Open Telemetry Uncategorized, 11 Jan 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2023-01-11 meeting

Description

Open Telemetry Meeting 1's Personal Meeting Room

A

Foreign hi Mark.

B

uh Lalit said he would be a bit late, so uh we can wait a bit.

A

C

Highlight it hey hi, good evening.

C

Yes, sorry for being late, let me let me quickly share the screen.

D

C

So Tom Tom has paid me that he will be a bit late, we're gonna add his name and.

C

And I think sorry for uh last meetings, confusion I, think we didn't realize me and Tom that we had an often Microsoft and we won't be able to join, but thanks Assan and Mark for continuing the meeting. I could see the meeting notes so yeah thanks for that yeah I. Think, let's, let's start discussing, feel free to add any items you want to add in the agenda. I can see. Mark has few items there. I have added mine, yeah specs, release, Here Mark. Please go ahead.

B

Yeah, so I just noticed that the spec is preparing a release, so I just opened an issue for that to adjust the semantic convention when it is done as a as a way to remember doing it.

C

C

Yeah, you can see lots of semantic conventions, breaking changes which should be fine. Parasitic.

C

Okay, uh yeah I, just added few of the GitHub issues and the PRS, which I probably felt that could be, would be interesting for us to discuss.

C

uh One is just wanted to highlight this that still there are changes happening in the log CPI uh kind of splitting between logs and events API in that's the reason I was delaying the the last PR which event has raised. So this is this is kind of again splitting. uh We do have separate logs in even API, but I. Think at the SDK level we will be still using. The log.

C

Sdk kind of I was delaying the last PR, which event has raised because that's PR is very specific to uncpi, so I was kind of waiting for the changes which will happen here, but I think Owen is saying that he's okay. If, if we go away for the review and I think he will afterwards, he will do those changes. Any changes which are happening as part of this.

C

This issue, so yeah just just wanted to highlight that and let nothing very substantial change, because events API is a very small API Delta change on top of logs API but yeah good to bring it.

C

uh The second one is specify the default Matrix view attributes in the API uh yeah. This is something which is uh there have been discussion happening on the hints. Api I mean I. um Hopefully, if anybody have seen that discussion uh as of now most of the configurations uh or not just metric or even for the traces and logs uh is happening at the SDK level, but there may be the cases where API or the instrumentation Library may want to give some recommendation or the hints in terms of configuration.

C

uh One of the examples could be uh the instrument for the histogram aggregation.

C

The instrumentation may want to recommend the range of the uh default instrument range of the histograms Windows to use a SDK may or may not adhere to that or may or may not use that, but that's kind of providing a hint that this is something which would be good if the SDK is using that configuration so that there's a hint API which has been in discussion for long and probably in continuation to that uh some more issues have been raised to provide say, provide a default metric view attributes in the API like as of now in The Matrix SDK.

C

We can filter out the attributes which uh the SDK fields are not really important to be sent to the exporters as part of the view configuration, but if the instrumentation Library does not really want that weapon. That's something uh probably the this is a issue which is in discussion for that, so yeah I mean I.

C

Don't think this is going to happen anything soon, but this probably I think there have been discussion happening on the hints, API and related uh changes required for that uh open Telemetry project, Road, Mac I, think I already shared that, but yeah feel free to have. When you have reviewed this and uh put your comments, I mean as a as an open, Telemetry com contributors. I think we all can give our comments on this.

C

So yeah feel free to provide that just wanted to highlight this. Okay, yeah I think we can start discussing with the more important topic which I feel probably duplication. uh uh Yeah marketing feel free to discuss on this control.

B

So, first of all, we know that there are things that we will deprecate, Like, Jaeger or maybe drop in C, plus plus 11 things like that. So, while we do that before asking ourselves every time again and again, how do we do that or how much time we we give users? How do we? How do we remove an API without breaking things?

B

um I would like to try proposing a uh some guideline and a process to do that, so um so whether at least, we have something to to detail how we will do things and also to communicate to users what they should expect yeah.

B

So if we say we duplicate things, maybe we will make a different document listing all the everything which is deprecated so that we know what not to use things like that. Okay,.

D

B

So I I raised the pr and I'm proposing to write the write, the process in the dock and when we review the pr we can review and discuss the process and see what we decide. Then.

C

Yeah sure I think that would that's a really good idea, I think to have some guidelines for us as of now I think we haven't really done much of the deprecation, but now I think there are a couple of things couple of bigger things which probably we have to do it so I think any guidelines I think Mark. If you can propose and probably we can discuss it, that puts it's. That's a welcome thing.

C

And Jagger uh okay Jager is has already deprecated I mean they are already. They started using otlp right and.

