other week. I mean this. Isn't a one-time meeting, I'm just trying to be clear about what you're.

G

Doing you're right, it won't be um for a couple of reasons. um One is assuming that you know a healthy number of the implementers actually participate. It'll take each implementer or each service mesh a little bit of time to yeah. The short answer is yeah, we'll need to meet a few times. That's every other.

D

I put a note on about your proposed time just because um friday, 1 pm central 11 am pacific is definitely night time on a friday in europe, and I guess I'll, ask michael to kind of speak to and staphon to speak to how useful that time. Slot is for them. But that's kind of my concern about fridays.

A

Yeah, I mean for me it would still kind of work for steven is like two hours ahead of me. So that's even later, maybe not the best time. Yeah.

G

I don't think it will work for me on fridays, the the wednesday every other wednesday, like this same time, slot, except for next week with kubecon, um and actually I would okay.

D

So you want to just put that on the smi calendar as the conformance working group.

F

D

And you want to start it the december whatever it would be. uh The first one I guess would be december 2nd. Is that what you would prefer.

G

Well, no, because that's another, like um uh it's unfortunate that like we, I wanted the initiative to be done this year to be over with it.

G

um So I guess I should ask, since we have three minutes so we've got a.

G

So meeting this week is pr anybody available to meet tomorrow. It's not, it doesn't necessarily need to be everybody I mean, but so not next week. That means the following week: we're going to have this meeting.

D

B

No, the following week we're.

D

Going to have u.s thanksgiving, we already decided we're not going to have this meeting. So if you want to hold a a different meeting on the 25th of november, I'm not going to stop you, but I'm also not going to come to it.

A

We don't do thanksgiving here in europe. I would certainly be up for it if you want to add a life if who else.

D

Awesome, so if you want to do if you, if michael do you want to run a meeting in this time, slot just uh run the meeting on wednesday, the 25th.

A

We don't need any specific things for the kicking off the scene that just works on its own right. We don't need some magic moderator, access or whatever.

D

You have, I can sync with you after this and make sure you have the the mod code cool.

A

And let's pencil it in and.

D

Nice all right.

A

Great cool all right, perfect thanks a lot that.

D

A

D

Minutes to bike shed over who's, gonna, moderate and or take notes on december, 9th.

D

I see matthew raising his hand.

B

I feel you knew to moderate, I can um I can. I can do the notes.

D

Awesome, please give me the uh the google doc enabled email address that you just send it to me, so that I can add it so that you can take the notes, cool.

B

Can I just send you that.

D

Yeah, damn it to me on here and we have a mod staphon. You want to moderate on december 9th.

E

Sorry, I I don't know if I will be able to make it.

D

All right, uh then, I guess it'll be me.

E

D

A

D

You're in charge on the 25th.

A

Oh yeah yeah, true, okay, cool thanks a lot. Everyone see you.

E

Thank you, bye.

A

A