B

So Jaeger is using hlp and so Jaeger, so the Jager Library itself is being deprecated in a girl and in respect. The pr to deprecate. Jaeger is more just to respect itself, so in respect is, is already deprecated as well. So the question is how much time we keep the the Jaeger exporter.

E

B

The protocol, because it's now totally redundant with with their people at work.

D

C

I think it's uh probably just like for C plus 11 I. Think it's good good.

C

uh Probably if we start with the documentation uh start here and let people know that we are soon going to replicate it and then give some time, at least for people to be prepared of that. Yes,.

B

So then, and if we announce that, then it would be a way to to see who is using it and who depends on what and if people are are saying, hey we need more time. A lot of things happens.

C

And Jagger is going to remove it the support altogether or they they will.

B

Keep it possible? No, it's! uh It's not even reported yet so if there is a crushing bug or a security bug,.

C

Okay, yeah I think it's good, probably Let's. uh Okay I haven't seen the timelines at jagat site till till what time they will keep it without security patches till what then they will keep it active, uh so yeah, but I think from our side uh and I think asan. You have already have this uh issue. I mean assigned to yourself right.

A

C

Yeah, probably I think it's good to at least to start with good good to uh document. It I'm not sure whether we should document it in the I, don't think it should be in the readme, uh probably inside the Jagger.

C

All right. Let's see.

C

Yeah, probably somewhere here at least we can put as a note as a warning or something that it soon would be. uh Deprecated.

C

Because this complete for I mean if this correct me I think we duplicate means we are going to remove this complete directory all together. Yes,.

B

Yeah well not when we deprecate it, but when we literally.

C

When we eventually yeah sorry, eventually, we are going to remove it. Yes, yeah so I think it's good good to probably put it here and then, let's, let's see uh the response on this issue. If somebody has any concerns, let's let them raise the issue in that and I think that, let's, let's discuss in subsequent meetings, uh how much time we need to give.

C

Yeah: okay, thanks thanks Mark for both this proposal, that is Raising those issues for both of them: yeah drop, C, plus plus 11 support.

B

So reciprocess 11 supports back in December. We decided to discuss that in January, based on the feedback from people and I'm, not aware of any any comments. Oh.

C

Yeah, so the only comment I saw was that Jagger is not Jagger. I. Think correct me, Hassan I think you.

A

From tears, images.

C

They will continue supporting C plus 11. um That's I mean that should not be something which should affect our decision.

C

uh I mean for us, it's I, think maintaining C, plus plus 11, and the CIA and I think uh that that's that's more of a maintenance over it. Supporting supporting uh the Legacy compilers in terms of GCC 4.8 is also causing lots of issues for us, so yeah. It's just that, probably when when should we drop it, uh and when did we raise the issue? I think it's good to understand how much time we have.

A

It's first of December.

C

Yeah first of December, yeah I'm, not sure, probably should we give couple of months for people and then probably around 1st of Feb or something if we say.

B

That we are not going to stop it. At least we should announce, maybe for these notes or somewhere over twins.

B

We are deprecating it and see. Yeah.

C

And for us, what? What does it mean for the for deprecation like we'll remove it from the CI.

A

Yeah, we will remove it from CI.

C

Or anything else, we will do immediately, I mean when we say that we don't support I, think that that's the.

A

uh I think we don't need to proactively remove stuff that works.

D

B

Operations there is nothing to remove it's only. We will have less combination of compilers to yeah.

C

B

Probably GCC.

C

4.8 is something definitely will remove it. Yes,.

B

um If we already want to, uh but in the longer term we can today, we have workarounds with no STD shared pointers as opposed to STD, shot pointers, and this is to for compilers that would not even have a c plus 11 Library.

B

So in the long term we may consider, for example, to to change all the shared pointers to be STD and nothing else. um But I don't see the need for now, because we can't just have a type difference.

C

Oh, will it be a good idea to change it to STD, because that will that can break our Avi compliance right.

B

Well, it will give you collection of the code, but if we do that, then we already forced people to have stdm.

C

B

So these are basically what I'm I will suggest in the in the process. Is that every time we have a remove or something which is removed, we have to mitigate a transition. Yeah give guidelines to people like okay. How do I uh transition from the whole thing to a new thing, and you know in a way which is um not not too inconvenient.

C

B

So in this case, we we can keep no STD for a while. It doesn't cost much yeah.

C

Yeah I think probably it's a good idea to at least I mean if we all agree, hey Tom. Is there or not? Let me see okay, he has not yet joined, but I think if we all agree that it's we haven't seen much of the opposition from the users of continuous supporting C plus plus 11 ovent has already said that yeah. He is fine. If.

B

We he's fine as long as you can build within his own package, I'm assuming that means keeping no STD. For example, yeah.

C

C

I mean if, if we all agree in that case, then probably it's good to announce in the next release uh that we are definitely going to drop. The C plus plus 11 support sure.

A

B

Okay, I I will do it here. For that then.

E

Okay, yeah thanks.

C

Your Mark dearest version and Cipher, you know TLP HTTP, you haven't.

B

Yes, so this is my new toy uh yeah, so I implemented uh basically SSL in the HTTP exporter, okay, and uh so we've heard and and and so on, and also a long time ago, in back in November, um uh sir Zoro uh asks for basically minimum and maximum maintenance version, as well as ciphers yeah. So we filed uh a PR on this pack itself, which got no attention whatsoever.

B

Yeah and uh basically I started to uh uh file a PR in the spec, with something often to see if it can get an interaction, and if people can have interest in that and in the meantime, I also started to implement that in a in a PR- and you know you know uh open Telemetry, CPP.

C

And that's that would be separate VR for this right, not not as part of the current SSL prior right, I.

B

Mean um to discuss today I'm doing that in the same code, in the same rights for testing. How.

C

B

This is a true discussion: okay,.

C

Got it here, yeah I, think this looks good I I did review this I had a glance on this I think it looks good to me.

B

um One thing Andrew sees today is that it's very open, Telemetry collector, which is implemented in go already. Exports I mean it. It exposes minerals, maxillus already, even okay, it's not in respect and Jager of a product which implements also the hotel. Peak connector uh also expose minion Max theories, and, and it also expose ciphers.

C

B

C

Saying you're saying you're saying that yeah, the uh The Collector and the Jager exporter both support, Min and Max TLS versions right.

B

C

uh Okay and then we have Zipkin exporter, I, don't know what what exactly that does.

B

uh I didn't look at that, but.

E

B

Don't even know if zipkey is using careless at all I'm, not sure. Oh.

C

They should be using it right now. I mean, if not TLS, what exactly what else they would be using for.

B

B

Basically, I'm mostly working on uh internally I'm, mostly.

E

D

B

Tntics portal, so.

E

It's okay, yeah.

B

C

I think hopefully, hopefully, by raising this PR, will have more traction on this. This feature, I can I mean it looks good to me. I can probably add some comments and even approve it, but I don't think my approval really works here uh yeah. So it's basically a technical committee. I can check with really if you can, if he can look into that, I mean he's from Microsoft. So I can. Let me see if he can do.

B

Review it- and this is also for me, a way to test a PR on respect to receive a process and how it goes. Okay,.

D

C

Okay, yeah Otep internal SDK, metrics, yeah I, didn't have chance to look into that, but.

B

Okay, so the story is like that: um I'm using open television, uh C plus plus, obviously, and when something is exported where it may or may not end up in the open, Telemetry corrector.

D

B

Sometimes say if your endpoint is not even configured correctly, uh the exporter will not correct connect to anything and the traces will be dropped but externally. So we print something in a in a log file saying we cannot connect to the do it on stream connector, but we don't have a way to show and to know if events are dropped or not so I'm looking to have something like that, so that I can I can know if my application and my configuration is working correctly.

B

Okay to do that, there is very open to just Implement a new feature in opportunity.

B

Other people in other languages might have the same needs where do I have to do not dip for that.

C

Okay, so did this I mean I mean if I understand this? uh Basically Otep uh PR would be not specific to C, plus plus it's more of a generic uh recommendation.

B

Right, yes, so it will be a general recommendation.

C

B

So uh say, for example, in the trace exporter, if you drop a trace for for whatever reason you would increment to metric to count the trace lost, what metric will be exported using the metric uh exporter to something.

D

C

It's something similar to uh trying to use I mean if it's, if supposed to trace export, is failed. We can, as of now plug in uh the hotel, open Telemetry, logs, SD API and use that to send to log the error messages, so something similar in so I mean so just like logging error message using open, Telemetry log API, we will use Matrix API too, send the metrics.

B

C

B

Yeah, so I've basically have two set of questions like that. First of all, whether the feature makes sense or not in general, and the second one is process wise how to do that as a no type.

B

If it's just a matter of writing everything and and submitting a PR in realtor repository or if it involves more meetings and discussion with people up front and on.

D

C

I think it would be good to join the specification meeting sometime to discuss this I mean there, for who at least, would be good if.

E

C

Specifications you can join and probably discuss this whole tip.

B

C

You'll get more more idea from the actual uh technical committee and the governance committee, at least they can help you out to really uh I understand more about the process and whether it's the right approach, whether it's the right feature or not,.

B

Yeah thanks good idea.

C

Yeah one thing was merge policy I just wanted to talk about that I mean uh with C plus plus I, think it's we we do have a constant in terms of the number of uh maintainers I mean as compared to the other six number of maintainers and the approvers uh I and I just solely wanted to talk about to ensure that I mean as an approvers and the maintainers. We should not have lots of load.

C

I mean lots of uh load and pressure to really review all everything which is coming into the place and also to ensure that we are not losing the quality of the quality of the pr reviews. uh I mean probably I can.

B

So is it a concern with the merge itself of the review.

C

Not much into the review like right now, we we do have some constraints, I mean if I am just trying to I'm opening the contribution uh documentation to say I mean as of now. We have I. Think I just have concern with one of the uh point which we have uh so normally we say the pr is considered to be ready to merge when it has received more than one more than two approvals from the core owners, which is fine. Probably we can.

C

We can add another point that for more of a cosmetic uh PRS or a small, PR, I think even the one approver one, the one approval is fine or okay. ah We can. We can talk about what exactly the small PR in terms of the number of lines or the number of files, and the second thing was if the pi. If- uh and there was another thing right.

C

Oh I'm just trying to see where it was uh from different company. We also say.

B

Well, I think it's for maintainers, but we should have maintenance from different companies. Yeah.

C

If everybody is from only one, not maintainers, but it's a approval or from only one company, they want count uh yeah. This is this.

C

uh This is this is something probably, uh and we also have another thing right: uh the if, if one of the approverses or maintainer is raising a PR, it can be in only one approval is required.

B

Well, that makes sense, because.

E

B

Maintainer I mean when you submit something yeah. It means it has been reviewed by yourself at the minimum.

C

So yeah yeah, okay, yeah! Let me let me check that uh so I mean my proposal was that if more than two approve we can at least remove this constraint from only one company, because we right now, we we only have I mean more of the four active reviewers and I'll say in uh two of us from one company.

C

uh I I really don't want to add this further constraint. That actually means that we only have three approvals right now, two in total, if um uh I mean, if we only count one of the approvers from from me and Tom, if only one approval is going to be counted so that that's not solely I mean I I again, I don't want that. This should be. This should look like that.

C

Somebody from Microsoft is trying to is trying to use use their their I mean uh maintenance power to really merge all the peers. I. Definitely don't want to do that and it won't it's not going to happen. It's just want to don't want to delay lots of PR's and also don't want that. We all have to review all the PRS all the time uh as a maintenance and I think we as a maintenance and approvers.

C

We already know I mean the quality of the quality of the pr which we are raising so I think, even if um we have one review, irrespective of the company I, think that should be also fine.

C

uh So just wanted to check your review, I mean what what.

B

Is so what you're looking at is the guideline from respect, so we it applies to every project, not only open Telemetry, no.

C

So specs does not have any guideline, it's basically, we as in maintainers, probably we uh I mean in agreement with the approvals. We have to read our own guidelines so in if I see in the JavaScript, they are going to change the guideline that they need on, because they also have right now, shortage of number of approvers, so they're just going to have only one approved one approval is enough for them, irrespective of the company, so we still have two approvals, but for them I think it's it's they're just going to keep fund approval.

C

B

uh So well so there are many different parts. First of all, uh it's only five of us. Basically, so we don't have a lot of time to spend on all the reviews and so.

D

B

um I still like the fact that we have two up overs with other set from the same company or not it's a different story. Yes, because at least first of all, myself had not I, don't know all the codes. So if I propose PR, it's, it means I, think it's okay to go, but someone else might might know that he doesn't apply to this in that place and whatnot. So it's still.

B

uh We still need review from from someone else anyway, yeah and it's also I think a good way to not only do the do the review, but also learn the code from different places. Okay,.

D

B

Do different tributes in different things, so it collectively, if we do two reviews for one PR I mean if we have two reviewers for for one PR. It offers forces us to know more and more code. Okay, so we see that as a good thing.

C

Okay, so you're saying that let's keep two approvers, irrespective of who is going to raise a PR, uh we'll keep two. We can remove the company wow.

A

um Maybe you can add something here: uh we make like an exception. If those are from the same company, we wait. I, don't know two days three days more. If there is no objection, then then it's good to.

C

Me, okay, okay, uh okay, I got it yeah, so so here I mean well.

B

I mean in parties, even if there is an objection, we can always find a PL of two driver to change later. Exactly.

C

C

So so I mean my proposal. Was that uh we'll keep two approvals at least.

C

um We can remove it from the same company uh here. uh Yeah I mean like if, if any for any of the pr I think this, this should be okay for it to at least have two approvals, uh even if, like suppose, Sr is raising a PR and suppose me and Tom has done the the review of that and we will have done approval I think it should be good to go. Oh.

B

And it's it's perfectly fine by me.

C

Yeah, that's perfectly fine and um my only concern was like for some of the cases like for Matrix PR's, no Sr, somebody who has done lots of open metrics along with me. We both are the uh have been doing the reviews a lot more most of the time.

C

Sometimes if I am busy or if, uh if Sr is not there, if I'm not there, then we are kind of stuck here uh for the in those scenarios yeah so that that's basically uh I have a concern and it should not put nscc pressure to either of us to really have to spend some time to review, even if we are busy on something else. uh So that's okay, that that, because of that, I wanted to add another thing like if the.

C

If the pr is raised by any of the eight uh any of the approvers or maintainers, then it's good to go even with the one approval. But probably we can wait. As Assad was saying, we can wait for a few days for three four days to give time to others. It should not happen that uh I raise the pr next day. Tom has approved, approved it and same day, I merge it. That should not happen, so it.

B

Will definitely give some time yeah. Basically, everyone should have a chance to take a look at.

C

Something yes exactly and comment.

B

No, of course, if, if your PR has been sitting there for a long time and uh any historic, we also need to be to be moving so I I would say something like if, after a week of PR, is only one apple War. As long as there are no negative comments, it can be merged.

C

Yeah, okay, let me raise a PR on that, uh for that PR will still use the guideline which is mentioned here and and let me raise a PR and then probably I think, let's discuss it. Any changes required on that I mean probably we can uh do it there.

C

B

Yeah I think in general it's it's, it's still open for interpretation. Also I mean for things trivial. Like a bill break. We know that we can. If we have a very break affecting CI, we can do a fix and push it right away, and we have done that yeah. So comments can still apply in all cases.

C

Okay, uh apart from that, you just wanted to say that 16th Jan is a holiday in U.S, so.

B

Is it a Martin Luther King day.

C

I, don't remember yes, Martin Luther, yeah, okay,.

B

So so should we cancel that and.

C

uh Yeah I mean I just wanted to check with you guys if you want to join I, think that's totally fine. Otherwise we can cancel it.

B

Well, let's cancel because it's.

C

B

C

Yeah so good point here, thank you, okay, uh I think for the next release. uh We already are one month more now more than one month on the previous release and I think it's I mean I felt that probably it's a good time to have another release.

E

B

So um if we do uh well, we can do that in one or two weeks, uh basically, okay, to wait for the spec to to release okay, 17 and.

C

B

Semantic conventions to be updated.

C

B

um In the meantime, I'm hoping to do more cement, cleanup.

E

B

If I, if we have more semi cleanup industry, is fine, otherwise it will just wait longer.

E

F

Should we create a milestone for the new release.

B

um I'm not sure do we have feet well, we have a change log that contains the list of everything which is majorities I'm, not aware of any major features, but we will deliver also. It will be mostly uh 1.8.2 release or something like that.

C

Yeah I think it's not I, don't think any major feature has would be part of this, so 1.8.2 would be.

C

Probably in couple of weeks we'll see if something breaking has come, then probably we may change the decision yeah. But so Mark is saying that let's wait for a couple of weeks or well.

B

Yeah at least wait for the spec release itself: okay,.

C

B

That we are aligned.

C

Okay sure yeah, okay, I.

B

Mean it's a proposal. I mean.

C

Now, that's totally fine I think, probably let's go for. Logically, let's, let's incorporate specs release changes and that's it. Let's, let's be in sync, with the specs release: okay,.

E

C

uh I just wanted to bring. Today there was a there was a slack-ish slack uh message: the build time for hotel, CTP I haven't seen that. But if it's something similar to the build time which we have in our own CI I think which is, we can't do much in that case, but yeah, that's I, think it was raised yesterday so yeah. If somebody want to really have a look into that, I I won't prioritize, it uh asan did had some.

A

Some comments, yeah yeah, just give me an image registry and I'll, make it faster. Oh.

C

Yeah, that's what you did. You have been talking about: caching, the rebuilt images and something and probably or or the docker HD, or something so yeah yeah, but.

B

That would be the best I mean there won't see eye anyway, I'm assuming.

A

They can use the same also.

F

Yeah I think the the task which takes the most time years they actually be building grpc. Other libraries should not be that that is significant yeah.

F

Yeah also wondering like some other Library, which depends on grpc. How do they improve the view time instead of building from scratch.

B

So is there any way to install pre-packaged libraries instead of building job easy form Source, all the time.

A

um It is possible.

C

I mean there are lots of ways we can. We can rebuild it uh for the platform for which we are doing a CI build, and then we can keep it in some storage and probably pull it from there and then use it.

C

So there are multiple ways of caching it and then use it from there.

C

I would yeah I think somebody has really had to do some work on that. Try to understand how I try to test whether we can do that.

C

Yeah: okay: uh let's quickly go to the PRS and.

C

We'll start with the issues, I think quickly won't spend much time you upgrade to version. We already discussed this yes, yeah I think this is you have already okay, yeah, you so you'll, be remarked, you'll be raising a PR for this like I. Think it's a good good cleanup, uh along with uh Json yeah.

B

So I've done one for.

C

B

I'm hoping to do the same one for her yeah.

C

How to remove a active span and top implicit child spans for a code path? I, don't know why this request is coming. um I really want to understand how they are doing it. I'm still still struggling to understand their complete code. I know they have multi-threaded application, but uh even sarjo is also asking something similar, and this guy is also asking something similar.

C

And it's kind of cleaning up the uh the current context, the thread local storage of the context so that they can start creating uh creating a new span um and they can just ignore all the currently existing paths.

B

Somehow of a feeling that we we don't use uh API the way it is entered intended to be used so exactly.

C

B

So respond I mean even, if afraid is suspended, the spun from the thread is not suspended to span it starts and intense, and it throws nothing else in the middle storage.

B

uh Typically, in my understanding, if you have a function that calls another function in your application, uh if you don't want to pass a span pointer out of our place in all your code, you put that in the federal core storage once and then you you can look at that stage to create the nested's.

B

The child's plan, uh in their case, if you are using IF, are implementing a scheduler or something that picks up a task from the work queue or something else, and if we want to change the power and span, then we should not use federal storage at all. We should we should know what is the parent and where it's coming from so.

D

B

The real idea of creating a scope to why drag things.

C

And to try to change them.

B

Is in my opinion, spoken if they should not be even doing that in the first place, yeah.

C

Exactly I think in this scenario I totally agree. They should not be using the scope, uh so they should not be using our context, storage for the spans and probably they should be maintaining. They should be creating these uh spans and directly specifying the parent span as option yes and yeah. So I think that that's something I think even I wanted to propose, probably in their scenario they should not be doing. This is not the right approach to do it, uh but yeah. Let's probably I, think I'll propose something similar here. Yes, see.

B

So overall there might be more discussion, but I don't think we need a code change because.

C

We should not be doing it, yeah exactly I mean, and the text does not say anything about cleaning up a context. uh Not that not that that means that we cannot add it. We can I think to be specs compliant. We can add any extra methods, but we should not be removing removing any method which is not there in the specs uh but said that I think this does not look right approach.

C

The way they are trying to implement so yeah.

C

C

Okay, yeah, okay, your instance ID. Probably we are not supporting it. I think I'm, not sure. Probably we have to add.

B

I have seen from this specs that there is the concept of a resource detector, so.

D

B

If you are part of Docker or you detect resources, this way, if you're part of something else, you detect results in our way. So in this case, what we are missing, what seems to be missing is that. Well, we are part of a C plus plus process, but we don't know much about it and we can have a resource detector but basically connects with PID of a process and stick that as a service ID, but I.

A

B

What it is we can do.

C

Let me let me just check it quickly and I think what this was proposing.

C

Generate a service, okay, I think it's not related to the resource detector or, let me correct it's probably I felt that it is something um okay yeah. We do detect the resource here and then we try to.

C

Okay, yeah I mean so so this this is basically we do direct. We do use resource to detect the right set of attributes, but if, if if instant ID is missing, then we should be using some default instance. Id is that it means.

B

Proposed to uses uid, but yes, okay,.

C

Yeah so probably I think that is something which, if, if this, if this field is not coming through resource detected, then probably we have to give a. We have to use probably a default. If it's a uuid, we have to provide provide a uuid just like something we are doing it here. We use a unknown Service as a default service. If it is not coming, the service name is not part of it's not coming from resource detector. Then we should use a default value as a non-service, so probably something similar.

C

We have to do it, for instance, ID yeah, but I think, let's look more into that. I can assign it to myself.

B

I would just choose the process. Id yeah.

C

Okay, yeah uh process ID would not be uuid right. I, don't know, probably, which is I. I, have to read that sorry, I, don't want to comment much without reading it. Yeah yeah, okay, yeah and your Primitives example. Cash yeah I saw nothing. You were looking into that right.

A

Yeah I think it's the name yeah it's an exception, because the port is used by some other process: okay, an exception that is not handled. Yes,.

B

So the question then becomes if we start a payment use exporter and it fails to initialize, because report is in use.

B

If positives itself is raising an exception, we should catch it at some point, and my question is what happened next I mean? uh Is there a return code return to application so that it knows nothing will be ever exported? And what is the application supposed to do then? Because the in in general, my understanding is that whatever, whatever an exporter, does it's not supposed to take down the application and if we have an exception so.

E

A

He had good good question on this and I think this is the only place where we will throw exception right because of the from this exploiter itself.

C

And this is the bootstrapping of the SDK right, which will happen when the process is starting right. So even if process goes down, the application goes down. It should be still okay, because it's it's, the process is actually application is going down when, when it's really going to start yeah.

A

It's so nice. Can you.

C

Correct me yeah, so it's not something that we are going to crash when the application is running and it's getting some Matrix and suddenly we crash it. Yeah.

A

Something at the yeah, so it starts to listen on on some part. Yeah.

D

A

Prompt use exporter can extract data from it coming.

E

A

This collector.

D

C

So probably I think I'll say that if you, even if, as of now, the current behavior is that it caches, but it should be more of a graceful graceful, a graceful exit.

B

A brexit of what? um Because we create, we create an exporter and we add it to uh other things, to create a metrics provider and.

E

B

Metrics provider is not working.

D

B

So a question to an example, but the question becomes in a real application. How would you instrument with moment use.

A

What can we do? We can just log something that the exporter is not able to.

B

Show sure we can we could lock something to start with. uh One thing we may do is expose an API on the properties exporter itself to check if it's, uh if it's running correctly or not, so that in the main verification, for example, can set up the parameters, exporter or install it and then check if it started or not, and then the application May then decide if it wants to continue without metrics or not.

C

uh Mark, can you just add your recommendation here? Probably let's discuss it. So let me let me also start uh looking more into this. I mean and probably let me propose something also in this and then probably let's, let's finalize it here,.

D

B

um I did not look at it, but.

D

B

Says something about that? Because it's.

C

Yeah, let's yeah.

B

The fact that he trades on uh on a given part, if this would be.

C

B

Everyone, not every language.

C

Yeah, probably it would be good to look into other six, how they are handling it. That also would be interesting to see, and yes,.

C

We already discussed internalization, Matrix and Trace ID confusion during concurrency.

C

What was that yeah I think this is probably less thinking. That's pinging away again: I.

B

Would say that let's wait a week if there is any.

C

D

We can close.

E

C

Scope should be default, constructible.

B

Same issue, research duplicates.

D

B

And it's the same person also discussing this yeah.

C

Okay, I think, let me see the crash which he is seeing I, don't know why it's happening, but I'll respond to that. I think I already assigned it to myself.

C

And yeah this is already for logs. This is all the logs I just created can conflict on.

C

Okay, I haven't seen that, but probably I, assume or went was discussing with him.

C

Okay, yeah I'll, look into that also, and probably it could check with the event if you can assign it to him.

C

Please add example: how to use baggage I mean again. This is something similar asking to replace a runtime context.

C

uh I, don't know we should not be doing that, but yeah I'll get more clarity on the scenario why why do he need this.

C

Instagram, do not count correctly raise a PR for this I think this is also raise a PR for this, and this three I think I've raised up here.

C

Okay, I haven't seen that yet I think it's with you.

A

Yeah, just I have to how many issues to go through yeah.

C

A

Very slow, sorry.

C

No, no problem, I, think I, know I know we all are working on other stuffs, other things apart from open Telemetry. So no no excuse me.

B

D

B

From early December yeah.

B

Well, we have discussed all that uh yeah last time already. This.

D

B

C

I think these three also we discussed.

C

I think I signed it with you, so whenever you have time.

C

I think that that's all, probably that let's have been discussing for long, let's quickly see a full requests. We just have five minutes left.

B

So working progress just forget it for now.

C

Open television yeah I have to look into that. It will take some time to really uh go through this.

C

I did tag one of the guys who has worked earlier, both in open tracing and also he was the main contributor of open tracing and also this guy started hotel, hotel, C plus plus work um I.

D

Think he's not.

C

Active that much so probably if we have to review this.

B

I quickly looked so it seemed to make sense, but I, don't know anything about open tracing to tell if a specific line of code is broken or not to make sense.

C

Okay, I have to look into that.

C

Some aggregation support, yeah, there's something I. Think I should be able to make it ready uh by by a couple of days.

C

Bad shade lip difference.

C

Okay, I think Owen has been reviewing it and he has already requested some changes.

C

Picksemic to build yeah I know this. This is a valid fix. I, don't know what issues is he is facing further.

C

Okay, I think it's a former tissue right now, I think yeah. Okay, we will ping him uh if it's it's good. If we can just fix these things and then probably you can merge it.

F

And for for g-test I think I have a question like um seems like our our current report doesn't work with the latest Benchmark Google, Benchmark I think and there I got some build Arrow, but I didn't look more yeah because, um usually our views, our report specifies which Google Benchmark to use. But if I use like some password customized Visa package to to my hotel CPP build, then I got the latest Google Benchmark. So there's some build error.

B

Yeah, there's a visibility or NCI and according to the log, it's uh because of a bill break uh something in in G, test or Benchmark. I, don't remember requires a C plus plus 14., so we should compile the the CI using C plus 14 for for the benchmark.

D

F

So, okay, um let's see I, think we discussed this before. Do we do we need to upgrade our standard requirement or outline even when you don't do that now, that's I, think more and more dependencies will require C plus plus 14., as in grpc. We look at the letter before.

C

Yeah so I think we already discussed.

F

C

Plus plus 11 duplication, right yeah, we already discussed I think and uh we eventually do need to remove it.

F

We don't have a fixed plan right so that.

C

um We do have it, I mean so we're going to announce it in the next release and then probably will deprecate it uh have another month or something and then I think it's too good to deprecate. It.

D

Okay, first thing:.

E

D

Okay sounds good.

F

C

But this will work if we use the same uh correct me if we use the right Benchmark version, this should not come right. It's just that we are using the.

B

Lead so trying to use. We are using a recent Benchmark which requires C plus plus 14, but if you use an old one, C plus plus 11 should be just fine, I guess.

C

B

Well, as long as you don't care about warnings.

D

F

Like a Korean like for now make the fix at little brown blocks, we see build with Master with the package. Can we make a fix like for for The Benchmark, build we use C plus 14, but for for the main SDK we don't, or should that work or.

B

uh That should work if it's only Benchmark itself.

F

Yeah so far it is I only got the brick other other build works, fine with latest with the package or the native dependencies. So maybe okay I can try that one to fix.

C

F

C

I mean sorry to Observer, just correct me, I think for for most of the latest compilers the default would be more than C, plus plus 11 right the default build flag. It will not be C plus plus 11. It would be more than that, so they would be still building with C, plus 14 or maybe 17 or something.

B

Oh, but um in the make file, I think this is possible.

C

Do we hardcore it.

B

We well, where is this um uh with STL flag and if we invest here is set, we say we need 14 or 7 or 17, and what 17 14 and 20..

B

But if nothing is said, I think we false C plus 11. So.

E

C

We explicitly ask for.

B

C plus plus 11, when um but.

C

We don't explicitly ask for C, plus plus 11 right. We only ask it in GCC for 4.8 builds.

B

No, no I think we do that in the in.

A

C

B

Will see make oh.

C

B

Different compiler, but in the top level C make file I think we said.

C

B

E

F

B

For reverse here.

B

B

C

This is, this is okay. This is for with the steel right, but if it is not with the steel.

B

Yeah but she wears I mean 127.

D

B

If it's not for this year, then we said we we compiled with with 11.

C

B

You said a standard yourself got.

C

It yeah I probably missed this. This thing, okay,.

B

E

Why yeah thanks.

C

Okay, yeah and.

B

Yeah so anyway, the CIA for Benchmark should be fixed.

C

Okay- let's, let's probably be already time, log SDK and eBay- to follow the specification yeah I'll review it uh histogram aggregation yeah. This is basically Assan. I think this is with you, I mean if you have some time to see.

A

C

A

I'll do it this week, sorry, okay,.

C

Okay, no no problems here, I mean I. Don't want you to be really pressurized on that. If you're, okay, I can merge it and then probably you can look in later, I mean either way is okay, uh just just probably, if you can see it this week, I think it's totally fine. Yeah I mean it's just. It has some uh very trivial changes in the performance fix, It's, Kind kind of I think good good to go um sooner so probably yeah.

C

Whenever you have time I think this week, thanks uh six uh hey I haven't seen these I mean. What's the status right I mean, is it good to review it right now? I know there was some work you were doing for Basil yeah.

C

This is ready to review it now yeah, okay and then this can be closed right.

A

uh This is an example, I mean if we are going to deprecate, C, plus plus 11th, and we don't need.

C

This okay, okay, yeah, okay. uh What was that Implement total.

B

C

Is also ready for review right, Mark, yes, ah okay, yeah, let's set probably I'll review it.

C

And okay, yeah I think it's good to review it on this one and yeah. This is no change on this.

B

So for the last one uh I propose we close it uh because there is no activity at all, and on top of that, uh it did a force push on the pr like you are losing the old history and all the comments.

C

Okay, you're saying it's good to close it right so.

B

We should just close it. I mean if.

A

E

B

Open it we can but yeah.

A

B

Won't swipe uh all every time and be discussed every time on every meeting.

E

C

Okay, I think that's all we have so we won't meet next week, we'll probably let's meet next to next week and.

C

Thanks for joining have a good rest of the week.

B

Thanks everyone thanks: everyone.

F

You thank you. Thank you. All bye.

D